article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
in this paper we first recall some essential tools concerning the theory of differential forms in the _ cartan _ , _ hodge _ and _ clifford _ bundles over a @xmath1-dimensional manifold @xmath2 equipped with a metric tensor @xmath3 of arbitrary signature @xmath4 , @xmath5 and also equipped with metric compatible connections , the levi - civita ( @xmath6 ) and a general riemann - cartan ( @xmath7 ) one - dimensional , the metric has signature @xmath8 and which is equipped with a levi - civita or a a riemann - cartan connection , orietability and time orientation .
see below and , e.g. , @xcite for more details , if needed . ] .
after that we solved with details some exercises involving different grades of difficult , ranging depending on the readers knowledge from kindergarten , intermediate to advanced levels .
in particular we show how to express the _ _ derivative ( _ _ @xmath9 ) and _ coderivative _ ( @xmath10 ) operators as functions of operators related to the levi - civita or a riemann - cartan connection defined on a manifold , namely the _ standard _ dirac operator ( @xmath11 ) and general dirac operator ( @xmath12 ) .
those operators are then used to express maxwell equations in both a lorentzian and a riemann - cartan spacetime .
we recall also important formulas ( not well known as they deserve to be ) for the square of the general dirac and standard dirac operators showing their relation with the _ hodge _ _ dalembertian _ ( @xmath13 ) , the _ covariant d alembertian _ ( @xmath14 ) and
the _ ricci operators _ ( @xmath15 , @xmath16 ) and _ einstein _ operator ( @xmath17 ) and the use of these operators in the _ einstein - hilbert _ gravitational theory . finally , we study the _ bianchi _
identities . recalling that
the first bianchi identity is @xmath18 , where @xmath19 and @xmath20 are respectively the torsion and the curvature @xmath0-forms and @xmath21 is a cotetrad we ask the question : who is @xmath22 ?
we find the correct answer ( eq.([du6 ] ) ) using the tools introduced in previous sections of the paper .
our result shows explicitly that the formula for .@xmath23 recently found in @xcite and claimed to imply a contradiction in einstein - hilbert gravitational theory is _
wrong_. two very simple counterexamples contradicting the wrong formula for @xmath22 are presented .
a detailed account of the issues discussed in the paper appears in the table of contents .
we call also the reader attention that in the physical applications we use natural units for which the numerical values of @xmath24,@xmath25 and the gravitational constant @xmath26 ( appearing in einstein equations ) are equal to @xmath27 .
let @xmath2 denotes a @xmath1-dimensional manifold unspecified for a while . ] .
we denote as usual by @xmath29 and @xmath30 respectively the tangent and the cotangent spaces at @xmath31 .
by @xmath32 @xmath29 and @xmath33 @xmath34 respectively the tangent and cotangent bundles . by @xmath35
we denote the bundle of @xmath36-contravariant and @xmath37-covariant tensors and by @xmath38 the tensor bundle . by @xmath39 and @xmath40 denote respectively the bundles of @xmath36-multivector fields and of @xmath36-form fields .
we call @xmath41 the bundle of ( non homogeneous ) multivector fields and call @xmath42 the exterior algebra ( cartan ) bundle . of course
, it is the bundle of ( non homogeneous ) form fields .
recall that the real vector spaces are such that @xmath43 @xmath44 @xmath45 and @xmath46 .
some _ additional _ structures will be introduced or mentioned below when needed .
let the space of the sections of a bundle @xmath47 .
note that all functions and differential forms are supposed smooth , unless we explicitly say the contrary . ]
@xmath3 a metric of signature @xmath4 and @xmath7 an arbitrary metric compatible connection on @xmath2 , i.e. , @xmath48 .
we denote by @xmath49 and @xmath50 respectively the ( riemann ) curvature and torsion tensors of the connection @xmath7 , and recall that in general a given manifold given some additional conditions may admit many different metrics and many different connections . given a triple @xmath28 : \(a )
it is called a riemann-_cartan space _ if and only if @xmath51
\(b ) it is called _
_ weyl space if and only if__@xmath52 \(c ) it is called a _ riemann space _ if and only if @xmath53 and in that case the pair @xmath54 is called _
riemannian structure_. _ _ ( _ _ d ) it is called _ riemann - cartan - weyl space _ if and only if @xmath55
\(e ) it is called _ _ ( _ _ riemann _ ) _ flat if and only if @xmath56
\(f ) it is called teleparallel if and only if@xmath57 for each metric tensor defined on the manifold @xmath2 there exists one and only one connection in the conditions of eq.([888 ] ) .
it is is called _ levi - civita connection _ of the metric considered , and is denoted in what follows by @xmath6 . a connection satisfying the properties in ( a ) above
is called a riemann - cartan connection . in general
both connections may be defined in a given manifold and they are related by well established formulas recalled below . a connection defines a rule for the parallel transport of vectors ( more generally tensor fields ) in a manifold , something which is conventional @xcite , and so the question concerning which one is more important is according to our view meaningless .
the author knows that this assertion may surprise some readers , but he is sure that they will be convinced of its correctness after studying section 15 .
more on the subject in @xcite . for implementations of these ideas for the theory of gravitation see @xcite when @xmath58 and the metric @xmath59 has signature @xmath8 we sometimes substitute riemann by lorentz in the previous definitions _ ( c)_,_(e ) _ and _ ( f)_. in order to represent a spacetime structure a lorentzian or a riemann - cartan structure @xmath28 need be such that @xmath2 is connected and paracompact _
@xcite _ and equipped with an orientation defined by the volume element @xmath60 and a time orientation denoted by @xmath61 .
we omit here the details and ask to the interested reader to consult , e.g. , _ @xcite_. a general spacetime will be represented by a pentuple @xmath62
given a differentiable manifold @xmath2 , let @xmath63 , any vector fields , @xmath64 any covector field .
let @xmath65 be the tensor bundle of @xmath2 and @xmath66 any general tensor field .
we now describe the main properties of a general connection @xmath7 ( also called absolute differential operator ) .
we have@xmath67 where @xmath68 the covariant derivative in the direction of the vector field @xmath69 satisfy the following properties : given , differentiable functions @xmath70 , vector fields @xmath63 and
@xmath71 we have @xmath72 given @xmath73 the relation between @xmath74 , the _ absolute differential _ of @xmath75 and @xmath76 the covariant derivative of @xmath75 in the direction of the vector filed @xmath69 is given by @xmath77 let @xmath78 and consider a chart of the maximal atlas of @xmath2 covering @xmath79 coordinate functions @xmath80 .
let ` ` @xmath3 be a metric field for @xmath2 .
let @xmath81 be a basis for @xmath82 , @xmath78 and let @xmath83 be the dual basis of @xmath81 .
the reciprocal basis of @xmath84 is denoted @xmath85 , and ` g`@xmath86 @xmath87 .
introduce next a set of differentiable functions @xmath88 such that : @xmath89 it is trivial to verify the formulas @xmath90 with @xmath91{c}\eta_{\mathbf{ab}}=\mathrm{diag}(\underbrace{1, ...
,1}\underbrace{-1, ... -1})\\ \text { \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ } p\text { { \footnotesize times \ \ \ \ \ \ } } q\text { { \footnotesize times}}\end{array } . \label{eta}\ ] ] moreover , defining @xmath92 the set @xmath93 with @xmath94 is an orthonormal basis for @xmath82 .
the dual basis of @xmath82 is @xmath95 , with @xmath96 . also , @xmath97 is the reciprocal basis of @xmath95 , i.e. @xmath98 .
when @xmath58 the basis @xmath93 of @xmath82 is called a _ tetrad _ and the _ _
( _ _ dual _ ) _ basis @xmath99 of @xmath100 is called a cotetrad .
the names are appropriate ones if we recall the greek origin of the word .
the connection coefficients associated to the respective covariant derivatives in the respective basis will be denoted as : @xmath101 the connection coefficients of the levi - civita connection in a coordinate basis are called christoffel symbols .
we write in what follows@xmath102 to understood how @xmath7 works , consider its action , e.g. , on the sections of @xmath103 .
@xmath104 for every vector field @xmath105 and a covector field @xmath106
we have @xmath107 and using the properties of a covariant derivative operator introduced above , @xmath108 can be written as:@xmath109 where it is to be kept in mind that the symbol @xmath110 is a short notation for @xmath111 also , we have @xmath112 where it is to be kept in mind that and @xmath113 . what is important is always to have in mind the meaning of the symbols . ] that the symbol @xmath114 is a short notation for @xmath115 the necessity of precise notation becomes obvious when we calculate @xmath116 thus verifying that @xmath117 and that @xmath118 moreover , if we define the object @xmath119 which is clearly the identity endormorphism acting on sections of @xmath82 , we find@xmath120 some authors call @xmath121 _ _ ( _ _ a single object _ _ ) _ _ a tetrad , thus forgetting the greek meaning of that word
. we shall avoid this nomenclature .
moreover , _
eq.([tetrad ] ) _ is presented in many textbooks _ _ ( _ _ see , e.g. , _
@xcite ) _ and articles under the name ` tetrad postulate ' and it is said that the covariant derivative of the `` tetrad '' vanish .
it is obvious that _ eq.([tetrad ] ) _ it is not a postulate , it is a trivial _ _
( _ _ freshman _ ) _ identity . in those books , since authors do not distinguish clearly the derivative operators @xmath122 and @xmath7 , _
eq.([tetrad ] ) _ becomes sometimes misunderstood as meaning @xmath123 or @xmath124 , thus generating a big confusion and producing errors _ _ ( _ _ see below__)__.
we call in what follows hodge bundle the quadruple @xmath126 .
we now recall the meaning of the above symbols .
we suppose in what follows that any reader of this paper knows the meaning of the exterior product of form fields and its main properties .
we simply recall here that if @xmath127 , @xmath128 then @xmath129 let be @xmath130 @xmath131 , @xmath132 where @xmath133 @xmath134 \(i ) the scalar product @xmath135 is defined by @xmath136{lll}a_{1}\cdot b_{1 } & .... & a_{1}\cdot b_{r}\\ .......... & .... & .......... \\
a_{r}\cdot b_{1 } & .... & a_{r}\cdot b_{r}\end{array } \right\vert .
\label{scalarprod}\ ] ] where @xmath137 .
we agree that if @xmath138 , the scalar product is simple the ordinary product in the real field . also ,
if @xmath139 , then @xmath140 . finally , the scalar product is extended by linearity for all sections of @xmath141 . for @xmath142 @xmath143 @xmath144we define the _ left contraction _ by @xmath145 where @xmath146 is the reverse mapping ( _ reversion _ ) defined by @xmath147 and extended by linearity to all sections of @xmath148 .
we agree that for @xmath149 the contraction is the ordinary ( pointwise ) product in the real field and that if @xmath150 , @xmath151 , @xmath152 then @xmath153 .
left contraction is extended by linearity to all pairs of elements of sections of @xmath148 , i.e. , for @xmath154 @xmath155 where @xmath156 means the projection of @xmath157 in @xmath158 .
it is also necessary to introduce the operator of _ right contraction _ denoted by @xmath159 .
the definition is obtained from the one presenting the left contraction with the imposition that @xmath160 and taking into account that now if @xmath127 , @xmath161then @xmath162 .
the hodge star operator is the mapping @xmath164 where for @xmath165@xmath166\tau_{{\mbox{\bfseries\slshape g}}}=\mathcal{b}_{k}\wedge\star\mathcal{a}_{k},\text { } \forall\mathcal{b}_{k}\in\sec\bigwedge\nolimits^{k}t^{\ast}m \label{11a}\ ] ] @xmath167 is the _ metric volume element_. of course , the hodge star operator is naturally extended to an isomorphism @xmath168 by linearity .
the inverse @xmath169 of the hodge star operator is given by : @xmath170 where @xmath59 @xmath171 denotes the sign of the determinant of the matrix @xmath172 ` ` @xmath173 , where @xmath174 is an _
arbitrary _ basis of @xmath82 .
we can show that ( see , e.g. , @xcite ) that @xmath175 where as noted before , in this paper @xmath176 denotes the _ reverse _ of @xmath177 .
let @xmath178 be the dual basis of @xmath174 ( i.e. , it is a basis for @xmath179 ) then ` g`@xmath180 , with @xmath181 .
writing @xmath182 , @xmath183 we have from eq.([11b ] ) @xmath184 some identities ( used below ) involving the hodge star operator , the exterior product and contractions are ) which uses the clifford product . ] : @xmath91{l}a_{r}\wedge\star b_{s}=b_{s}\wedge\star a_{r};\quad
r = s\\ a_{r}\cdot\star b_{s}=b_{s}\cdot\star a_{r};\quad r+s = n\\ a_{r}\wedge\star b_{s}=(-1)^{r(s-1)}\star(\tilde{a}_{r}\lrcorner b_{s});\quad r\leq s\\ a_{r}\lrcorner\star b_{s}=(-1)^{rs}\star(\tilde{a}_{r}\wedge b_{s});\quad r+s\leq n\\ \star\tau_{{\mbox{\bfseries\slshape g}}}=\mathrm{sign}\text { } { \mbox{\bfseries\slshape g}};\quad\star1=\tau_{{\mbox{\bfseries\slshape g}}}. \end{array } \label{440new}\ ] ] the _ exterior derivative is a mapping _ @xmath185 satisfying : @xmath91{ll}\text{(i ) } & d(a+b)=da+db;\\ \text{(ii ) } & d(a\wedge b)=da\wedge b+\bar{a}\wedge db;\\ \text{(iii ) } & df(v)=v(f);\\ \text{(iv ) } & d^{2}=0 , \end{array } \label{ext deriv}\ ] ] for every @xmath186 , @xmath187 and @xmath188 .
the _ hodge codifferential _ operator in the hodge bundle is the mapping @xmath189 , given for homogeneous multiforms , by:@xmath190 where @xmath163 is the hodge star operator .
the operator @xmath10 extends by linearity to all @xmath191 the _ hodge laplacian ( or hodge dalembertian ) _ operator is the mapping @xmath192 given by : @xmath193 the exterior derivative , the hodge codifferential and the hodge d alembertian satisfy the relations : @xmath91{l}dd=\delta\delta=0;\quad\diamond=(d-\delta)^{2}\\ d\diamond=\diamond d;\quad\delta\diamond=\diamond\delta\\ \delta\star=(-1)^{r+1}\star d;\quad\star\delta=(-1)^{r}d\star\\ d\delta\star=\star\delta d;\quad\star d\delta=\delta d\star;\quad\star \diamond=\diamond\star{.}\end{array } \label{545}\ ] ]
let @xmath194 be a riemannian , lorentzian or riemann - cartan structure may be the levi - civita connection @xmath6 of @xmath59 or an arbitrary riemann - cartan connection @xmath7 . ] .
as before let @xmath195 be the metric on the cotangent bundle associated with @xmath3
. then @xmath196 , where @xmath197 is a vector space equipped with a scalar product @xmath198 of signature @xmath4 .
the clifford bundle of differential forms @xmath199 is the bundle of algebras , i.e. , @xmath200 , where @xmath201 , @xmath202 , a real clifford algebra . when the structure @xmath194 is part of a lorentzian or riemann - cartan spacetime @xmath203 the so called _ spacetime _ _ algebra_. recall also that @xmath204 is a vector bundle associated with the _ _ _ _ @xmath205-__orthonormal coframe bundle _ _ @xmath206 , i.e. , @xmath207 @xmath208 ( see more details in , e.g. , @xcite ) . for any @xmath31
, @xmath209 is a linear space over the real field @xmath210 .
moreover , @xmath211 is isomorphic as a real vector space to the cartan algebra @xmath212 of the cotangent space
. then , sections of @xmath207 can be represented as a _ sum _ of non homogeneous differential forms .
let now @xmath93 be an orthonormal basis for @xmath82 and @xmath99 its dual basis . then , ` g`@xmath213 the fundamental _ clifford product _
( in what follows to be denoted by juxtaposition of symbols ) is generated by@xmath214 and if @xmath215 we have @xmath216 where @xmath217 is the volume element and @xmath37 , @xmath218 , @xmath219 , @xmath220 , @xmath221 .
let @xmath222 . for @xmath223 , we define the _ scalar product _ as follows : for @xmath224@xmath225 we identify the _ exterior product _ ( @xmath226 of homogeneous forms ( already introduced above ) by @xmath227 where @xmath228 is the _ component _ in @xmath229 ( projection ) of the clifford field .
the exterior product is extended by linearity to all sections of @xmath207 . the scalar product , the left and the right are defined for homogeneous form fields that are sections of the clifford bundle in exactly the same way as in the hodge bundle and they are extended by linearity for all sections of @xmath207 .
in particular , for @xmath230 we have @xmath231 the main formulas used in the present paper can be obtained ( details may be found in @xcite ) from the following ones ( where @xmath232 ) : @xmath233 two other important identities to be used below are : @xmath234 for any @xmath235 and @xmath236 , and @xmath237 for any @xmath238 the dirac operator associated to a general riemann - cartan structure @xmath28 acting on sections of @xmath199 is the invariant first order differential operator @xmath240 for any @xmath241 we define @xmath242 recall that the _ reciprocal _ basis of @xmath21 is denoted @xmath244 with @xmath245 ( @xmath246 ) and that @xmath247 with @xmath248 , and @xmath249 @xmath250 .
defining @xmath251 we have ( by linearity ) that @xcite for any @xmath252 @xmath253 , \label{12nnn}\ ] ] where @xmath254 is the pfaff derivative , i.e. , for any @xmath255 it is : @xmath256\theta^{_{\mathbf{i}_{1}}} ...
\theta^{_{.\mathbf{i}_{p}\text { } } } . \label{pfaff}\ ] ] using eq.([12nnn ] ) we can show that for the case of a riemannian or lorentzian structure @xmath258 the standard dirac operator defined by:@xmath259 for any @xmath260 is such that @xmath261 i.e.,@xmath262
as we said in the beginning of section 1 a given structure @xmath263 may admit many different metric compatible connections .
let then @xmath6 be the levi - civita connection of @xmath59 and @xmath7 a riemann - cartan connection acting on the tensor fields defined on @xmath2 .
let @xmath78 and consider a chart of the maximal atlas of @xmath2 covering @xmath79 with arbitrary coordinates @xmath264 .
let @xmath81 be a basis for @xmath82 , @xmath78 and let @xmath265 be the dual basis of @xmath81 .
the reciprocal basis of @xmath84 is denoted @xmath84 , and ` g`@xmath266 @xmath267 .
let also @xmath93 be an orthonormal basis for @xmath268 with @xmath269.@xmath270the dual basis of @xmath82 is @xmath99 , with @xmath96 . also , @xmath97 is the reciprocal basis of @xmath99 , i.e. @xmath98 .
an arbitrary frame on @xmath271 , coordinate or orthonormal will be denote by @xmath174 .
its dual frame will be denoted by @xmath272 ( i.e. , @xmath273 ) .
the _ torsion and curvature operators _ @xmath274 and @xmath275 of a connection @xmath7 , are respectively the mappings:@xmath276,\label{top}\\ \mathbf{\rho(u , v ) } & = d_{\mathbf{u}}d_{\mathbf{v}}-d_{\mathbf{v}}d_{\mathbf{u}}-d_{[\mathbf{u , v } ] } , \label{cop}\ ] ] for every @xmath277 .
the _ torsion andcurvature _ tensors of a connection @xmath7 , are respectively the mappings : @xmath278 for every @xmath279 and @xmath280 .
we recall that for any differentiable functions @xmath281 and @xmath25 we have@xmath282 note that it is quite obvious that @xmath283 define the tensor field @xmath284 as the mapping such that for every @xmath285 and @xmath280 . @xmath286
it is quite ovious that @xmath287 where @xmath288 we now show that for any structure @xmath28 such that @xmath48 we have for @xmath289 , @xmath290 we start recalling that for every metric compatible connection it holds : @xmath291 exachanging @xmath292 in the last equation we get@xmath293 subtracting eq.([ri7 ] ) from eq.([ri8 ] ) we have @xmath294(c\cdot c)=}2([d_{\mathbf{u}},d_{\mathbf{v}}]\mathbf{c)\cdot c } \label{ri9}\ ] ] but since @xmath294(c\cdot c)=}d_{[\mathbf{u , v]}}(\mathbf{c\cdot c})=2(d_{[\mathbf{u , v]}}\mathbf{c)\cdot c , } \label{ri10}\ ] ] we have from eq.([ri9 ] ) that @xmath295\mathbf{c-}d_{[\mathbf{u , v]}}\mathbf{c)\cdot c}=0\text { , } \label{ri11}\ ] ] and it follows that @xmath296 as we wanted to show .
prove that for any metric compatible connection , @xmath297 given an arbitrary frame @xmath174 on @xmath271 , let @xmath272 be the _
dual frame_. we write : @xmath91{ccl}\lbrack e_{\alpha}\mathbf{,}e_{\beta } ] & = & c_{\alpha\beta}^{\rho}e_{\rho}\\ d_{e_{\alpha}}e_{\beta } & = & \mathbf{l}_{\alpha\beta}^{\rho}e_{\rho } , \end{array}\ ] ] where @xmath298 are the _ _ structure coefficients__of the frame @xmath174 and @xmath299 are the _ _
connection coefficients__in this frame . then , the components of the torsion and curvature tensors are given , respectively , by : @xmath91{c}\mathbf{t}(\vartheta^{\rho},e_{\alpha}\mathbf{,}e_{\beta})=t_{\alpha\beta } ^{\rho}=\mathbf{l}_{\alpha\beta}^{\rho}-\mathbf{l}_{\beta\alpha}^{\rho } -c_{\alpha\beta}^{\rho}\\ \mathbf{r(}e_{\mu},\vartheta^{\rho},e_{\alpha}\mathbf{,}e_{\beta}\mathbf{)}=r_{\mu}{}^{\rho}{}_{\!\alpha\beta}=e_{\alpha}(\mathbf{l}_{\beta\mu } ^{\rho})-e_{\beta}(\mathbf{l}_{\alpha\mu}^{\rho})+\mathbf{l}_{\alpha\sigma } ^{\rho}\mathbf{l}_{\beta\mu}^{\sigma}-\mathbf{l}_{\beta\sigma}^{\rho } \mathbf{l}_{\alpha\mu}^{\sigma}-c_{\alpha\beta}^{\sigma}\mathbf{l}_{\sigma \mu}^{\rho}. \end{array } \label{585}\ ] ] it is important for what follows to keep in mind the definition of the ( symmetric ) ricci tensor , here denoted @xmath300 and which in an arbitrary basis is written as @xmath301 it is crucial here to take into account the _ place _ where the contractions in the riemann tensor takes place according to our conventions .
we also have : @xmath91{l}d\vartheta^{\rho}=-\frac{1}{2}c_{\alpha\beta}^{\rho}\vartheta^{\alpha}\wedge\vartheta^{\beta}\\ d_{e_{\alpha}}\vartheta^{\rho}=-\mathbf{l}_{\alpha\beta}^{\rho}\vartheta ^{\beta}\end{array } \label{608}\ ] ] where @xmath302 are the _ connection 1-forms , _
@xmath299 are said to be the connection coefficients in the given basis , and the @xmath303 are the _ torsion 2-forms _ and the @xmath304 * * * * are the _ curvature 2-forms _ , given by : @xmath305 multiplying eqs.([585 ] ) by @xmath306 and using eqs.([608 ] ) and ( [ 620 ] ) , we get : @xmath91{l}d\vartheta^{\rho}+\omega_{\beta}^{\rho}\wedge\vartheta^{\beta}=\mathcal{t}^{\rho},\\ d\omega_{\mu}^{\rho}+\omega_{\beta}^{\rho}\wedge\omega_{\mu}^{\beta } = \mathcal{r}_{\mu}^{\rho}. \end{array } \label{559}\ ] ] we can show that the torsion and ( riemann ) curvature tensors can be written as@xmath307
sometimes , eqs.([559 ] ) are written by some authors @xcite as : @xmath309 and @xmath310 is said to be the _ exterior covariant derivative _ related to the connection @xmath7 .
now , eq.([559b ] ) has been printed with quotation marks due to the fact that it is an _ incorrect _ equation . indeed ,
exterior covariant derivative operator is a concept that can be defined for @xmath311-indexed @xmath36-form fields - forms @xmath312 , despite the name .
more precisely , the @xmath312 are not true indexed forms , i.e. , there does not exist a tensor field @xmath313 such that @xmath314 @xmath315 as follows .
suppose that @xmath316 and let @xmath317 such that for @xmath318 @xmath319 , @xmath320 the exterior covariant differential @xmath308 * * of @xmath321 on a manifold with a general connection @xmath7 is the mapping : @xmath322 such that ) ) means that the corresponding symbol is missing in the expression.]@xmath323 then , we may verify that @xmath324 note that if _ eq.([559new4 ] ) _ is applied on any one of the connection @xmath27-forms @xmath325 we would get @xmath326 .
so , we see that the symbol @xmath327 in _ eq.([559b ] ) _ , supposedly defining the curvature @xmath0-forms is simply wrong despite this being an equation printed in many physics textbooks and many professional articles _ _ '' only as a short notation .
unfortunately this is not the case for some other authors . ] _ _ ! .
the exterior covariant derivative @xmath308 satisfy the following properties : * ( a ) * for any @xmath328 @xmath329 and @xmath330 @xmath331 are sets of indexed forms and @xmath332 .
] , then@xmath333 * ( b ) * for any @xmath334 then@xmath335 * ( c ) * for any metric - compatible connection @xmath7 if @xmath336 then,@xmath337 in an orthonormal cobasis @xmath99 we have ( see , e.g. , @xcite ) for the connection @xmath27-forms @xmath339 , \label{conn}\ ] ] or taking into account that @xmath340 @xmath341
as we said above a given structure @xmath263 in general admits many different connections .
let then @xmath6 and @xmath7 be the levi - civita connection of @xmath59 on @xmath2 and @xmath7 and arbitrary riemann - cartan connection .
given an arbitrary basis @xmath174 on @xmath271 , let @xmath272 be the dual frame .
we write for the connection coefficients of the riemann - cartan and the levi - civita connections in the arbitrary bases @xmath174,@xmath272 : @xmath342 moreover , the structure coefficients of the arbitrary basis @xmath174 are : @xmath343=c_{\alpha\beta}^{\rho}e_{\rho}. \label{lr}\ ] ] let moreover,@xmath344 where @xmath345 is the lie derivative in the direction of the vector field @xmath346 .
then , we have the noticeable formula ( for a proof , see , e.g. , @xcite):@xmath347 where the tensor @xmath348 is called the strain tensor of the connection and can be decomposed as : @xmath349 where @xmath350 is its traceless part , is called the _ shear _ of the connection , and @xmath351 is its trace part , is called the _ dilation _ of the connection .
we also have that connection coefficients of the levi - civita connection can be written as:@xmath352 moreover , we introduce the _ contorsion tensor _ whose components in an arbitrary basis are defined by @xmath353 and which can be written as @xmath354 we now present the relation between the riemann curvature tensor @xmath355 associated with the riemann - cartan connection @xmath7 and the riemann curvature tensor @xmath356 of the levi - civita connection @xmath6 .
@xmath357 } , \label{1070}\ ] ] where : @xmath358 multiplying both sides of eq.([1070 ] ) by @xmath359 we get : @xmath360 where @xmath361}\theta^{\alpha}\wedge\theta^{\beta}.\ ] ] from eq.([1070 ] ) we also get the relation between the ricci tensors of the connections @xmath7 and @xmath6 .
we write for the _ ricci tensor of _ @xmath7 @xmath362 then , we have @xmath363 with @xmath364 observe that since the connection @xmath7 is arbitrary , its ricci tensor will be _ not _ be symmetric in general . then , since the ricci tensor @xmath365 of @xmath6 is necessarily symmetric , we can split eq.([1174 ] ) into : @xmath366{c}$r_{[\mu\alpha]}=j_{[\mu\alpha]},$\\ \\ $ r_{(\mu\alpha)}=\mathring{r}_{(\mu\alpha)}+j_{(\mu\alpha)}.$\end{tabular } \ \ \ \ \ \label{1190}\ ] ]
we have the following noticeable formulas whose proof can be found in , e.g. , @xcite .
let @xmath367 . then as we already know @xmath368 we have also the important formulas @xmath369
we now investigate the square of a dirac operator . we start recalling that the square of the standard dirac operator can be identified with the hodge d alembertian and that it can be separated in some interesting parts that we called in @xcite the dalembertian , ricci and einstein operators of @xmath258 .
the square of standard dirac operator @xmath11 is the operator , @xmath370 @xmath371 @xmath372 given by : @xmath373 it is quite obvious that@xmath374 and thus we recognize that @xmath11 @xmath375 is the _
hodge dalembertian _ of the manifold introduced by eq.([705 ] ) on the other hand , remembering the standard dirac operator is @xmath376 , where @xmath377 is the dual basis of an arbitrary basis @xmath174 on @xmath268 and @xmath6 is the levi - civita connection of the metric @xmath59 , we have : @xmath378 then defining the operators : @xmath366{c}(a)\\ ( b ) \end{tabular}\begin{array } [ c]{ccl}{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\cdot{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$ } } } & = & g^{\alpha\beta}(\mathring{d}_{e_{\alpha}}\mathring{d}_{e_{\beta}}-\mathbf{\mathring{l}}_{\alpha\beta}^{\rho}\mathring{d}_{e_{\rho}})\\{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\wedge{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$ } } } & = & \vartheta^{\alpha}\wedge\vartheta^{\beta}(\mathring{d}_{e_{\alpha}}\mathring{d}_{e_{\beta}}-\mathbf{\mathring{l}}_{\alpha\beta}^{\rho}\mathring{d}_{e_{\rho } } ) , \end{array } \label{1792}\ ] ] we can write : @xmath379 or,@xmath380 it is important to observe that the operators @xmath381 and @xmath382 do not have anything analogous in the formulation of the differential geometry in the cartan and hodge bundles .
the operator @xmath381 can also be written as : @xmath383 .\ ] ] applying this operator to the 1-forms of the frame @xmath384 , we get : @xmath385 where : @xmath386 the proof that an object with these components is a tensor may be found in @xcite . in particular , for every @xmath36-form field @xmath387 , @xmath388 , we have : @xmath389 where @xmath390 are the components of the covariant derivative of @xmath391 , i.e. , writing @xmath392 , it is:@xmath393 in view of eq.([1831 ] ) , we give the call the operator @xmath394 the _ covariant _ _ dalembertian_.
note that the covariant dalembertian of the 1-forms @xmath395 can also be written as : @xmath396 and therefore , taking into account the eq.([1812 ] ) , we conclude that : @xmath397 by its turn , the operator @xmath382 can also be written as : @xmath398 .\ ] ] applying this operator to the 1-forms of the frame @xmath399 , we get : @xmath400 where @xmath401 are the components of the curvature tensor of the connection @xmath6 . then using the second formula in the first line of eq.([10 ] )
we have @xmath402 the second term in the r.h.s . of this equation
is identically null because due to the first bianchi identity which for the particular case of the levi - civita connection ( @xmath403 ) is @xmath404 . the first term in eq.([rrr ] ) can be written @xmath405 where @xmath406 are the components of the ricci tensor of the levi - civita connection @xmath6 of ` \mathbf{g } ` .
thus we have a really beautiful result : @xmath407 where @xmath408 are the ricci 1-forms of the manifold .
because of this relation , we call the operator @xmath382 the _ _ ricci operator__of the manifold associated to the levi - civita connection @xmath6 of @xmath59
. we can show @xcite that the ricci operator @xmath382 satisfies the relation _ _ : _ _ @xmath409 where the curvature @xmath0-forms are @xmath410 and @xmath411 observe that applying the operator given by the second term in the r.h.s . of eq.([1918 ] ) to the dual of the 1-forms @xmath395 , we get : @xmath412 where we have used the eqs.([440new ] ) .
then , recalling the definition of the curvature forms and using the eq.([7 ] ) , we conclude that : @xmath413 where @xmath414 is the scalar curvature of the manifold and @xmath415the @xmath416 may be called the _ einstein _
@xmath27__-form fields .
_ _ that observation motivate us to introduce in @xcite the _ einstein operator _ of the levi - civita connection @xmath6 of @xmath59 on the manifold @xmath2 as the mapping @xmath417 @xmath418 given by : @xmath419 obviously , we have : @xmath420 in addition , it is easy to verify that @xmath421 and @xmath422 . thus we can also write the einstein operator as : @xmath423 where@xmath424 for any @xmath425
. we recall @xcite that if @xmath426 are the levi - civita connection 1-forms fields in an arbitrary moving frame @xmath399 on @xmath258 then : _ _ @xmath366{c}(a)\\ ( b ) \end{tabular}\begin{array } [ c]{rcl}({\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\cdot{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}})\vartheta^{\mu } & = & -({\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\cdot\mathring{\omega}_{\rho}^{\mu}-\mathring{\omega}_{\rho}^{\sigma}\cdot\mathring{\omega}_{\sigma}^{\mu})\vartheta^{\rho}\\ ( { \mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\wedge{\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}})\vartheta^{\mu } & = & -({\mathop{\partial\kern -4pt\raisebox{.8pt}{$|$}}}\wedge\mathring{\omega}_{\rho}^{\mu}-\mathring{\omega}_{\rho}^{\sigma}\wedge\mathring{\omega}_{\sigma}^{\mu})\vartheta^{\rho } , \end{array } \label{2017}\ ] ] _ _ and _ _ @xmath427 show that @xmath428 where @xmath414 is the curvature scalar .
consider the structure @xmath28 , where @xmath7 is an arbitrary riemann - cartan - weyl connection and the clifford algebra @xmath430 .
let us now compute the square of the ( general ) dirac operator @xmath431 . as in the earlier section , we have , by one side , @xmath432 and we write @xmath433 , @xmath434 and @xmath435 so that : @xmath436 the operator @xmath437 when applied to scalar functions corresponds , for the case of a riemann - cartan space , to the wave operator introduced by rapoport @xcite in his theory of stochastic mechanics .
obviously , for the case of the standard dirac operator , @xmath437 reduces to the usual hodge d alembertian of the manifold , which preserve graduation of forms . for more details
see @xcite .
on the other hand , we have also : @xmath438 and we can then define : @xmath91{ccl}{\mbox{\boldmath$\partial$}}\cdot{\mbox{\boldmath$\partial$ } } & = & g^{\alpha\beta}(d_{e_{\alpha}}d_{e_{\beta}}-\mathbf{l}_{\alpha\beta}^{\rho } d_{e_{\rho}})\\ { \mbox{\boldmath$\partial$}}\wedge{\mbox{\boldmath$\partial$ } } & = & \theta^{\alpha}\wedge\theta^{\beta}(d_{e_{\alpha}}d_{e_{\beta}}-\mathbf{l}_{\alpha\beta}^{\rho}d_{e_{\rho } } ) \end{array}\ ] ] in order to have : @xmath439 the operator @xmath440 can also be written as : @xmath441 \label{gda}\ ] ] or , @xmath442 where @xmath443 has been defined in eq.([thes ] ) . by its turn
, the operator @xmath444 can also be written as : @xmath445\end{aligned}\ ] ] or , @xmath446 for the case of a levi - civita connection we have similar formulas for @xmath381 _ ( eq.([gda ] ) ) _ and @xmath382 _ ( eq.([gricci ] ) ) _ with @xmath447 , and of course , @xmath448 , as follows directly from _
eq.([1792 ] ) .
we now take @xmath263 as a lorentzian manifold , i.e. , @xmath58 and the signature of @xmath59 is @xmath8 .
we consider moreover a lorentzian spacetime structure on @xmath263 , i.e. , the pentuple @xmath449 and a riemann - cartan spacetime structure @xmath450 . now ,
in both spacetime structures , maxwell equations in vacuum read:@xmath451 where @xmath452 is the faraday tensor ( electromagnetic field ) and @xmath453 is the current .
we observe that writing @xmath454 we have using eq.([hodge dual ] ) that @xmath455 thus@xmath456 the homogeneous maxwell equation @xmath457 can be writing as @xmath458 .
the proof follows at once from the definition of @xmath10 ( eq.([hodge ] ) ) .
indeed , we can write@xmath459 then @xmath460 and we end with@xmath461 ( a ) * * we now express the equivalent equations @xmath462 and @xmath463 in arbitrary coordinates @xmath264 covering @xmath78 using first the levi - civita connection and noticeable formula in eq.([d and delta ] ) .
we have @xmath464 \\ & = \frac{1}{2}\theta^{\alpha}\wedge\left [ ( { \partial}_{\alpha}f_{\mu\nu } ) \theta^{\mu}\wedge\theta^{\nu}-f_{\mu\nu}\mathring{\gamma}_{\alpha\rho}^{\mu}\theta^{\rho}\wedge\theta^{\nu}-f_{\mu\nu}\mathring{\gamma}_{\alpha\rho } ^{\nu}\theta^{\mu}\wedge\theta^{\rho}\right ] \\ & = \frac{1}{2}\theta^{\alpha}\wedge\left [ ( \mathring{d}_{\alpha}f_{\mu\nu } ) \theta^{\mu}\wedge\theta^{\nu}\right ] \\ & = \frac{1}{2}d_{\alpha}f_{\mu\nu}\theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & = \frac{1}{2}\left [ \frac{1}{3}\mathring{d}_{\alpha}f_{\mu\nu}\theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}+\frac{1}{3}\mathring { d}_{\mu}f_{\nu\alpha}\theta^{\mu}\wedge\theta^{\nu}\wedge\theta^{\alpha } + \frac{1}{3}\mathring{d}_{\nu}f_{\alpha\mu}\theta^{\nu}\wedge\theta^{\alpha } \wedge\theta^{\mu}\right ] \\ & = \frac{1}{2}\left [ \frac{1}{3}\mathring{d}_{\alpha}f_{\mu\nu}\theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}+\frac{1}{3}\mathring { d}_{\mu}f_{\nu\alpha}\theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu } + \frac{1}{3}\mathring{d}_{\nu}f_{\alpha\mu}\theta^{\alpha}\wedge\theta^{\mu } \wedge\theta^{\nu}\right ] \\ & = \frac{1}{6}\left ( \mathring{d}_{\alpha}f_{\mu\nu}+\mathring{d}_{\mu } f_{\nu\alpha}+\mathring{d}_{\nu}f_{\alpha\mu}\right ) \theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}.\end{aligned}\ ] ] so , @xmath465 if we calculate @xmath457 using the definition of @xmath9 we get:@xmath466 from where we get that@xmath467 next we calculate @xmath468 .
we have @xmath469 \right\ } \nonumber\\ & = -\frac{1}{2}\theta^{\alpha}\lrcorner\left\ { ( { \partial}_{\alpha}{}^{\star}f_{\mu\nu})\theta^{\mu}\wedge\theta^{\nu}-{}^{\star}f_{\mu\nu } \mathring{\gamma}_{\alpha\rho}^{\mu}\theta^{\rho}\wedge\theta^{\nu}-{}^{\star}f_{\mu\nu}\mathring{\gamma}_{\alpha\rho}^{\nu}\theta^{\mu}\wedge \theta^{\rho}\right\ } \nonumber\\ & = -\frac{1}{2}\theta^{\alpha}\lrcorner\left\ { ( { \partial}_{\alpha}{}^{\star}f_{\mu\nu})\theta^{\mu}\wedge\theta^{\nu}-{}^{\star}f_{\rho\nu } \mathring{\gamma}_{\alpha\mu}^{\rho}\theta^{\mu}\wedge\theta^{\nu}-{}^{\star } f_{\mu\rho}\mathring{\gamma}_{\alpha\nu}^{\rho}\theta^{\mu}\wedge\theta^{\nu } \right\ } \nonumber\\ & = -\frac{1}{2}\theta^{\alpha}\lrcorner\left\ { ( \mathring{d}_{\alpha}{}^{\star}f_{\mu\nu})\theta^{\mu}\wedge\theta^{\nu}\right\ } \nonumber\\ & = -\frac{1}{2}\left\ { ( \mathring{d}_{\alpha}{}^{\star}f_{\mu\nu})g^{\alpha\mu}\theta^{\nu}-(\mathring{d}_{\alpha}{}^{\star}f_{\mu\nu } ) g^{\alpha\nu}\theta^{\mu}\right\ } \nonumber\\ & = -(\mathring{d}_{\alpha}{}^{\star}f_{\mu\nu})g^{\alpha\mu}\theta^{\nu } \nonumber\\ & = -[\mathring{d}_{\alpha}{}(^{\star}f_{\mu\nu}g^{\alpha\mu})]\theta^{\nu } \nonumber\\ & = -(\mathring{d}_{\alpha}{}^{\star}f_{\,\nu}^{\alpha})]\theta^{\nu}. \label{delf}\ ] ] then we get that @xmath470 \(b ) _ _ also , the non homogenoeous maxwell equation @xmath471 can be written using the definition of @xmath10 ( eq.([hodge ] ) ) as @xmath472:@xmath473 we now express @xmath474 in arbitrary coordinates . ] using first the levi - civita connection .
we have following the same steps as in eq.([delf ] ) @xmath475 \right\ } + j_{\nu}\theta^{\nu } \label{df6}\\ & = ( -\mathring{d}_{\alpha}f_{\,\nu}^{\alpha}+j_{\nu})\theta^{\nu}.\nonumber\end{aligned}\ ] ] then@xmath476 we also observe that using the symmetry of the connection coefficients and the antisymmetry of the @xmath477 that @xmath478 . also , @xmath479 and @xmath480 then@xmath481 and@xmath482 show that in a lorentzian spacetime maxwell equations become maxwell equation , i.e. , @xmath483 from time to time we see papers ( e.g. , @xcite ) writing maxwell equations in a riemann - cartan spacetime using arbitrary coordinates and ( of course ) the riemann - cartan connection . as we shall see such enterprises are simple exercises , if we make use of the noticeable formulas of eq.([d delta bis ] ) .
indeed , the homogeneous maxwell equation @xmath457 reads @xmath484 or @xmath485\\ & = \frac{1}{6}(d_{\alpha}f_{\mu\nu}+d_{\mu}f_{\nu\alpha}+d_{\nu}f_{\alpha\mu } ) \theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & -\frac{1}{2}t_{\rho\sigma}^{\alpha}f_{\mu\nu}\theta^{\rho}\wedge \theta^{\sigma}\wedge\delta_{\alpha}^{\mu}\theta^{\nu}\\ & = \frac{1}{6}(d_{\alpha}f_{\mu\nu}+d_{\mu}f_{\nu\alpha}+d_{\nu}f_{\alpha\mu } ) \theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & -\frac{1}{2}t_{\alpha\mu}^{\sigma}f_{\sigma\nu}\theta^{\alpha}\wedge \theta^{\mu}\wedge\theta^{\nu}\\ & = \frac{1}{6}(d_{\alpha}f_{\mu\nu}+d_{\mu}f_{\nu\alpha}+d_{\nu}f_{\alpha\mu } ) \theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & -\frac{1}{6}(t_{\alpha\mu}^{\sigma}f_{\sigma\nu}+t_{\mu\nu}^{\sigma } f_{\sigma\alpha}+t_{\nu\alpha}^{\sigma}f_{\sigma\mu})\theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & = \frac{1}{6}(d_{\alpha}f_{\mu\nu}+d_{\mu}f_{\nu\alpha}+d_{\nu}f_{\alpha\mu } ) \theta^{\alpha}\wedge\theta^{\mu}\wedge\theta^{\nu}\\ & + \frac{1}{6}(f_{\alpha\sigma}t_{\mu\nu}^{\sigma}+f_{\mu\sigma}t_{\nu\alpha } ^{\sigma}+f_{\nu\sigma}t_{\alpha\mu}^{\sigma})\theta^{\alpha}\wedge \theta^{\mu}\wedge\theta^{\nu}.\end{aligned}\ ] ] i.e.,@xmath486 also , taking into account that @xmath487 we have using the second noticeable formula in eq.([d delta bis ] ) _ _ that @xmath488 now,@xmath489 and @xmath490\\ & = \frac{1}{4}t_{\beta\rho}^{\alpha}{}^{\star}f^{\mu\nu}\theta^{\beta } \lrcorner(\delta_{\alpha}^{\rho}\theta_{\mu}\wedge\theta_{\nu}-\delta_{\mu } ^{\rho}\theta_{\alpha}\wedge\theta_{\nu}+\delta_{\nu}^{\rho}\theta_{\alpha } \wedge\theta_{\mu})\\ & = \frac{1}{4}t_{\beta\alpha}^{\alpha}{}^{\star}f^{\mu\nu}\theta^{\beta } \lrcorner(\theta_{\mu}\wedge\theta_{\nu})-\frac{1}{4}t_{\beta\mu}^{\alpha}{}^{\star}f^{\mu\nu}\theta^{\beta}\lrcorner(\theta_{\alpha}\wedge\theta_{\nu } ) + \frac{1}{4}t_{\beta\nu}^{\alpha}{}^{\star}f^{\mu\nu}\theta^{\beta}\lrcorner(\theta_{\alpha}\wedge\theta_{\mu})\end{aligned}\]]@xmath491 using eqs.([df13 ] ) and ( [ df14 ] ) in eq.([df12 ] ) we get@xmath492 and we have@xmath493 finally we express the non homogenous maxwell equation @xmath494 in arbitrary coordinates using the riemann - cartan connection .
we have @xmath495\theta_{\nu}=-j^{\nu}\theta_{\nu } , \label{df18}\ ] ] i.e.,@xmath496 show _ _ ( _ _ use _ eq.([d delta bis ] ) ) _ that in a riemann - cartan spacetime maxwell equations become maxwell equation , i.e.,@xmath497
we rewrite cartan s structure equations _ for an arbitrary riemann - cartan structure _ @xmath498 where @xmath499 and @xmath59 is a metric of signature @xmath4 , with @xmath5 using an arbitrary cotetrad @xmath99 as @xmath91{l}\mathcal{t}^{\mathbf{a}}=d\theta^{\mathbf{a}}+\omega_{\mathbf{b}}^{\mathbf{a}}\wedge\theta^{\mathbf{b}}=\mathbf{d}\theta^{\mathbf{a}},\\ \mathcal{r}_{\mathbf{b}}^{\mathbf{a}}=d\omega_{\mathbf{b}}^{\mathbf{a}}+\omega_{\mathbf{c}}^{\mathbf{a}}\wedge\omega_{\mathbf{b}}^{\mathbf{c}}\end{array } \label{cartan strc}\ ] ] where @xmath500 since the @xmath19 and the @xmath20 are _ index form fields _ we can apply to those objects the exterior covariant differential ( eq.([559new4 ] ) ) .
we get @xmath501 also,@xmath502 so , we have the general bianchi identities which are valid for any one of the metrical compatible structures classified in section 2,@xmath503 taking advantage of the calculations we done for the coordinate expressions of maxwell equations we can write in a while : @xmath504 now , @xmath505 and using the freshman identity ( eq.([freshman ] ) ) we can write @xmath506 so , @xmath507 now , recalling that @xmath508 we can write@xmath509 using these formulas we can write@xmath510 now , the coordinate representation of @xmath511 is:@xmath512 and thus the coordinate expression of the first bianchi identity is:@xmath513 which we can write as@xmath514 with @xmath515 denoting as usual the sum over cyclic permutation of the indices @xmath516 . for
the particular case of a levi - civita connection @xmath6 since the @xmath517 we have the standard form of the first bianchi identity in classical riemannian geometry , i.e. , @xmath518 if we now recall the steps that lead us to eq.([df16 ] ) we can write for the torsion @xmath0-form fields @xmath19 , @xmath519 with @xmath520
. then we can write the first bianchi identity as@xmath521 , \label{bi10}\ ] ] and taking into account that @xmath522 we end with@xmath523 this is the first bianchi identity written in terms of duals . to calculate its coordinate expression , we recall the steps that lead us to eq.([df16 ] ) and write directly for the torsion @xmath0-form fields @xmath19 @xmath524 also , writing@xmath525 we have : @xmath526 on the other hand we can
also write:@xmath527 from where we get in agreement with eq.([hodge dual ] ) the formula@xmath528 which shows explicitly that @xmath529 are not the components of the _ ricci _ tensor .
moreover,@xmath530 collecting the above formulas we end with@xmath531 which is another expression for the first bianchi identity written in terms of duals .
consider , e.g. , the term @xmath532 in the above equation and write @xmath533 we now show that @xmath534 indeed , recall that we already found that @xmath535 and taking into account the second formula in _ eq.([d delta bis ] ) _ we can write @xmath536 now , writing @xmath537 @xmath538 and get @xmath539\nonumber\\ & = \frac{1}{2}\vartheta^{\alpha}\lrcorner\lbrack\partial_{\alpha}(q_{\rho } ^{\mathbf{a}}{}^{\star}t^{\mathbf{\rho}\mu\nu})\theta_{\mu}\wedge\theta_{\nu } + q_{\rho}^{\mathbf{a}}{}^{\star}t^{\mathbf{\rho}\mu\nu}d_{\mathbf{\partial } _ { \alpha}}(\theta_{\mu}\wedge\theta_{\nu})]\nonumber\\ & = \frac{1}{2}\vartheta^{\alpha}\lrcorner\lbrack(\partial_{\alpha}q_{\rho } ^{\mathbf{a}}){}^{\star}t^{\mathbf{\rho}\mu\nu}\theta_{\mu}\wedge\theta_{\nu } + q_{\rho}^{\mathbf{a}}\partial_{\alpha}({}^{\star}t^{\mathbf{\rho}\mu\nu } ) \theta_{\mu}\wedge\theta_{\nu}+q_{\rho}^{\mathbf{a}}{}^{\star}t^{\mathbf{\rho}\mu\nu}d_{\mathbf{\partial}_{\alpha}}(\theta_{\mu}\wedge \theta_{\nu})]\nonumber\\ & = \frac{1}{2}\vartheta^{\alpha}\lrcorner\lbrack(\partial_{\alpha}q_{\rho } ^{\mathbf{a}}){}^{\star}t^{\mathbf{\rho}\mu\nu}\theta_{\mu}\wedge\theta_{\nu } + q_{\rho}^{\mathbf{a}}d_{\alpha}({}^{\star}t^{\mathbf{\rho}\mu\nu})\theta_{\mu}\wedge\theta_{\nu}]\nonumber\\ & = ( \partial_{\alpha}q_{\rho}^{\mathbf{a}}){}^{\star}t^{\mathbf{\rho}\mu\nu } \delta_{\mu}^{\alpha}\theta_{\nu}+q_{\rho}^{\mathbf{a}}d_{\alpha}({}^{\star } t^{\mathbf{\rho}\mu\nu})\delta_{\mu}^{\alpha}\theta_{\nu}. \label{bi22}\ ] ] comparing the _ eq.([bi20 ] ) _ with _ eq.([bi21 ] ) _ using _ eq.([bi22 ] ) _ we get@xmath540 thus proving our statement and showing the danger of applying a so called `` tetrad postulate '' asserting without due care on the meaning of the symbols that the covariant derivative of the tetrad is zero , and thus using @xmath541 .
show that the coordinate expression of the second bianchi identity @xmath542 is@xmath543 calculate @xmath544 in an orthonormal basis .
* solution : * first we recall the @xmath545 and then use the formula in the third line of eq.([440new ] ) to write:@xmath546 \nonumber\\ & = -\star\lbrack r_{\mathbf{b\;cd}}^{\;\mathbf{a}}\eta^{\mathbf{bc}}\theta^{\mathbf{d}}]\nonumber\\ & = -\star\lbrack r_{\;\;\mathbf{cd}}^{\mathbf{ca}}\theta^{\mathbf{d}}]=-\star\lbrack r_{\;\;\mathbf{dc}}^{\mathbf{ac}}\theta^{\mathbf{d}}]\nonumber\\ & = -\star\lbrack r_{\mathbf{d}}^{\mathbf{a}}\theta^{\mathbf{d}}]=-\star\mathcal{r}^{\mathbf{a } } \label{bi25}\ ] ] of course , if the connection is the levi - civita one we get@xmath547
( [ bi17 ] ) or its equivalent eq.([bi23 ] ) is to be compared with a wrong one derived by evans from where he now claims that the einstein - hilbert ( gravitational ) theory which uses in its formulation the levi - civita connection @xmath6 is incompatible with the first bianchi identity .
evans conclusion follows because he thinks to have derived from first principles that @xmath366{l}$\mathbf{d\star}\mathcal{t}^{\mathbf{a}}=\star\mathcal{r}_{\mathbf{b}}^{\mathbf{a}}\wedge\theta^{\mathbf{b}},$\end{tabular } \ \ \ \ \ \ \
\label{evans 24}\ ] ] an equation that _ if _ true implies as we just see from eq.([bi25 ] ) that for the levi - civita connection for which the @xmath549 the ricci tensor of the connection @xmath6 is null .
we show below that eq.([evans 24 ] ) is a false one in two different ways , firstly by deriving the correct equation for @xmath550 and secondly by showing explicit counterexamples for some trivial structures . before doing that
let us show that we can derive from the first bianchi identity that@xmath551 an equation that eventually may lead evans in believing that for a levi - civita connection the first bianchi identity implies that the ricci tensor is null . as we know
, for a levi - civita connection the first bianchi identity gives ( with @xmath552 : @xmath553 contracting this equation with @xmath554 we get@xmath555\\ & = \mathcal{\mathring{r}}_{\mathbf{a}}^{\mathbf{a}}-\mathring{r}_{\mathbf{b\;ad}}^{\;\mathbf{a}}\theta^{\mathbf{b}}\wedge\theta^{\mathbf{d}}\ ] ] now , the second term in this last equation is null because according to the eq.([ricci ] ) , @xmath556 are the components of the ricci tensor , which is a symmetric tensor for the levi - civita connection . for the first term
we get @xmath557 which implies that as we stated above that@xmath558 but according to eq.([ricci ] ) the@xmath559 are not the components of the ricci tensor , and so there is not any contradiction . as an additional verification recall that the standard form of the first bianchi identity in riemannian geometry is@xmath560 making @xmath561 we get@xmath562
we now present using results of clifford bundle formalism , recalled above ( for details , see , e.g. , @xcite ) a calculation of @xmath563 .
we start from cartan first structure equation @xmath564 by definition @xmath565 now , if we recall eq.([545 ] ) , since the @xmath566 we can write @xmath567 we next calculate @xmath568 .
we have:@xmath569 next we recall the definition of the hodge dalembertian which , recalling eq.([ssd ] ) permit us to write the first two terms in eq.([du3 ] ) as the negative of the square of the standard dirac operator ( associated with the levi - civita connection ) .
we then get:@xmath570 where we have used eq([1174 ] ) to write @xmath571 so , we have@xmath572 and finally@xmath573 or equivalently recalling eq.([440new ] ) @xmath574 _ eq.([du7b ] ) _ does not implies that @xmath575 because taking into account _
eq.([bi25 ] ) _ @xmath576 in general .
so , for a levi - civita connection we have that @xmath577 and then eq.([du6 ] ) implies @xmath578 or since @xmath579 for a levi - civita connection,@xmath580 or yet @xmath581 an identity that we already mentioned above ( eq.([ssd1 ] ) ) .
the reader can easily verify that einstein equations in the clifford bundle formalism is written as:@xmath582 where @xmath414 is the scalar curvature and @xmath583 are the energy - momentum @xmath27-form fields .
comparing eq.([du8 ] ) with eq.([e1 ] ) .
we immediately get the `` wave equation '' for the cotetrad fields : @xmath584 which does not implies that the ricci tensor is null .
we see from eq.([e2 ] ) that a ricci flat spacetime is characterized by the equality of the hodge and covariant d alembertians acting on the coterad fields , i.e.,@xmath585 a non trivial result . using _
eq.([1831 ] ) _ and _ eq.([1832 ] ) write _ @xmath586 in terms of the connection coefficients of the riemann - cartan connection .
consider the well known riemannian structure on the unit radius sphere @xcite @xmath589 .
let @xmath590 , @xmath591 , @xmath592 , @xmath593 , @xmath594 , be spherical coordinates covering @xmath595 , where @xmath596 is the curve joining the north and south poles .
a coordinate basis for @xmath82 is then @xmath81 and its dual basis is @xmath83 .
the riemannian metric @xmath597 is given by @xmath598 and the metric ` g`@xmath599 of the cotangent space is@xmath600 an orthonormal basis for @xmath82 is then @xmath93 with@xmath601 with dual basis @xmath99 given by@xmath602 the structure coefficients of the orthonormal basis are @xmath603=c_{\mathbf{ij}}^{\mathbf{k}}\mathbf{e}_{\mathbf{k } } \label{ce4a}\ ] ] and can be evaluated , e.g. , by calculating @xmath604 .
we get immediately that the only non null coefficients are@xmath605 to calculate the connection @xmath27-form @xmath606 we use eq.([conna ] ) , i.e. , @xmath607 then , @xmath608 then @xmath609@xmath610 now , from cartan s second structure equation we have@xmath611 and ] @xmath612 now , let us calculate @xmath613 . we have @xmath614 and @xmath615 now
, evans equation implies that @xmath616 for a levi - civita connection and thus as promised we exhibit a counterexample to his wrong equation .
we recall that the first bianchi identity for @xmath617 , i.e. , @xmath618 which translate in the orthonormal basis used above in @xmath619 is rigorously valid .
indeed , we have @xmath620 consider the manifold @xmath622 @xmath623@xmath624 , it is an sphere excluding the north pole .
let ` ` @xmath625 be the standard riemann metric field for @xmath622 ( eq.([ce1 ] ) ) .
now , consider besides the levi - civita connection another one , @xmath7 , here called the nunes ( or navigator @xcite ) connection .
it is defined by the following parallel transport rule : a vector is parallel transported along a curve , if at any @xmath626 the angle between the vector and the vector tangent to the latitude line passing through that point is constant during the transport ( see figure [ fignunes ] ) [ h ] .. /documents and settings / waldir / meus documentos / gs / s2-plano.jpg as before @xmath627 @xmath593 , @xmath594 , denote the standard spherical coordinates of a @xmath628 of unitary radius , which covers @xmath629 , where @xmath596 is the curve joining the north and south poles .
now , it is obvious from what has been said above that our connection is characterized by @xmath630 then taking into account the definition of the curvature tensor we have @xmath631}^{c}\right ] \mathbf{e}_{\mathbf{k}}\right ) = 0.\ ] ] also , taking into account the definition of the torsion operation we have@xmath632\nonumber\\ & = [ \mathbf{e}_{\mathbf{i}},\mathbf{e}_{\mathbf{j}}]=c_{\mathbf{ij}}^{\mathbf{k}}\mathbf{e}_{\mathbf{k}},\end{aligned}\]]@xmath633 it follows that the unique non null torsion @xmath0-form is : @xmath634 if you still need more details , concerning this last result , consider figure [ fignunes](b ) which shows the standard parametrization of the points @xmath635 in terms of the spherical coordinates introduced above @xcite .
according to the geometrical meaning of torsion , we determine its value at a given point by calculating the difference between the ( infinitesimal ) vectors @xmath636 and @xmath637 determined as follows .
if we transport the vector @xmath638 along @xmath639 we get the vector @xmath640 such that @xmath641 . on the other hand ,
if we transport the vector @xmath639 along @xmath642 we get the vector @xmath643 .
let @xmath644 .
then , @xmath645 also , @xmath646 then , the connection @xmath7 of the structure @xmath621 has a non null torsion tensor @xmath647 .
indeed , the component of @xmath648 in the direction @xmath649 is precisely @xmath650 .
so , we get ( recalling that @xmath651 @xmath652 show that @xmath7 is metrical compatible , i.e. , @xmath48 . * * solution:**@xmath653 our counterexamples that involve the parallel transport rules defined by a levi - civita connection and a teleparallel connection in @xmath622 show clearly that we can not _
mislead _ the riemann curvature tensor of a connection defined in a given manifold with the fact that the manifold may be bend as a surface in an euclidean manifold where it is embedded .
neglecting this fact may generate a lot of wishful thinking .
in this paper after recalling the main definitions and a collection of tricks of the trade concerning the calculus of differential forms on the cartan , hodge and clifford bundles over a riemannian or riemann - cartan space or a lorentzian or riemann - cartan spacetime we solved with details several exercises involving different grades of difficult and which we believe , may be of some utility for pedestrians and even for experts on the subject .
in particular we found using technology of the clifford bundle formalism the correct equation for @xmath563 .
we show that the result found in @xcite , namely @xmath654 is wrong since it contradicts the right formula we found . besides that
, the wrong formula is also contradicted by two simple counterexamples that we exhibited in section 15 .
the last sentence before the conclusions is a crucial remark , which each one seeking truth must always keep in mind : do not confuse the riemann curvature tensor of a connection defined in a given manifold with the fact that the manifold may be bend as a surface in an euclidean manifold where it is embedded .
we end the paper with a necessary explanation .
an attentive reader may ask : why write a bigger paper as the present one to show wrong a result not yet published in a scientific journal ?
the justification is that dr .
evans maintains a site on his ( so called ) ece theory which is read by thousand of people that thus are being continually mislead , thinking that its author is creating a new mathematics and a new physics . besides that ,
due to the low mathematical level of many referees , dr .
evans from time to time succeed in publishing his papers in sci journals , as the recent ones . , @xcite .
in the past we already showed that several published papers by dr .
evans and colleagues contain serious flaws ( see , e.g. , @xcite ) and recently some other authors spent time writing papers to correct mr .
evans claims ( see , e.g.,@xcite ) it is our hope that our effort and of the ones by those authors just quoted serve to counterbalance dr .
evans influence on a general public which being anxious for novelties may be eventually mislead by people that claim among other things to _ know _
@xcite how to project devices to withdraw energy from the vacuum .
the author is grateful to prof .
e. a. notte - cuello , for have checking all calculations and discovered several misprints , that are now corrected .
moreover , the author will be grateful to any one which point any misprints or eventual errors . | in this paper after recalling some essential tools concerning the theory of differential forms in the cartan , hodge and clifford bundles over a riemannian or riemann - cartan space or a lorentzian or riemann - cartan spacetime we solve with details several exercises involving different grades of difficult .
one of the problems is to show that a recent formula given in @xcite for the exterior covariant derivative of the hodge dual of the torsion @xmath0-forms is simply wrong .
we believe that the paper will be useful for students ( and eventually for some experts ) on applications of differential geometry to some physical problems .
a detailed account of the issues discussed in the paper appears in the table of contents . |
in wilson s renormalization group @xcite , renormalized theories come as renormalized trajectories of effective actions .
the renormalization group leaves invariant a renormalized trajectory up to a flow of renormalization parameters .
this invariance prescribes a renormalized theory without a detour to limit procedures .
renormalization group invariance was used in @xcite as a first principle to deduce a renormalized running coupling expansion for the renormalized trajectory of massless @xmath0theory on four dimensional euclidean space time .
the construction yielded a pair consisting of an effective potential @xmath2 and a step @xmath3function @xmath4 , both in perturbation theory in @xmath5 , which obey the discrete flow equation @xmath6 here @xmath7 stands for a momentum space renormalization group transformation , based on the decomposition of @xmath8 by means of an exponential regulator , which rescales by a factor of @xmath9 .
( [ 1.1 ] ) was shown to possess a unique solution consisting to first order in @xmath5 of a @xmath0vertex together with a wave function term , @xmath10 where @xmath11 , and consisting to all higher orders of polynomial vertices with all properties of a renormalized theory
. the higher orders of ( [ 1.2 ] ) where determined inductively .
the step @xmath3function came out as @xmath12 showing asymptotic freedom for the forward flow on the renormalized @xmath0trajectory .
the construction relied on the analysis of @xmath7 with a fixed scale @xmath9 .
the scale parameter @xmath13 is the ratio of an ultraviolet and an infrared cutoff .
remarkably , the effective potential @xmath2 comes out independent of @xmath13 .
the only @xmath13dependence remains in the step @xmath3function ( [ 1.3 ] ) .
the differential @xmath3function @xmath14 on the other hand comes out independent of @xmath13 .
this suggests a differential approach for the pair @xmath2 and @xmath15 which makes no reference to the scale @xmath13 at all . that is the present program .
it yields an alternative construction for the renormalized @xmath0trajectory .
the differential form of the discrete flow equation ( [ 1.1 ] ) is a functional differential equation .
for the normal ordered potential @xmath16 it reads @xmath17 the operator @xmath18 generates scale transformations . the right hand side of ( [ 1.5 ] ) is a bilinear renormalization group bracket @xmath19 .
it consists of contractions between @xmath20 and @xmath21 , and is independent of @xmath5 .
restricted to the renormalized trajectory , the renormalization group becomes a one dimensional dynamical system given by a flow of @xmath5 according to the ordinary differential equation @xmath22 the step @xmath3function ( [ 1.3 ] ) is retained as a solution of ( [ 1.6 ] ) .
the differential approach parallels completely its discrete brother @xcite .
it yields an alternative construction of the same renormalization group invariant .
the scheme will be to solve ( [ 1.5 ] ) inductively by perturbation theory in @xmath5 .
the @xmath0trajectory will be selected as the unique solution which is given by ( [ 1.2 ] ) to first order in @xmath5 , and higher orders of appropriate type .
we remark that the first order wave function renormalization in ( [ 1.2 ] ) is peculiar to the four dimensional case .
both the aim and the setup of this paper are identical to those in @xcite . however , the difference equations of @xcite are here replaced by differential equations .
the formulation to be preferred is a matter of taste .
the differential version has the advantage that it involves a quadratic non linearity only . as a consequence ,
the inductive solution involves only pairwise convolution of lower order vertices .
the setup will be as in @xcite , with the difference that the scale @xmath13 is here variable . consider the following momentum space renormalization group transformation @xmath7 , depending on a scale parameter @xmath9 .
let @xmath7 be composed of a gaussian fluctuation integral , with covariance @xmath23 and mean @xmath24 , and a dilatation @xmath25 of @xmath24 .
let the fluctuation covariance be defined by @xmath26 where @xmath27 .
let @xmath28 be the associated gaussian measure on field space with mean zero .
let @xmath25 be defined by @xmath29 .
we then define @xmath7 as the renormalization group transformation @xmath30 we will restrict our attention to @xmath31symmetric potentials , with @xmath32 . the field independent constant will be understood to be removed in ( [ 2.2 ] ) .
the composition of two renormalization group transformations with scale @xmath13 is equal to one renormalization group transformation with scale @xmath33 .
moreover , the transformation ( [ 2.2 ] ) satisfies @xmath34 it therefore defines a representation of the semi group of dilatations with scale factors @xmath9 on the space of effective potentials . due to the semi - group property the iteration of renormalization group transformations with fixed scale is identical with an increase of the scale in a single transformation .
this interpolation is the motive of the present investigation , in conjunction with the previous construction in @xcite .
the continuous point of view has the advantage to allow for infinitesimal renormalization group transformations , which can be expected to be close to the identity .
the renormalization group transformation ( [ 2.2 ] ) associates with a bare potential @xmath35 an orbit @xmath36 .
this orbit satisfies the functional differential equation @xmath37 here @xmath38 generates dilatations of the field @xmath39 .
the operator @xmath40 generates scale transformations of the potential , @xmath41 the operator @xmath42 is given by @xmath43 and is @xmath13independent . with an exponential regulator
we have that @xmath44 .
( [ 2.5 ] ) follows from a functional heat equation for convolutions by a parameter dependent gaussian measure , see @xcite .
the functional laplacian on the left hand side of ( [ 2.5 ] ) can be removed by normal ordering at the expense of a more complicated bilinear term .
we define a normal ordered potential @xmath45 by @xmath46 strictly speaking @xmath45 is the pre image of @xmath47 under the normal ordering operator and is therefore not decorated by normal ordering colons . to remove the functional laplacian , the normal ordering covariance has to be chosen such that @xmath48
specifically , we choose @xmath49 a massless covariance with unit ultraviolet cutoff .
the normal ordered potential then satisfies the functional differential equation @xmath50 the non linearity comes in form of a bilinear renormalization group bracket @xmath51 where @xmath52 and @xmath53 denote two independent copies of @xmath39 .
it consists of contractions between @xmath20 and @xmath21 , each contraction being made of one hard propagator @xmath42 and any number of soft propagators @xmath54 .
consider the following renormalization problem .
we seek an effective potential @xmath55 and a differential @xmath3function @xmath15 , both depending on a coupling parameter @xmath5 , but not on @xmath13 , with the following properties of a renormalized theory . + * 1 .
power series : * the effective potential @xmath55 and the differential @xmath3function @xmath15 are both formal power series @xmath56 in @xmath5 .
the question of the summability of ( [ 3.1 ] ) and ( [ 3.2 ] ) will be left aside . + * 2 .
@xmath0theory : * the effective potential @xmath55 is to first order a @xmath0vertex together with a wave function term , @xmath57 where @xmath58 is a first order wave function parameter .
the @xmath59th order effective potential is a polynomial @xmath60 in the field @xmath39 . + * 3 .
regularity : * the kernels in ( [ 3.4 ] ) are euclidean invariant symmetric distributions .
they are given by fourier integrals @xmath61 with the @xmath62function removed , their fourier transforms are euclidean invariant symmetric @xmath63functions on momentum space @xmath64 .
they satisfy @xmath65bounds @xmath66 for all @xmath67 . here
@xmath68 is a multi index and @xmath69 .
the @xmath65norm is defined as @xmath70 where @xmath71 denotes the hyperplane of momenta with @xmath72 .
coupling parameter : * the four point kernel at zero momenta is @xmath73 zero in higher orders than one .
invariance : * ( [ 3.1 ] ) and ( [ 3.2 ] ) satisfy the renormalization group equation @xmath74 to all orders in @xmath5 .
+ the differential @xmath3function ( [ 3.2 ] ) defines a running ( scale dependent ) coupling @xmath75 as solution to the ordinary differential equation @xmath76 the effective potential @xmath77 , as a function of @xmath39 and @xmath13 , then satisfies the flow equation ( [ 2.11 ] ) . in perturbation theory
, there exists a unique solution to this renormalization problem
. we will construct it by induction on the order of @xmath5 .
the renormalization group equation ( [ 3.9 ] ) is independent of @xmath13 .
its perturbative solution will also be independent of @xmath13 .
in other words , the solution will depend on cutoffs only through a running coupling .
this property is called scaling .
a potential @xmath55 , together with a differential @xmath3function @xmath15 , which satisfies ( [ 3.9 ] ) is said to scale .
inserting the power series ( [ 3.1 ] ) and ( [ 3.2 ] ) into ( [ 3.9 ] ) , we obtain a system of first order functional differential equations for their coefficients .
it reads @xmath78 the right hand side of ( [ 4.1 ] ) depends only on lower orders @xmath79 , with @xmath80 , and is zero to first order .
it is given by @xmath81 if @xmath79 is given by ( [ 3.3 ] ) to first order and is a polynomial ( [ 3.4 ] ) for all orders @xmath82 , then also ( [ 4.2 ] ) is a polynomial @xmath83 the polynomial form ( [ 3.4 ] ) is therefore preserved by the renormalization group equation ( [ 3.9 ] ) .
( [ 4.1 ] ) will be understood as a system of first order partial differential equations for the kernels in ( [ 3.4 ] ) .
it reads @xmath84 with real space scaling dimensions @xmath85 we will look for solutions to ( [ 4.4 ] ) whose fourier transforms ( [ 3.5 ] ) obey ( [ 3.6 ] ) .
we therefore switch from real space to momentum space . in momentum space (
[ 4.4 ] ) becomes a system of first order partial differential equations @xmath86 with momentum space scaling dimensions @xmath87 the contribution @xmath88 in ( [ 4.7 ] ) comes from translation invariance .
anticipating @xmath89 , the scaling dimension of the @xmath0vertex , the scaling dimensions ( [ 4.7 ] ) become order independent in four dimensions .
they are given by @xmath90 the renormalized @xmath0trajectory is a particular solution to the scaling equation ( [ 4.6 ] ) .
it is distinguished by the properties listed above .
we will solve the scaling equations ( [ 4.1 ] ) by induction on the order of perturbation theory .
the induction step consists of solving a system of first order renormalization group pdes ( [ 4.6 ] ) of the general form @xmath91 where @xmath92 , and where @xmath93 is a given function of @xmath94 .
the integration of ( [ 5.1 ] ) calls for initial data .
it is usually supplied in form of a bare action together with renormalization conditions at a distinguished scale .
we will instead look for smooth solutions to ( [ 5.1 ] ) . recall that smoothness was one condition on the momentum space kernels ( [ 3.5 ] ) in our renormalization problem .
therefore , we will assume that @xmath95 and look for solutions @xmath96 of ( [ 5.1 ] ) .
the right hand side of ( 4.6 ) is a sum of multiple convolutions of lower order vertices with cutoff propagators . with an exponential cutoff ,
smoothness therefore iterates through the induction .
but it is luxury in the sense that no higher momentum space derivatives are needed than those required by the taylor expansions of the non irrelevant momentum space kernels .
the reader is invited to diminish the demands on regularity to this situation in the following statements .
consider first the case when @xmath97 .
we can then immediately integrate ( [ 5.1 ] ) .
it is equivalent to @xmath98 a special solution thereto is @xmath99 the integral converges for @xmath97 and yields a smooth function @xmath100 .
furthermore , @xmath100 inherits the symmetries of @xmath93 .
the integral ( [ 5.3 ] ) exhausts the irrelevant case .
suppose that @xmath101 and @xmath102 are two different smooth solutions to ( [ 5.2 ] ) .
their difference @xmath103 obeys the homogeneous equation @xmath104 and is consequently a homogeneous function in @xmath105 of degree @xmath106 .
but regularity at @xmath107 demands @xmath108 , when @xmath97 .
+ _ let @xmath97 and @xmath95 .
then we have that : ( 1 ) there exists a unique solution @xmath100 of the pde ( [ 5.1 ] ) in the space @xmath109 .
( 2 ) it is given by the integral ( [ 5.3 ] ) . _ + consider then the case @xmath110 .
the integral ( [ 5.3 ] ) now fails to converge unless @xmath93 provides a sufficiently high power of @xmath105 .
this can be achieved by a taylor expansion of order @xmath106 .
the remainder term behaves as @xmath111 and compensates the power @xmath112 in ( [ 5.3 ] ) .
we thus expand @xmath113 where @xmath114 is an integer valued multi
index , @xmath115 , @xmath116 , and @xmath117 .
taking derivatives of ( [ 5.1 ] ) , we find that @xmath118 each @xmath105derivative thus reduces @xmath106 by one unit
. the derivatives of order @xmath119 in ( [ 5.5 ] ) therefore fall into the irrelevant case . from ( [ 5.3 ] ) it follows that @xmath120 for all derivatives of order @xmath121 ( in fact @xmath122 ) . the remainder term in ( [ 5.5 ] )
then involves a convergent integral .
it yields a smooth function .
the taylor coefficients in ( [ 5.5 ] ) on the other hand are determined by evaluation of ( [ 5.6 ] ) at @xmath107 .
we have that @xmath123 for all solutions to ( [ 5.6 ] ) which are regular at @xmath107 .
( [ 5.8 ] ) determines all relevant taylor coefficients with @xmath124 ( in fact all non marginal ones with @xmath125 ) .
the marginal taylor coefficients are not only undetermined by ( [ 5.8 ] ) but also impose the condition @xmath126 on the taylor coefficients of @xmath93 of order @xmath106 .
smooth solutions exist only if the pde ( [ 5.6 ] ) satisfies ( [ 5.9 ] )
. if this is the case , as will be supposed here , then the taylor coefficients of @xmath100 of order @xmath106 are free parameters . in the perturbation theory for the @xmath0trajectory
, we will use this freedom of the marginals at a given order of perturbation theory to satisfy the conditions on the existence of smooth solutions one order below . in this sense ,
the condition ( [ 5.9 ] ) on the marginal taylor coefficients determines both the @xmath3function and the wave function renormalization .
+ _ let @xmath110 and @xmath95 .
then we have that : ( 1 ) the pde ( [ 5.1 ] ) has smooth solutions @xmath100 if and only if all derivatives of @xmath93 of order @xmath106 satisfy ( [ 5.9 ] ) .
let this be the case .
( 2 ) then @xmath100 can be written in the form of a taylor expansion ( [ 5.5 ] ) of order @xmath106 .
the taylor remainder follows from ( [ 5.7 ] ) .
the taylor coefficients of order @xmath106 are free parameters .
the taylor coefficients of order less than @xmath106 follow from ( [ 5.8 ] ) . _ + there is a neat reformulation for the non irrelevant case in terms of subtracted functions .
( [ 5.1 ] ) is equivalent to @xmath127 smooth solutions to ( [ 5.10 ] ) also obey @xmath128 for all @xmath129 . in particular , we have the identities @xmath130 as polynomials in @xmath105
. we can then define subtracted functions @xmath131 from ( [ 5.10 ] ) and ( [ 5.12 ] ) it follows that they satisfy the subtracted version of ( [ 5.10 ] ) , namely @xmath132 but the subtracted function has the property @xmath133 for all @xmath94 .
therefore , the integral @xmath134 converges for @xmath135 .
the taylor remainder can thus be integrated as in the irrelevant case .
renormalized perturbation theory is the inductive solution of the scaling equations ( [ 4.6 ] ) .
we have included a separate treatment of the first and second order scaling equations to illustrate the method . in the functional representation , the scaling equation ( [ 4.1 ] )
requires for @xmath136 that @xmath137 in other words , @xmath138 has to be an eigenvector of @xmath139 , and @xmath140 has to be the eigenvalue .
the @xmath0vertex @xmath141 is an eigenvector of @xmath139 .
its eigenvalue is @xmath89 . in four dimensions ,
the @xmath0vertex becomes marginal .
the wave function term @xmath142 is marginal in any dimension .
we therefore conclude that @xmath143 is in agreement with ( [ 6.1 ] ) , whence @xmath144 .
the first order wave function parameter @xmath58 is not determined by ( [ 6.1 ] ) . the first order equation ( [ 6.1 ] ) is special as it is homogeneous .
all higher order equations are inhomogeneous .
to second order , ( [ 4.1 ] ) reads @xmath145 where @xmath146 eq . ( [ 6.6 ] ) contains two unknowns , @xmath147 and @xmath58 .
they belong to the two marginal couplings . in the momentum space kernel representation , ( [ 6.5 ] ) becomes @xmath148 with @xmath149 where @xmath150 denotes convolution in momentum space times @xmath151 .
the kernels are understood to be symmetrized in their entries and to be restricted to the hyperplane of zero total momentum .
put @xmath152 for instance .
the constant @xmath153 in ( [ 6.8 ] ) stands for the convergent one loop integral @xmath154 all higher kernels with @xmath155 are zero in accordance with ( [ 3.4 ] ) .
the six point kernel has scaling dimension @xmath156 and is irrelevant . from ( [ 5.3 ] ) we learn that it is given by .
this expression is regular at zero momentum and a bounded function of the momenta .
it has the form of a cutoff propagator . ]
@xmath157 notice that ( [ 6.10 ] ) is indeed a smooth function of the momenta .
the four point kernel has scaling dimension @xmath158 and is marginal .
it requires a separate treatment of its zero momentum value , the four point coupling .
we determine @xmath147 such that @xmath159 from ( [ 6.9 ] ) it follows that the condition ( [ 6.13 ] ) reads @xmath160 notice that it is independent of @xmath58 .
the value of the second order @xmath0coupling is not determined by renormalization invariance but rather by the choice of the expansion parameter .
a natural choice is ( [ 3.8 ] ) .
the four point kernel is then equal to its first subtraction .
it is therefore integrated to @xmath161 the integral converges due to ( [ 6.13 ] ) .
the flow of the coupling parameter thus saves us from a logarithmic singularity of ( [ 6.15 ] ) .
we remark that , in three dimensions , the four point kernel is already irrelevant to second order and needs no subtraction at all .
the quadratic kernel finally has scaling dimension @xmath162 and is relevant .
it calls for a taylor expansion of order two with remainder term . using euclidean invariance
, we write @xmath163 furthermore , we trade @xmath164 for a new variable @xmath165 .
the scaling equation for the quadratic kernel then takes the form @xmath166 to solve it , we expand @xmath167 in a taylor formula @xmath168 the taylor coefficients follow from evaluating ( [ 6.18 ] ) and its @xmath165derivatives at @xmath169 .
we find @xmath170 with a convergent two loop integral , which comes out as @xmath171 the other taylor coefficient is the second order wave function parameter @xmath172 we leave it undetermined for the moment .
taking a @xmath165-derivative of ( [ 6.18 ] ) , it follows that @xmath173 regularity of @xmath174 at @xmath169 requires that @xmath175 this condition determines the first order wave function parameter .
its value in four dimensions is @xmath176 eq .
( [ 6.24 ] ) explains the presence of a wave function term in ( [ 3.3 ] ) .
it is necessary to fulfill the condition under which smooth solutions of the second order equation ( [ 6.18 ] ) exist .
the second order wave function parameter ( [ 6.22 ] ) is then fixed by the condition on the third order pendant to ( [ 6.24 ] ) .
the value of the second order mass parameter ( [ 6.20 ] ) is then @xmath177 a second @xmath165derivative turns ( [ 6.23 ] ) into @xmath178 the second derivative is thus irrelevant . eq .
( [ 6.27 ] ) is equivalent to @xmath179 therefrom we conclude that @xmath180 the integrals ( [ 6.19 ] ) and ( [ 6.29 ] ) will not be evaluated .
we content ourselves with the ascertainment that they converge and yield a smooth remainder term .
the second order scheme is now complete .
the second order scheme generalizes to all higher orders . to order @xmath59 , we first compute @xmath181 , then @xmath182 , and thereafter @xmath183 , except for @xmath184
. then we proceed to the order @xmath185 .
each step consists of solving a system of scaling equations ( [ 4.6 ] ) for the order @xmath59 momentum space kernels . the momentum space kernels with non
negative scaling dimension will have to be taylor expanded . in four dimensions , the quadratic needs to be expanded to second order , whereas the quartic kernel needs to be expanded to first order . a convenient way of writing the taylor expansions is in terms of projectors @xmath186 they are defined to act linearly on polynomial functionals of the form ( [ 3.4 ] ) .
we have that @xmath187 with coupling parameters @xmath188 and local vertices @xmath189 they are all eigenvectors of @xmath190 with eigenvalues @xmath191 when applied to ( [ 4.1 ] ) , the projectors yield the scaling equations @xmath192 by means of ( [ 4.2 ] ) , these are given by @xmath193 eqs .
( [ 6.45 ] ) , ( [ 6.46 ] ) , and ( [ 6.47 ] ) are linear equations for the coupling parameters ( [ 6.36 ] ) , ( [ 6.37 ] ) , and ( [ 6.38 ] ) .
they are algebraic equations rather than first order differential equations . in four dimensions , ( [ 6.47 ] ) reads @xmath194 since @xmath144 .
the choice ( [ 3.8 ] ) of the coupling parameter means @xmath195 . from ( [ 6.50 ] ) we then have @xmath196 its right hand side is known from the induction hypothesis , and does not depend on @xmath182 because @xmath197 due to the regularity of @xmath198 at @xmath107 .
( [ 6.52 ] ) determines @xmath181 .
consider then ( [ 6.46 ] ) .
since @xmath144 , it becomes @xmath199 from ( [ 6.49 ] ) it follows that @xmath200 the right hand side is again independent of @xmath182 because @xmath201 but then ( [ 6.35 ] ) determines @xmath182 .
the last coefficient is the effective mass parameter .
( [ 6.45 ] ) tells that @xmath203 and using ( [ 6.48 ] ) we then have @xmath204 the non irrelevant part of @xmath183 is now complete , except for the wave function parameter @xmath184 which we leave undetermined until the next order . the irrelevant part is directly integrated . for @xmath205 , the momentum space scaling dimension ( [ 4.8 ] )
is negative .
those kernels are therefore all irrelevant .
the scaling equation ( [ 4.6 ] ) is in this case integrated to @xmath206 all kernels with @xmath207 are zero .
the integrals converge due to the negative power counting .
the zero momentum value of the four point kernel has been transfered to the differential @xmath3-function through ( [ 6.51 ] ) .
its subtracted remainder is irrelevant and integrated to @xmath208 the integral converges due to the subtraction at zero momentum , finally , the two point kernel is reconstructed with the help of @xmath209 and @xmath210 through @xmath211 the iterative scheme is now complete , aside of the question of the convergence and smoothness of the right hand side of ( [ 4.6 ] ) .
the bilinear renormalization group bracket is responsible for inhomogeneous terms in our scaling equations . we write it out explicitly for even monomials @xmath212 of the general form
@xmath213 applied to two monomials of this form ( [ 7.1 ] ) , the bilinear bracket decomposes into @xmath214 and is itself a sum of monomials @xmath215 whose kernels are given by a multiple convolutions @xmath216 the kernels are understood to be symmetric under permutations of their entries .
the multiple convolution in ( [ 7.4 ] ) always involves one hard propagator @xmath217 and @xmath218 soft propagators @xmath54 , which is at the same time the number of loops .
furthermore , ( [ 7.2 ] ) involves a combinatorial factor @xmath219 coming from the number of ways in which the contractions can be made .
( [ 7.4 ] ) can be interpreted as a fusion of two vertices . in the process of fusion links
are created , consisting of propagators .
we present an elementary estimate on this fusion product .
the estimate uses the @xmath65-norm in momentum space .
the estimate works at this point for @xmath220 .
later it will be used for @xmath67 only . notice to begin with that @xmath221 for the propagators with exponential cutoff . and @xmath222- and @xmath223-norms in momentum space .
] at @xmath224 we have for instance @xmath225 and @xmath226 .
if the fourier transformed kernels now satisfy the bounds @xmath227 that is , are finite in the @xmath65-norm ,- norm for the momentum space kernels is defined as @xmath228 with @xmath229 the hyperplane of total zero momentum . ]
then it follows that all summands ( [ 7.4 ] ) in the decomposition ( [ 7.2 ] ) of the bilinear operation have finite @xmath65-norms in momentum space .
they obey @xmath230 therefore , the renormalization group flow preserves the @xmath65-norm of momentum space kernels .
the estimate immediately follows from the fourier transform @xmath231 of ( [ 7.4 ] ) .
the @xmath62-functions from translation invariance have been removed .
the sums of momenta in the kernels are zero through @xmath232 the idea with the parameter @xmath233 is to use part of the exponential large momentum decay of the fluctuation and normal ordering propagators to compensate a possible large momentum growth of the kernels . in an initial value problem for a scale dependent renormalization group flow with @xmath234bounded initial data this might seem unnecessary .
for instance , a pure @xmath0vertex is given by a constant kernel and is thus @xmath234bounded .
the evolution preserves the @xmath234bound for all _ finite _ scales @xmath13 .
however , we can not expect the solution to be @xmath234bounded _ uniformly _ in @xmath13 .
the limit @xmath235 requires a separate treatment of zero momentum derivatives and taylor remainders of the non - irrelevant kernels .
the price to pay is a growth in momentum space . in the four dimensional case it is polynomial in powers and logarithms of momenta . with an exponential bound we are very far on the safe side .
we prove a large momentum bound for the solution to the renormalization group pde ( [ 5.1 ] ) under the assumption of a large momentum bound on the inhomogeneous side .
the bound uses the @xmath65norm for @xmath67 . although
being rather wasteful , it suffices to prove finiteness of the bilinear renormalization group bracket .
let @xmath97 .
suppose that the function @xmath93 in ( [ 5.1 ] ) has a finite @xmath65norm @xmath236 its solution ( [ 5.3 ] ) then inherits an @xmath65bound .
from @xmath237 it follows that @xmath238 eq .
( [ 8.3 ] ) shows that the irrelevant solution to the renormalization group pde is not only finite but also decreases in the @xmath65-norm .
let @xmath239 . in this case
we assemble @xmath100 using a first order taylor formula .
suppose then that we have @xmath65-bounds on the first derivatives @xmath240 if @xmath100 is marginal , then its first derivatives are irrelevant with scaling dimension minus one .
it follows that @xmath241 therefrom it follows that @xmath242 the result is an @xmath65-bound @xmath243 this estimate is not uniform in @xmath233 .
it works for @xmath233 arbitrary small , but the bound grows with an inverse power of @xmath233 .
the large momentum growth is a consequence of the split in derivatives and taylor remainder .
let @xmath244 .
this case requires a generalization of the bound in the marginal case .
the taylor expansion is pushed to order @xmath245 .
then the derivatives become irrelevant .
we assume @xmath65-estimates on all derivatives @xmath246 of order @xmath247 .
since they are irrelevant with scaling dimension minus one it follows that the corresponding derivatives of @xmath100 obey @xmath248 and are also @xmath65-bounded . from the taylor formula
it then follows that @xmath249 and thus @xmath250 with constants @xmath251 thus we again have an @xmath65-bound on the function @xmath100 .
this completes the large momentum bound on @xmath100 .
exactly the same strategy applies to the derivatives of @xmath100 as well .
the irrelevant derivatives inherit immediately large momentum bounds .
the relevant derivatives require taylor expansions .
we omit to spell out explicitely the necessary bounds on the derivatives of @xmath93 .
the iterative scheme determines order by order @xmath252 , @xmath253 , @xmath254 , and the irrelevant remainders @xmath255 .
it is finite to all orders of perturbation theory because of the following iteration of regularity .
suppose that we have shown the following to all orders @xmath256 : + _ i ) @xmath252 , @xmath253 , and @xmath254 are finite numbers .
ii ) @xmath257 is a smooth function on @xmath258 for all @xmath259 , symmetric in the momenta , and @xmath260-invariant .
iii ) @xmath261 is finite for all @xmath67 , @xmath259 , and @xmath262 . here
@xmath263 is a multi - index which labels momentum derivatives . _
+ then the same statements hold at order @xmath264 . since they are trivially fulfilled to order one they iterate to all orders of perturbation theory . to prove the iteration of regularity we once more inspect each step of the iterative scheme .
first , the irrelevant remainders @xmath265 are smooth functions on @xmath258 , symmetric under permutations and @xmath260-invariant .
they and all their momentum derivatives satisfy @xmath65-bounds .
they are composed of two contributions .
the first immediately inherits a bound from the induction hypothesis .
the second is a sum of renormalization group brackets of lower orders .
therefore , they consist of multiple convolutions with propagators .
the integrals converge , are smooth functions of the external momenta , and satisfy @xmath65-bounds .
second , we have linear equations for the coefficients @xmath266 , @xmath267 , and @xmath202 with finite coefficients . third , the integration of the inhomogeneous renormalization group pdes , yields solutions with the desired properties .
the aim of perturbative renormalization theory is to derive power series expansions for green s functions which are free of divergencies .
the bphz theorem states that this can be accomplished by writing the green s functions in terms of renormalized parameters .
an elegant proof of the bphz theorem was given by callan @xcite
. a polished version of which is due to lesniewski @xcite .
their method is similar to ours in that it is based on renormalization group equations for the renormalized green s functions , the callan - symanzik equations .
the method proposed here is different in that it does _ not _ resort to any kind of graphical analysis , not to analysis of sub - graphs , and not to skeleton expansions .
a new generation of proofs of the bphz theorem was initiated with the work of polchinski @xcite .
his proof has been simplified further by keller , kopper , and salmhofer @xcite .
our method is similar in that it uses an exact renormalization differential equation .
the details are however quite different .
the main difference is that polchinski begins with a cutoff theory .
he then shows how the cutoff can be removed in a way such that the effective interaction remains finite .
our method directly addresses the limit theory without cutoffs , expressed in terms of a renormalization group transformation with cutoffs . roughly speaking , we are here simultaneously changing polchinski s renormalization conditions and integrating an amount of fluctuations . unlike polchinski and followers
we use a renormalization group differential equation with dilatation term .
a way to think of ( [ 1.5 ] ) is as an equation for a renormalization group fixed point of a system which has been enhanced by one degree of freedom , the running coupling .
this fixed point problem can only be formulated with rescaling and with dilatation term .
another renormalization group approach to renormalized perturbation theory comes from gallavotti @xcite and collaborators .
pedagocial accounts of tree expansions can be found in @xcite . there
the result of renormalization is expressed in terms of a renormalized tree expansion .
the program of @xcite with an iterated transformation with fixed @xmath13 is related to the tree expansion .
both are built upon a cumulant expansion for the effective interaction .
the renormalization procedure is however quite different .
like polchinski , gallavotti starts from a cutoff theory .
it is organized in terms of trees , which describe the sub - structure of divergencies in feynman diagrams .
the divergencies are transformed into a flow of the non - irrelevant couplings .
this part is similar to ours .
the basic difference with gallavotti is that we do not organize our expansion in terms of trees .
a hybrid approach between polchinski and gallavotti is due to hurd @xcite .
an important question is whether this constuction of renormalized trajectories extends beyond perturbation theory .
another important question is whether it extends to renormalized trajectories at non - trivial fixed points .
we hope to return with answers to these questions in the future .
xxxxx g.benfatto , g.gallavotti , renormalization group , physics notes no . 1 , princeton university press 1995 c.g.callan , introduction to renormalization theory , les houches lecture notes 1975 , 41 - 77 , r.balian and j.zinn-justin eds .
j.s.feldman , t.r.hurd , l.rosen , j.d.wright , qed : a proof of renormalizability , lecture notes in physics 312 , springer verlag 1988 g.gallavotti , renormalization theory and ultraviolet stability for scalar fields via renormalization group methods , rev .
57 no . 2 ( 1985 )
471 - 562 g.gallavotti and f.nicolo , renormalization in four dimensional scalar fields i , commun .
( 1985 ) 545 - 590 ; renormalization renormalization in four dimensional scalar fields ii , commun .
( 1985 ) 247 - 282 j.glimm and a.jaffe , quantum physics , springer verlag 1987 t.hurd , a renormalization group proof of perturbative renormalizability , commun .
124 ( 1989 ) 153 - 168 g.keller , c.kopper , m.salmhofer , perturbative renormalization and effective lagrangeans , mpi - pae / pth 65/90 a.lesniewski , on callan s proof of the bphz theorem , helv .
acta , vol .
56 ( 1983 ) 1158 - 1167 j.polchinski , renormalization and effective lagrangeans , nucl .
b231 ( 1984 ) 269 - 295 k.wilson , renormalization group and critical phenomena i and ii , phys . rev . b4 ( 1971 ) 3174 - 3205 c.wieczerkowski , the renormalized @xmath1-trajectory by perturbation theory in the running coupling i : the discrete renormalization group , hep - th/9601142 k.wilson and j.kogut , the renormalization group and the @xmath233-expansion , phys .
c12 no . 2 ( 1974 ) 75 - 200 | the renormalized trajectory of massless @xmath0theory on four dimensional euclidean space
time is investigated as a renormalization group invariant curve in the center manifold of the trivial fixed point , tangent to the @xmath0interaction .
we use an exact functional differential equation for its dependence on the running @xmath0coupling .
it is solved by means of perturbation theory .
the expansion is proved to be finite to all orders .
the proof includes a large momentum bound on amputated connected momentum space green s functions .
the renormalized @xmath1-trajectory by + perturbation theory in a running coupling ii : + the continuous renormalization group + c. wieczerkowski + institut fr theoretische physik i , universitt mnster , + wilhelm - klemm - strae 9 , d-48149 mnster , + wieczer@@ uni-muenster.de + ms - tp1 - 96 - 14 |
the layered transition - metal dichalcogenides ( tmds ) have been investigated for over fifty years now .
however , recent interest in this class of materials has been spurred by possible applications in nanoelectronics , photovoltaics , catalysis to name a few @xcite .
further advances in the growth of one or more layers @xcite has led to the exploration of the properties of these materials as a function of thickness .
analogous to the nanomaterials where one finds a size dependence of the band gap @xcite , one finds thickness dependent changes in the electronic structure of the layered transition metal dichalcogenides .
additionally one finds a thickness dependent band gap which changes character .
the bulk band gap of mos@xmath0 is found to be an indirect one of 1.3 ev @xcite which increases to 1.6 ev in the bilayer limit @xcite .
the nature of the band gap changes and becomes a direct one of 1.9 ev at the monolayer limit @xcite .
the fact that the monolayers of transition metal dichalcogenides have a direct band gap ( with the exception of wse@xmath0 @xcite ) is evident from the sharp peak that one finds in the photoluminescence spectra@xcite .
mose@xmath0 also has an indirect band gap of 1.1 ev @xcite in the bulk limit whereas in the monolayer limit it has direct band gap of about 1.66 ev @xcite .
there could be different types of mx@xmath0 ( m = mo , w , ti etc .
, x = s , se , etc . )
sandwiches depending on the coordination of the transition metal atom with the chalcogens as well as the stacking of atoms @xcite . in this work
we focus our attention on the 2h polymorph of mox@xmath0 ( x = s , se ) where the symmetry about the mo site is trigonal prismatic , though certain generic features are found to be valid across different types of stacking .
the bonding within each monolayer is strongly covalent .
however , the coupling between layers is believed to be due to weak van der waal s interaction . the size dependence of the band gap in semiconductor quantum dots has been extensively studied and is well understood .
it has been interpreted as arising because the electron and hole wavefunction begin to feel the effects of the boundary and hence a band gap different from what is found in the bulk for these materials .
this has been referred to as quantum confinement effect on the band gap and happens in a size range less than the bohr radius for the material .
consequently , when electronic structure calculations found a thickness dependence of the band gap in transition metal dichalcogenides , this was interpreted as arising due to quantum confinement effects .
this view has prevailed in the community in the absence of any proper quantification .
@xcite fit the _ ab - initio _ band structure for a monolayer of mos@xmath0 to a nearest neighbor tight binding model . using the same parameters within each layer and introducing interactions between the layers , the _ ab - initio _ band structure for bulk mos@xmath0 was fit and the relevant parameters extracted . from this model
, it was inferred that interlayer interactions were responsible for the direct to indirect band gap transition in these materials .
however , only a subset of bands were fit in addition to a very narrow energy window .
the problem with fitting a small subset of bands is that they allow you to arrive at any conclusion you would like .
the issue still does nt seem to be settled . recently
zhang and zunger @xcite examined various aspects of the thickness dependent changes in the electronic structure and arrived at the conclusion that both interlayer interactions as well as quantum confinement effects were responsible for the observed changes .
su - huai wei and coworkers @xcite attributed the transition to the absence of interlayer interactions at k point , a general feature they attributed to the crystal symmetry of hexagonal materials . in view of the differing viewpoints existing in the literature ,
there is a need to reexamine this problem .
the direct band gap originates from a transition from the valence band maximum at k to the conduction band minimum at the same point .
this was found to show a weak dependence on size .
the indirect band gap arises from a transition from the top of the valence band at @xmath1 to the bottom of the conduction band between @xmath1 and k , and exhibits a strong dependence on the number of layers . in order to examine the origin of the observed variations in the band gap with size , padhila
_ @xcite chose an internal reference in each case and monitored the size dependence of the band extrema points .
they found that while the band extrema at k point hardly showed any size dependence , the band extrema at @xmath1 and t showed significant variations with thickness which determined the band gap .
examining the charge density associated with the band extrema at @xmath1 point , they found that they were contributed by orbitals directed out of plane , hence explaining the changes in the position of the band extrema as each additional layer was added . in this manuscript
we have studied the variation of the band gap in mose@xmath0 as a function of number of layers .
the band structures obtained from _ ab - initio _ electronic structure calculations are found to reproduce the experimental observation @xcite that this system is a direct band gap semiconductor in the monolayer limit , but as we move to the bilayer , the system becomes an indirect band gap material .
as mentioned earlier , the evolution in the electronic structure of the transition metal dichalcogenides with thickness have been discussed in terms of geometric confinement effects , interlayer interactions with an important role played by crystal symmetry in determining the nature of interactions . in order to understand the role of each of these quantities in bringing about the band gap crossover
, our approach has been to carry out an _ ab - initio _ mapping onto a tight binding model .
this method uses maximally localized wannier functions as its basis functions .
geometric confinement effects are expected to renormalize the onsite energies of the orbitals which are directed into the vacuum as against those orbitals which are in - plane , in line with the belief that these orbitals feel the effect of the boundary .
additionally , one knows that as one builds a solid atom by atom , the onsite energies get renormalized because of the potential felt by the electrons due to the neighboring atoms ( madelung effect ) . as a result
there should be a change in the onsite energies as one went from monolayer to bilayer to trilayer .
the tight binding fittings carried out for monolayer and bilayer are found to have similar onsite energies .
hence one concluded that there was no renormalization of the energies by geometric confinement or madelung potential effects .
the changes in the electronic structure as a function of the number of layers arose from interlayer interactions present when a layer was added to a monolayer . to check this point
we used the tight binding hamiltonian of the bilayer , switched off interlayer interactions and recovered the monolayer band structure , as one would expect .
a similar construction of the trilayer band structure within the tight binding model was carried out using the monolayer hamiltonian for the layers and the interlayer interactions derived from the bilayer hamiltonian .
the electronic structure constructed this way gave perfect match with the _ ab - initio _ band structure .
hence our results showed quantitatively that interlayer interactions were responsible for the changes in the electronic structure that one found as a function of the number of layers , with geometric confinement effects playing a negligible role .
similar conclusions were arrived at for the other transition metal dichalcogenides mx@xmath0 where m = mo , w and x = s , se , te also .
while the 2h stacking was considered in each of the cases listed above , considering alternate stacking did not change the onsite energies . this therefore led us to the rather puzzling conclusion that interlayer interactions dictated the electronic structure changes even in these van de waal s materials .
while the hexagonal symmetry of the lattice is attributed for the absence of interlayer interactions at the k point , we show that it is the character of the bands which determines the presence / absence of interlayer interactions at a symmetry point .
the electronic structure calculations of monolayer , bilayer as well as trilayer mose@xmath0 as well as monolayer and bilayer mos@xmath0 are carried out within a plane wave implementation of density functional theory ( dft ) using the vasp @xcite code .
we have taken the 2h stacking of mose@xmath0 as it is found to be the most stable structure @xcite . while , the lattice constants are kept at the experimental values of @xmath2=3.299 and @xmath3=12.939 for mose@xmath0 @xcite and @xmath2=3.16 and c=12.296 for mos@xmath0 @xcite , all the atoms are allowed to relax through a total energy minimization that is guided by the calculated atomic forces . a vacuum of 20 is used along @xmath4 direction to minimize the interaction among the periodic images .
projected augmented wave @xcite potentials are used to solve the electronic structure self - consistently using a k - points mesh of 12@xmath512 @xmath51 with a cutoff energy for the plane wave basis states equal to 280 ev .
perdew - burke - ernzerhof ( pbe ) @xcite approximation was used for the exchange - correlation functional .
there is a weak van der waals interaction between the layers which has an effect in the determination of the interlayer distances .
a dispersion correction based on grimme s dft - d2 method @xcite is used on top of the pbe potentials . in order to quantify the results , we setup a tight binding model with mo @xmath6 and se @xmath7/s @xmath7 states in the basis . in this model , the maximally localized wannier functions @xcite are used for the radial parts of the basis functions .
technically , the degree of localization and the symmetry of these wannier functions can be controlled in the projection procedure .
all on - site energies and hopping interaction strengths in this case are determined from the interface of vasp to wannier90 @xcite .
apart from the 2h stacking , we also explored other stacking geometries aa , a@xmath8b , ab@xmath8 and ab to explore the renormalization of the onsite energies due to differing madelung potentials .
an explanation for the notation for the stacking may be found in ref .
@xcite . in order to generalize our findings
, we did the similar analysis mentioned earlier for mote@xmath0 , ws@xmath0 , wse@xmath0 and wte@xmath0 .
the experimental lattice parameters of a= 3.522 , 3.1532 , 3.282 , and c= 13.967 , 12.323 , 12.96 , were considered for mote@xmath0 , ws@xmath0 , wse@xmath0 and wte@xmath0 @xcite respectively .
the internal positions were optimized in each case for both monolayer and bilayer .
the van der waals interactions implemented using the dft - d2 method of grimme were considered for positional optimization in the case of bilayer .
the @xmath9 band structure was then calculated for each of these cases and the results for the bilayers were then mapped onto a tight binding model with mo / w @xmath6 and s , se and te @xmath7 orbitals in the basis . the radial part of the tight binding basis functions corresponded to maximally localized wannier functions
the _ ab - initio _ band dispersions for monolayer mose@xmath0 along various symmetry directions are plotted in fig .
[ fig:1 ] .
one finds that the valence band maximum ( vbm ) and the conduction band minimum ( cbm ) are both located at k point and the system is a direct band gap semiconductor .
this is consistent with experiment which also finds the system to be a direct band gap semiconductor with a gap of 1.66 ev .
the present calculations which use gga for the exchange correlation functional find a gap of 1.59 ev which is close to the experimental value .
the agreement is however fortuitous as one usually has an underestimation of the band gap due to self - interaction effects among various other approximations which enter the use of the generalized gradient approximation in the absence of an exact exchange correlation functional . in order to quantify the changes in the electronic structure , we have mapped the _ ab - initio _ band structure onto a tight binding model with mo @xmath6 and se @xmath7 states in the basis .
the tight binding band structure shown by red line with circles is superposed on the calculated _ ab - initio _ band structure in fig .
[ fig:1 ] .
we have a good description of the _ ab - initio _ band structure in the energy window from -3.5 ev to 4 ev .
this gives us confidence in the extracted parameters and allows us to discuss changes in the electronic structure in terms of these parameters .
there are several ways to construct the bilayer of mose@xmath0 .
each monolayer can be visualized as a three atomic layer stacking of mo and se atoms where mo atoms are sandwiched between layers of se atoms .
the se atoms generate a trigonal prismatic crystal field at the mo site .
the stacking that we have considered has the mo atom in one layer above that in the layer beneath .
however , the mo - se motif is rotated by 180@xmath10 in the upper layer with respect to the layer beneath .
this is referred to as the aa@xmath8 stacking and has been shown to have the lowest energy among various stacking patterns considered . considering a bilayer of mose@xmath0
, we have calculated the band dispersions along various symmetry directions .
this is shown in fig .
[ fig:2 ] .
we find that the vbm which was at k point has now shifted to @xmath1 .
the cbm is also shifted to t point which lies along the line from @xmath1 to k. this leads to an indirect band gap of 1.25 ev in contrast to the experimental value of 1.55 ev @xcite .
the changeover in the vbm positions can easily be understood by examining the character of the states contributing to this point .
this is shown in fig .
[ fig:3 ] where we plot the charge density for the highest occupied band at @xmath1 in panel ( a ) .
these are seen to emerge from the interactions between the @xmath11 orbitals on mo and @xmath12 orbitals on se . as these involve orbitals which are directed out of plane
, one finds that these levels in the lower layer interact with the @xmath11/@xmath12 orbitals in the layer above . as a result , the highest occupied band at @xmath1
point moves to higher energies relative to that at k point , and consequently the vbm shifts to @xmath1 point when we move from monolayer to bilayer .
the highest occupied band at the k point is contributed by interactions between in - plane orbitals as is evident from fig .
hence it shows no shift from the bilayer from its position for the monolayer . a similar analysis of the charge density contributing to the lowest unoccupied band at t and
k symmetry points is shown in panels ( c ) and ( d ) of fig .
[ fig:3 ] .
we find that in - plane orbitals contribute to the lowest unoccupied band at k point while out - of - plane orbitals contribute to the lowest unoccupied band at t point .
hence , the increased interaction arising from the presence of the second layer moves t point relative to k point , making the former the conduction band minimum .
these ideas are consistent with the analysis of padhila _ et al .
_ @xcite who examined the movement of various band extrema as a function of the number of layers .
while these ideas are qualitative , we map the @xmath9 band structure to a tight binding model and extract the onsite energies and hopping interaction strengths in order to make a more quantitative statement of the role of various effects .
the tight binding band structure is superposed on the _ ab - initio _ band structure calculated for the monolayer in fig.[fig:1 ] and for the bilayer in fig .
[ fig:2 ] .
the quality of the fit in both cases gives us confidence in the extracted parameters ( table [ onsite ] ) which we use to discuss changes in the electronic structure in going from the monolayer to the bilayer .
the onsite energies for se @xmath7 as well as mo @xmath6 orbitals extracted by the tight binding mapping for monolayer as well as bilayer mose@xmath0 are given in table [ onsite ] .
the energies are similar within deviations associated with the error bars on the energies . hence we can conclude that geometric confinement effects do not modify the onsite energies when one goes from monolayer to bilayer . in order to examine what is it that changes leading to the differences in the electronic structure in going from monolayer to bilayer , we have considered the tight binding hamiltonian for the bilayer .
interlayer interactions have been switched off in this hamiltonian and the ensuing band structure has been plotted along various symmetry directions in fig .
[ fig:4 ] . the band structure for the monolayer
has been superposed for comparison .
the two band structures look almost identical suggesting that the only difference between the electronic structure of the monolayer and that for the bilayer emerge from interlayer interactions .
the dominant interlayer hopping interaction strengths of bilayer mose@xmath0 are between the first nearest neighbor mo-@xmath13 and se-@xmath12 ( 4.88 ) and also between the se-@xmath7 orbitals which are first ( 3.72 ) as well as second ( 5.92 ) neighbors . in order to examine this hypothesis
further , we constructed a trilayer heterostructure of mose@xmath0 .
the _ ab - initio _ band structure for the trilayer was calculated along various symmetry directions .
this is shown in fig .
[ fig:5 ] . for comparison and to examine the hypothesis made vis - a - vis the origin of the changes in the electronic structure as each layer is added , we set up the tight binding hamiltonian for the trilayer .
this was done by considering the hamiltonian for the monolayer for each of the layers .
the interlayer interactions extracted for the bilayer were then used to couple the layers .
the band structure calculated within this model is superposed on the _ ab - initio _ band structure for the trilayer in fig .
[ fig:5 ] .
the comparison is excellent , justifying our hypothesis .
an important aspect emerges from this study , which is that inspite of the widely used nomenclature of van der waal s heterostructures for few layer mose@xmath0 , one finds that there is significant covalent interaction between the layers .
there have been alternate explanations offered in the literature .
considering the crystal symmetry , su - huai wei and coworkers @xcite have suggested that it is the underlying symmetry that dictates whether the interlayer matrix elements will be finite or not for the band extrema at k - point .
however , we point out that depends entirely on the character of the wavefunction contributing to the band extrema .
a comparison of the _ ab - initio _ band structure for the monolayer and the bilayer ( fig.[fig:6 ] ) which have been aligned at the highest occupied band at k - point reveals the presence of other bands at k -point which are contributed by orbitals with mo - @xmath14,@xmath15 and se - @xmath16 , @xmath17 character which are perturbed in going from monolayer to bilayer .
so , the @xmath6-filling determines the character of the highest occupied band at k - point
. there could be other hexagonal compounds with entirely different bands contributing to the band extrema .
while we have examined the onsite energies in going from monolayer to bilayer for the 2h stacking , there are other stackings possible . in order to examine whether our conclusions were general enough , we considered bilayers with aa , a@xmath8b , ab@xmath8 and ab stacking with the nomenclature used as given in ref.@xcite . in each case
the _ ab - initio _ band structure was fit to a tight binding model with mo @xmath6 and se @xmath7 states in the basis .
the extracted onsite energies are given in table [ mose3_all_stack ] .
these onsite energies for different stacking of layers in mose@xmath0 bilayer look just like what we found in the case of 2h stacking ( table [ onsite ] ) .
this analysis clearly shows that renormalization of onsite energies does not occur for differently stacked bilayer structures of these materials .
the evolution of the electronic structure with number of layers that we find here is not specific to mose@xmath0 alone .
we find similar changes when we have examined the electronic structure as a function of the number of layers for mos@xmath0 also . in fig .
[ fig:7 ] we have plotted the _ ab - initio _ band structure for the monolayer in panel ( a ) and for the bilayer in panel ( b ) . the tight - binding band structure has been superposed in each case and we have a good description of the _ ab - initio _ band structure .
geometric confinement effects are expected to modify the onsite energies of the orbitals directed out of plane as against those which lie in plane .
a comparison of the extracted onsite energies is given in table [ onsitemos2 ] . here
again , we find that the energies for the monolayer and bilayer are similar .
this shows that geometric confinement effects are not responsible for the changes in the electronic structure . when we considered the tight binding hamiltonian for the bilayer and switched off interlayer interactions , we recovered the monolayer band structure .
the comparison between the model hamiltonian results with interlayer interactions switched off and the _ ab - initio _ band structure for the monolayer is given in fig .
[ fig:8 ] . in order
to demonstrate that the conclusions obtained from our analysis of mose@xmath0 and mos@xmath0 are general , we have considered a monolayer as well as bilayers of mote@xmath0 , ws@xmath0 , wse@xmath0 and wte@xmath0 .
the @xmath9 band dispersions calculated for the monolayer of each of these systems along various symmetry directions is shown in fig.[fig:9 ] .
the @xmath9 band structure for the bilayer was mapped onto a tight binding model with maximally localized wannier functions for the radial part .
the ensuing band structure for the bilayers , with interlayer interactions switched off , has been superposed on the monolayer band structure in fig.[fig:9 ] .
the two are found to be almost identical for each of the systems shown here , indicating that the differences in the electronic structure between monolayer and bilayer arise due to interlayer interaction alone .
in conclusion , we have examined the evolution in the electronic structure of transition metal dichalcogenides as a function of layers .
the changes in the structure that one finds have been discussed in terms of a combination of interlayer interactions and geometric confinement effects .
considering the examples of mos@xmath0 and mose@xmath0 , a mapping onto a tight binding model with mo @xmath6 and se / s @xmath7 states in the basis allows us to quantify the role of each of these effects . even in these layered materials which are referred to as van der waal heterostructures
, we find that interlayer interactions play the primary role in bringing about changes in the electronic structure as a function of thickness .
geometric confinement effects on the other hand , we find , play no role in the observed changes in electronic structure . band structure ( solid line ) for monolayer mose@xmath0 compared with the tight binding band structure ( red circles ) of bilayer with interlayer interactions switched off .
the zero of energy is the valence band maximum.,width=283,height=207 ] ( solid line ) and tight binding ( red circles ) band structure for trilayer mose@xmath0 .
the tight binding hamiltonian for the trilayer has been constructed from monolayer for the layers and interlayer interactions taken from the bilayer .
zero of energy is the valence band maximum.,width=283,height=207 ] band structure ( solid line ) for monolayer mos@xmath0 compared with the tight binding band structure ( red circles ) of bilayer with interlayer interactions switched off .
the zero of energy is the valence band maximum . , width=283,height=207 ] band structure ( solid line ) for monolayer compared with the tight binding band structure ( red circles ) of bilayer with interlayer interactions switched off , for ( a ) mote@xmath0 , ( b ) ws@xmath0 , ( c ) wse@xmath0 and ( d ) wte@xmath0 .
the zero of energy is the valence band maximum.,width=529,height=377 ] .onsite energies obtained from tight binding mapping of the @xmath9 band structure .
a basis consisting of mo _ d _ and se _ p _ states has been considered for monolayer and bilayer mose@xmath0 . the respective se @xmath16 is taken as reference for the energies given in ev . [ cols="^,^,^ " , ] | we have examined the origin of the electronic structure changes as a function of the number of layers in transition metal dichalcogenides ( mx@xmath0 , where m = mo , w and x = s , se and te ) .
the belief has been that both geometric confinement effects as well as interlayer interactions play an important role in determining the electronic structure of transition metal dichalcogenide layers as the thickness is varied . by carrying out a mapping onto a tight binding model , we show that the evolution in the electronic structure with thickness is determined by interlayer interactions with geometric confinement effects playing almost no role . |
soft x - ray transients ( sxts ) are compact binary systems in which a low - mass secondary ( either a main - sequence star or a subgiant ) transfers mass via roche - lobe overflow onto a black hole ( bh ) or a neutron star ( ns ) primary ( see reviews by tanaka & lewin 1995 ; van paradijs & mcclintock 1995 ; white , nagase & parmar 1995 ) .
sxts have highly variable luminosities .
they spend most of their lifetime in a low luminosity quiescent state , but occasionally undergo dramatic outbursts during which both their optical and x - ray emission increase by several orders of magnitude ( see , e.g. , tanaka & shibazaki 1996 ; chen , shrader & livio 1997 ) . the x - ray emission during outburst of an sxt
is typically dominated by relatively soft , thermal emission from the accretion disk surrounding the compact object , while the optical emission is usually interpreted as reprocessed x - rays from the disk and/or the companion star .
the new source xte j1118 + 480 belongs to the class of sxts , but it also possesses some rather unusual characteristics .
xte j1118 + 480 was discovered on march 29 2000 with the rxte all - sky monitor as a brightening x - ray source .
pointed rxte observations confirmed the presence of this high galactic latitude source ( @xmath10 , @xmath11 ) , with a rather hard power law emission spectrum of photon index @xmath12 up to 30 kev .
it was subject to rapid x - ray flares , but no pulsation was detected .
retrospective asm analysis revealed that the source experienced another modest outburst in january 2000 ( remillard et al .
2000 ; see fig . [
fig : lightcurve ] ) .
batse observations showed that the source is visible up to 120 kev ( wilson & mccollough 2000 ) , and a 6.2 mjy ( at 15 ghz ) variable radio counterpart was later discovered ( pooley & waldram 2000 ) . the @xmath13 optical counterpart of xte j1118 + 480 was found to correspond to an @xmath14 mag star in the usno catalog by uemura , kato & yamaoka ( 2000 ; see also uemura et al .
a photometric modulation on a 4.1 hr period was observed by cook et al .
( 2000 ) and was later confirmed as a plausible orbital period by patterson ( 2000 ) and uemura et al .
( 2000 ; see stull , ioannou & webb in haswell et al .
2000b for a discrepant claim at twice this value ) .
the photometric modulation was later reported changing shape and period , possibly showing the development of superhumps in this source ( uemura 2000 ) .
if these are normal superhumps , then the orbital period would be smaller than the photometric superhump modulation by a few percent .
garcia et al . ( 2000 ) reported the first optical spectroscopic results on xte j1118 + 480 .
they found an optical spectrum typical of sxts in outburst , with very broad h@xmath15 , h@xmath16 and heii lines ( fwhm @xmath17 km s@xmath18 ) .
these observations also indicated the presence of absorption features and a very low interstellar absorption ( @xmath19 ) to the source .
these authors suggested that the surprisingly low x - ray ( @xmath20 mcrab , 2 - 12 kev ) to optical ( @xmath21 ) flux ratio of xte j1118 + 480 could be due to a nearly edge - on viewing angle .
additional x - ray observations revealed the presence of a strong quasi - periodic oscillation ( qpo ) at 0.085 hz in the x - ray lightcurve of xte j1118 + 480 .
the shape of the power density spectrum and the hard emission spectrum of xte j1118 + 480 prompted revnivtsev , sunyaev & borozdin ( 2000 ) to propose that the source is a bh transient , by analogy with other such systems .
the qpo was confirmed by asca observations , which also suggest the presence of a soft component in the spectrum ( below 2 kev ) , possibly due to emission from the accretion disk in the system ( yamaoka et al .
soft x - ray observations by _ chandra _ did not confirm this ( mcclintock et al . 2000 ) .
the qpo frequency was reported to have shifted from 0.085 hz to @xmath22 hz in subsequent observations ( yamaoka et al .
2000 ; wood et al .
2000 ) . pointed xte observations at the end of may do not reveal significant changes in the emission spectrum ( p. jonker , private communication ) .
xte j1118 + 480 has also been the subject of an extensive multiwavelength observation campaign with hst , euve , ukirt and rxte .
thanks to the very low interstellar absorption to the source , the first euve spectrum of a bh - candidate x - ray transient was obtained ( mauche et al . 2000 ) .
no periodic modulation was found in the euve data ( hynes et al .
an hst spectrum revealed a very broad ( @xmath23 km s@xmath18 ) ly@xmath15 absorption feature , suggestive of a massive accretor ( haswell et al .
haswell et al . ( 2000c ) obtained a near - uv power density spectrum of xte j1118 + 480 with a qpo and an overall shape in agreement with previous rxte timing data .
in addition , the near - uv variability was found to lag by 1 - 2 seconds behind the x - ray variations , as would be expected from light echoes in a system with @xmath24 hr .
hynes et al .
( 2000b ) , combining data from hst , euve , ukirt and rxte , find that the ir to uv data suggests emission from both an optically - thick disk and another , flat spectrum component ( possibly synchrotron ) , while the euve and x - ray data suggest a power law emission typical of a galactic x - ray binary in a low - hard state .
they conclude that xte j1118 + 480 , rather than experiencing a full outburst approaching the eddington luminosity , seems to be in a mini - outburst state ( but see kuulkers 2000 ) . in this paper
, we describe the results of our optical spectroscopic campaign to monitor the evolution of xte j1118 + 480 during its recent outburst ( and early decline ) . in 2 , we describe our observations and the data reduction techniques used .
our results concerning the continuum and line emission , a partial s - wave pattern and the interstellar absorption to the source are presented in 3 and discussed in the context of the current knowledge on xte j1118 + 480 in 4 .
our main conclusions are summarized in 5 .
we obtained optical spectra of xte j1118 + 480 from april 7 , 2000 to july 4 , 2000 with the arc 3.5 m telescope at apache point observatory .
we mostly used the double imaging spectrograph ( dis ) , but spectra with the echelle spectrograph were also obtained on april 7 , 2000 .
table [ tab : obs ] summarizes the dates and other characteristics of our observational campaign .
most of our observations with dis were carried out with the high resolution gratings ( hereafter _ hires _ ) with a 1.5 slit ( dispersion 1.6 pixel@xmath18 in the blue , 1.1 pixel@xmath18 in the red , and a resolution of 2 pixels ) , but we also obtained spectra with the low - resolution gratings ( same slit size , dispersion 6.2 pixel@xmath18 in the blue , 7 pixel@xmath18 in the red , and a resolution of 2 pixels ; hereafter _ lowres _ ) on april 7 , 2000 .
for the dis high resolution observations , the blue and red gratings were centered on slightly different wavelengths during our various nights of observations , but we generally centered the blue side to cover the h@xmath16 and heii @xmath1 lines , and the red side to cover the h@xmath15 line .
the complete list of spectral coverage for our observations can be found in table [ tab : obs ] .
the echelle spectrograph covers the ( fixed ) wavelength range 3500 - 10000 with @xmath25 ( 10 km s@xmath18 at 5000 ) and a resolution element of @xmath26 2.5 pixels .
all the exposure times can be found in table [ tab : obs ] .
the dis observations were reduced in the standard way using _ iraf _ and the spectra were optimally - extracted and dispersion corrected without any particular difficulty .
the echelle spectra were reduced using the iraf _ ecspec _ package .
direct extraction of the object spectra proved difficult for the highest ( bluest ) orders where the trace was hard to follow .
we decided to use the flat field as a guide and validated the method with a standard star observed on the same night .
this resulted in significantly higher s / n in the blue part of the spectrum .
a total of 115 orders were extracted covering the spectral range 3500-10000 , in which the source is detected , albeit with varying sensitivity . the dispersion varied between @xmath26 0.05 / pixel to 0.1 / pixel and the usable spectral range in each order between @xmath26 70 to 150 from the blue to the red end .
the emission lines in the spectrum of xte j1118 + 480 are both broad and weak ( see below , fig .
[ fig : lowres ] ) and their identification in the echelle spectra turned out to be challenging .
the broad lines can cover almost half of the spectral range in one order making a continuum fit unreliable .
we therefore proceeded by normalizing to an interpolated continuum determined from the two closest orders where no lines are expected .
the normalised summed spectra clearly show h@xmath15 ( in one order , left panel of fig .
[ fig : echha ] ) and heii @xmath274686 ( spread over two orders , right panel of fig .
[ fig : echha ] ) .
these lines are also detectable in the individual continuum - subtracted echelle spectra . the different profile of the heii line in the _ lowres _ and echelle spectra is most probably due to the s - wave modulation discussed below .
this s - wave is much less prominent in h@xmath15 .
the rxte asm lightcurve of xte j1118 + 480 is shown in fig .
[ fig : lightcurve ] with the dates of our spectroscopic observations indicated by dashed lines .
the source was in outburst during the entire period covered by our optical observations , with a flux about 40 mcrab in the 2 - 12 kev band ( fig .
[ fig : lightcurve ] ) .
optical photometry also shows the source in outburst with a mean magnitude that decreased at most by @xmath28 mag ( from its peak value @xmath29 ) at the end of our observational program .
since then , the optical and x - ray flux have decreased significantly , indicating that the source may soon enter quiescence .
figure [ fig : lowres ] shows the sum of the two dis _ lowres _ spectra obtained on april 7 , 2000 , which is also representative of other spectra of the source obtained later on during our campaign . the strong , blue power - law continuum and the weak emission lines are typical of an sxt in outburst ( see also garcia et al
a power - law fit to the blue part of the spectrum yields a slope @xmath30 ( where the flux f@xmath31 ) , while a similar fit to the red part of the spectrum yields @xmath32 .
these values are consistent with the expectation @xmath33 for an optically - thick accretion disk ( see , e.g. , frank , king & raine 1992 ) .
hynes et al .
( 2000b ) find that the spectrum is flat using a wider wavelength range .
the broad , double - peaked h@xmath15 emission line is the only one easily identified on the spectrum of fig .
[ fig : lowres ] , but a closer examination reveals the presence of heii @xmath34 , and of additional balmer and he lines in the other spectra that we collected ( see 3.2 below ) .
we find only marginal evidence for continuum variability in xte j1118 + 480 , both during the same night and from night to night .
the lack of obvious variability is consistent with the long - term variations in the optical vsnet lightcurve ( @xmath35 0.5 mag ) , the short timescale ( @xmath3510 s ) of the large amplitude ( @xmath360.3 mag ) flickering and the low amplitude ( 0.052 mag ) of the 4 hr modulation ( patterson 2000 ) .
we identify three balmer lines , three he lines and the bowen blend in our set of spectra .
these are : h@xmath37 , hei @xmath38 , the bowen blend at @xmath39 , heii @xmath40 , h@xmath41 , hei @xmath42 , h@xmath43 and hei @xmath44 the lines are weak , broad and double - peaked .
h@xmath16 , h@xmath45 and also the weaker hei @xmath274471.5 have double - peaked emission clearly embedded in a large absorption trough .
there is no clear evidence of such absorption around h@xmath15 .
[ fig : echha ] shows the h@xmath15 and heii line profiles in the summed echelle spectrum with the low resolution spectrum taken on the same night overplotted .
we find no evidence for structures at high spectral resolution in the lines , except for an excess at rest wavelength which would not be expected in a pure double - peaked disk profile .
this component also seems to be present in the high resolution dis spectra ( see fig . [
fig : sumlines ] ) .
the bowen blend , which is very weak in the dis _ hires _ spectra , was not detected on april 7 in either the dis _ lowres _ or the echelle spectra .
measuring the lines proved difficult , not because of a lack of counts in the spectra but because of their intrinsic broadness and weakness . as a rule , we performed two - component gaussian fits to the emission lines . for h@xmath45 and
h@xmath16 the two emission components were subtracted to measure the equivalent width ( ew ) and fwhm of the absorption trough .
the ew of the lines varies ( with large associated errors ) within a night but we could find no periodic behaviour linked to the photometric period .
we also tried to estimate the semi - amplitude velocity of the primary @xmath46 by fitting the wings of the heii line ( expected to trace the motion of the compact star ) and folding around the suspected @xmath47 of 0.17 day .
this did not lead to any result as could be expected from the usually low @xmath46 of sxts ( @xmath48 km s@xmath18 ) .
the evolution of these lines during the 3 months of observations is shown in fig .
[ fig : sumlines ] , after the underlying continuum was normalized to unity .
when available , multiple spectra were summed up to increase the s / n of the spectra . a comparison between the most comprehensive datasets ( apr .
17 and jul .
4 ) , where measurement errors can be minimized , shows that the two strongest lines , h@xmath15 and heii , did not change significantly .
we conclude that there do not seem to be any long term variations .
the average properties of the lines are summed up in table .
[ tab : lines ] ( the quoted errors are the rms of the measurements ) . in all our spectra
the blue side of the absorption troughs in h@xmath45 and h@xmath16 is less strong than the red side .
the measured central wavelengths of the absorptions are redshifted compared to the rest wavelength ( which could be an artefact of the method we used to remove the double peaked emission ) .
fitting a gaussian to the red part of the absorption while ignoring ( but not subtracting ) the emission lines gave higher equivalent widths ( 2.4for h@xmath45 and 2.0 for h@xmath16 ) and better agreement with the rest wavelength ( 7 redshift for both instead of 15 ) .
similar absorption redshifts were reported by callanan et al .
( 1995 ) for gro j0422 + 32 .
a plausible explanation would be that the shifts are due to distorted line profiles .
such asymetric lines are observed in dwarf novae when the disk becomes eccentric and the system shows superhumps ( warner 1995 ) .
there is evidence for such superhumps in xte j1118 + 480 ( see discussion in 4.3 ) .
the heii emission may be used to obtain a crude estimate of the extreme ultraviolet ( euv ) flux from the source ( e.g. patterson & raymond 1985 ) if one assumes the @xmath274686 line stems from the recombination of hei photoionized in the disk by photons with energies between 55 - 280 ev . on the other hand ,
the doppler map discussed below in 3.5 shows most of the heii emission is localized and in all likelihood associated with stream - disk interaction . in this case the @xmath274686 line would be pumped by collisional excitation rather than photoionization .
the following estimate therefore only yields an upper limit on the euv flux .
this is still of interest since the upper limit is independent of the column density to the source .
we derive from our observations @xmath49 erg s@xmath18 @xmath50 and , from the values given by haswell et al .
( 2000b ) , @xmath51 erg s@xmath18 @xmath50 .
this favours case b recombination for which one would expect @xmath52 ( seaton 1978 ) .
therefore , the region is optically - thick to the ionizing flux but thin to the line and a fraction @xmath53 of the photoionized hei recombinations lead to @xmath274686 emission ( hummer & storey 1987 ) .
following previous applications to soft x - ray transients ( marsh , robinson & wood 1994 ; hynes et al .
1998 ) , the euv flux @xmath54 and the heii line flux @xmath55 are related through @xmath56 where @xmath15 is the fraction of euv photons intercepted by the disc while @xmath57 ev and @xmath58 ev refer to the mean energy of the photons .
we have assumed @xmath59 , which may be incorrect if the euv emission is not isotropic or if we see only a fraction of the heii emission . in both cases
this assumption leads to an underestimate of the true euv luminosity .
observations suggest that a fraction @xmath60 of the soft x - ray flux is intercepted and reprocessed in the optical by the accretion disk in sxts ( e.g. dubus et al .
1999 ) . by analogy , we take @xmath61 from which we finally derive an upper limit to the 55 - 280 ev flux of @xmath62 erg s@xmath18 @xmath50 .
this is obviously a very crude estimate .
an euve spectrum was obtained for this source , the first time ever for a sxt .
the source flux at these wavelengths is heavily dependent on the extinction , but the upper limit on @xmath54 derived above appears consistent with both a simple extrapolation of the optical uv flux observed and the euve flux inferred if a value of @xmath63 @xmath50 is assumed when dereddening the euve spectrum ( mauche et al . 2000 ; hynes et al .
2000 ) .
the mean peak - to - peak separation for all the observations is 18(1240 km s@xmath18 ) for h@xmath45 , 23 ( 1470 km s@xmath18 ) for heii , 21 ( 1300 km
s@xmath18 ) for h@xmath16 and 28 ( 1280 km s@xmath18 ) for h@xmath15 .
the s - wave is much less prominent in the balmer lines and we assume this represents the projected keplerian velocity of the outer disk .
the peak - to - peak separations have been found in well - known systems to overestimate the outer disk radius by about 20% because of sub - keplerian motion or local broadening ( marsh 1998 ) .
this uncertainty is acceptable considering other assumptions made below . for a keplerian disk with an emissivity @xmath64
, @xmath65 can be related to the outer ( emitting ) disk radius @xmath66 ( smak 1981 ) : @xmath67 where @xmath68 is the mass of the primary and @xmath69 the system inclination .
we can also express @xmath70 as a function of the velocity semi - amplitude of the secondary @xmath71 and the mass ratio @xmath72 by combining the above equation with the mass function ( see below , eq .
[ eq : fm ] ) and kepler s third law0.17 day , kepler s third law gives a binary separation @xmath73 cm with a weak dependence on @xmath68 and @xmath74 . ] , resulting in : @xmath75 the disk outer radius is expected to reach the tidal truncation radius for @xmath76 and to be limited to @xmath77 by the 3:1 resonance for @xmath78 ( e.g. warner 1995 ) .
papaloizou & pringle ( 1977 ; their tab .
1 ) give a theoretical estimate of the outer disk radius that we recklessly use for @xmath70 .
eq . [ eq : rova ] can therefore be used to give the semi - amplitudes @xmath79 for which the disk radius is consistent with the theoretical expectations .
for @xmath80 the disk radius varies between 0.5 and 0.3 in units of the binary separation @xmath81 .
the expected @xmath71 as a function of @xmath74 is shown as a solid line in the left panel of fig .
[ fig : cons1 ] .
we find @xmath82 ( km s@xmath18 ) @xmath83 for @xmath80 .
the observed velocity fwhm ( @xmath84 ) of the lines provides an upper limit to the minimum emitting keplerian disk radius : @xmath85 where @xmath84 is expressed in units of 1000 km s@xmath18 .
the absorption features from the optically - thick disk ( see 4.1 and table . [
tab : lines ] ) have fwhm @xmath86 km s@xmath18 implying @xmath87 cm .
haswell et al . ( 2000b ) detect broad ly@xmath15 absorption with @xmath88 km s@xmath18 which gives a stricter @xmath89 cm or @xmath90 in schwarzschild units .
spectral models of the x - ray low / hard state of bh candidates tend to predict larger values of @xmath91 in schwarzschild units ) at which the transition from a thin disk to a hot advection - dominated flow occurs ( see , e.g. , esin et al .
the euv flux of xte j1118 + 480 could also imply a smaller transition radius ( see 4.2 ) .
figure [ fig : swave ] shows the evolution of the heii @xmath92 line profile over an approximately continuous 2.9 hr period on april 20 , 2000 .
a partial s - wave pattern moves from the red to the blue side of the heii @xmath1 line rest wavelength and appears consistent with the claimed photometric period of @xmath93 hr ( cook et al .
2000 ; patterson 2000 ) .
the modulated component is particularly strong towards the middle of the observation at hjd 51654.876 where it is blueshifted by about 900 km s@xmath18 .
the trailed spectrum clearly shows that the bowen blend , hei @xmath94 , h@xmath45 and h@xmath16 follow the same s - wave pattern ( see fig .
[ fig : swave ] ) .
we note that identical behaviour is seen in the smaller continuous set ( 1.8 hr ) of july 4 although with poorer s / n ( not shown here ) .
however , the presence of the s - wave in the h@xmath15 emission profile is not clear on either dates . despite the incomplete phase coverage and the absence of a reference for superior conjunction of the secondary
, we attempted to locate the emission site of the s - wave in the binary velocity plane using the doppler tomography technique ( marsh & horne 1988 ) .
maps were reconstructed for heii and h@xmath15 for the april 20 observations using the software developed by spruit ( 1998 ) , assuming an orbital period of 0.1708 day and a null systemic velocity .
it was not possible to combine this dataset with our other observations as this would have required @xmath47 to be known much more accurately ( @xmath95 day with @xmath96 days and an error on the phase set by the exposure time of @xmath97 ) .
furthermore , the 0.1708 day modulation is probably related to superhumps so that the orbital period can be expected to be close to this value with an uncertainty of a few percent .
we tried several different orbital periods around 0.17 day but this had no significant impact on the maps .
the reconstructed trailed profiles and corresponding velocity maps are shown in fig .
[ fig : prof ] and fig .
[ fig : map ] respectively .
each pixel on the projected doppler map corresponds to an s - wave in the trailed spectra .
the binary components are located on the y - axis of the map at positions corresponding to their semi - amplitude velocities @xmath98 for the primary and @xmath99 for the companion ( indicated by crosses ) .
however , since the orbital reference phase is arbitrary ( here we chose @xmath10051654.420 in mjd ) , the map can be rotated in any fashion around the velocity origin .
the secondary position , the ballistic gas stream trajectory and the corresponding keplerian velocity at the location of the stream shown on fig .
[ fig : map ] result from particular choices of @xmath71 and @xmath101 discussed further below .
the trailed emission of heii shows a nice s - wave pattern which is consistent with the reported 0.1708 day photometric period ( cook et al .
2000 ; patterson 2000 ) .
most of the heii emission originates from a localized region but there is also an underlying ( double - peaked ) disk component which is visible in the original data .
on the other hand , the disk dominates the h@xmath15 profile producing a ring of emission in the velocity map .
we also note the regions of enhanced h@xmath15 emission at about ( -910,0 ) km s@xmath18 ( consistent with the heii s - wave ) and @xmath102 km s@xmath18 .
the peak at the velocity origin ( the center of mass ) is due to the h@xmath15 emission component at rest wavelength discussed in 3.2 . the bright emission region is typically associated with emission arising from the stream - disk interaction region or from the x - ray heated surface of the secondary ( e.g. smak 1985 ; marsh , robinson & wood 1994 ; casares et al . 1995ab ; harlaftis et al .
1996 , 1997ab ; hynes et al .
some sxts also have low - velocity emission regions in their doppler maps , possibly associated with a magnetic propeller ( hynes et al .
2000a ) . here ,
the velocity of @xmath103 km s@xmath18 makes it doubtful that we observe such a phenomenon . assuming the heii emission arises from the heated hemisphere of the secondary , the semi - amplitude @xmath104 km s@xmath18 ( underestimating the real @xmath71 ) gives a rough constraint on the mass function through : @xmath105 this gives a _ minimum _ mass for the primary of 12 m@xmath106 , which would make xte j1118 + 480 at least as massive as the black hole in v404 cyg .
lower inclinations would imply an even greater mass making this system rather unusual compared to the dynamical masses inferred in other sxts ( see , e.g. , van paradijs & mcclintock 1995 ) .
although we can not formally reject this possibility , the heii emission is unlikely to originate from the heated surface of the secondary .
the heii emission is more likely to originate from the stream - disk interaction .
typically , the hotspot is located ( see above references ) somewhere between the ballistic gas stream trajectory and its corresponding keplerian velocity ( which defines a circularization radius ) .
we derive a lower limit on @xmath79 ( hence a lower limit on the primary mass ) by assuming that the emission is located at the intersection of the two trajectories .
the values of @xmath71 and @xmath74 for which the intersection matches the heii emission are plotted as a dashed line in the left panel of fig .
[ fig : cons1 ] .
a comparison between the two estimates of @xmath79 shows that values of @xmath107 best fit the data in the sense that they minimize the discrepancies .
formal agreement requires @xmath108 ( @xmath109 km s@xmath18 ) but this is , of course , model - dependent . knowing @xmath74 and @xmath71 , the mass function ( eq . [ eq : fm ] ) gives the inclination @xmath69 as a function of @xmath68 .
this relation is plotted in fig .
[ fig : cons1 ] ( right panel ) for the two extreme cases ( @xmath110,@xmath111 km s@xmath18 ) and ( @xmath112,@xmath113 km s@xmath18 ) .
clearly , a ns would require large inclinations while a bh gives intermediate inclinations .
given the assumptions made for the interpretation of the peak - to - peak velocity and the heii emission , we are likely to underestimate @xmath71 in both cases .
this would only strengthen the case for a black hole primary .
the stream - disk interaction apparently extends some way along the keplerian velocity trace as previously seen in other sxts ( e.g. harlaftis et al . 1996 ) .
this could also be a numerical artifact possibly due to an inaccurate orbital period .
the second weaker emission region in the h@xmath15 velocity map could be due to continued disturbance of the accretion disk further along in azimuth . in this picture
the stream - disk interaction takes place inside the disk as emphasized by the larger velocity amplitude of the s - wave with respect to the peak - to - peak amplitude of the balmer emission lines .
this interpretation suggests significant stream overflow above the outer disk and an interaction region close to the circularization radius . alternatively
, systematic effects leading to lower observed peak - to - peak velocities ( e.g. local line broadening , marsh 1998 ) would place the hot spot closer to the disk outer edge .
the echelle spectra of xte j1118 + 480 obtained on april 7 , 2000 reveal the presence of weak caii @xmath114 absorption features , which can be used to estimate the neutral hydrogen absorption column to the source .
the spectra also show caii @xmath115 in absorption , with an equivalent width of order half that of caii @xmath114 features , as expected for this doublet ( see , e.g. , cohen 1975 ) .
both features are seen on two overlapping orders with similar structure .
the caii lines are detected in some of the individual echelle exposures with identical profiles .
figure [ fig : ca2 ] shows the caii @xmath116 absorption features in the summed echelle spectra .
the two orders have also been summed for this plot with weighting corresponding to the different sensitivities at the locations of the caii lines . a gaussian fit to each of the three absorption components yields heliocentric velocities @xmath117 , @xmath118 and @xmath119 km s@xmath18 .
these velocities are consistent with an absorption component clearly seen at @xmath120 km s@xmath18 and the hint of another component at @xmath118 km s@xmath18 in nai @xmath121 .
the presence of strong na emission at the rest wavelength ( sky emission ) precludes the detection of the presumed third absorption component at @xmath119 km s@xmath18 and the use of na absorption for the determination of the hi absorption column , n@xmath122 , to xte j1118 + 480 .
we proceed as follows to determine n@xmath122 .
we assume that the gas in each absorbing cloud has a maxwellian velocity distribution centered on its heliocentric velocity @xmath123 , so that the caii optical depth of the cloud can be described by ( spitzer 1978 ) : @xmath124 = { \rm n_{ca{\sc ii } } } \frac{\sqrt{\pi } e^2 f \lambda_0}{m_e c b } e^{-[(v - v_0)/b]^2},\ ] ] where @xmath125 is the total cloud absorption column , @xmath126 is the rest wavelength of the caii k line , @xmath127 is the oscillator strength of this transition ( cardelli & wallerstein 1986 ) , @xmath128 is the velocity dispersion of the cloud , @xmath129 is the charge of the electron , @xmath130 is the mass of the electron and @xmath131 is the speed of light .
relative to a unity continuum , the line profile of a given absorbing cloud component has a depth given by : @xmath132}\ ] ] at any speed @xmath133 ( or equivalently wavelength ) around the central value @xmath123 . to deduce precise values of @xmath128 and @xmath125 ,
we have compared the observed ca@xmath134 line profiles with theoretical , multi - component absorption line profiles .
initial profiles were calculated using the values of @xmath128 and @xmath125 derived from eqs .
( 6 ) & ( 7 ) , which were then convolved with a gaussian instrumental line spread function whose fhwm was measured from the arc lines used to wavelength calibrate the data
. the best values of @xmath128 and @xmath125 were found by minimizing @xmath135 between the data and the theoretical fits .
the resulting values for the ca@xmath134 lines are : @xmath136 and 6.2 km s@xmath18 , @xmath137 and 11.67 for the components at @xmath138 and @xmath139 km s@xmath18 respectively .
the translation of the caii absorption column to an hydrogen absorption column is somewhat delicate , especially in view of the possibly efficient deposition of gaseous caii onto dust grains ( see , e.g. , vallerga et al .
1993 ) . however , there is a substantial body of observational data on caii absorption which allows us to use an average value for the ratio @xmath140 . in their independent studies ,
both cohen ( 1975 ) and sembach & danks ( 1994 ) find an average average value of @xmath141 \simeq -8.3 $ ] for high latitude line - of - sights , with approximate errorbars of @xmath142 in log in the two cases . assuming this average is representative of the line of sight to xte j1118 + 480
, this translates in a value @xmath143 = 20.45 \pm 0.2 $ ] , where the errorbars come from the conversion of @xmath144 to @xmath145 absorption .
we note that this estimate is somewhat larger than the values @xmath146 \approx 20 $ ] considered as reasonable by hynes et al .
( 2000 ; see also mcclintock et al .
2000 ; esin et al .
2000 ) when dereddening the low - energy spectrum of xte j1118 + 480 .
this discrepancy raises the possibility that the value obtained from the ca@xmath134 measurement overestimates the true value of n@xmath147 on the line of sight to the source .
most likely , this would come from the conversion from ca@xmath134 to h@xmath148 which may not be `` average '' as assumed above for xte j1118 + 480 .
it is also possible to measure the total @xmath149(h i ) along the sightline towards xte j1118 + 480 from 21 cm emission observations . using data from the leiden / dwingeloo survey ( hartmann & burton 1997 ) , the h i column at ( @xmath150)=(157.5,62.0 ) is @xmath151 @xmath50 , a factor of two lower than we derive .
this value represents an _ upper _ limit of course , since the 21 cm emission is integrated along the line of sight through the entire milky way .
this therefore also suggests our value of @xmath149(h i ) is over - estimated , although the dwingeloo beam size of 36@xmath152 is too large to measure higher @xmath149(h i ) over smaller angular scales , and it remains possible that a small , dense knot of h i with @xmath149(h i)@xmath153 @xmath50 could be lying directly along the line of sight . it is encouraging to see that the properties derived above for the clouds are consistent with the average properties of high latitude line - of - sights described in sembach & danks ( 1994 ) : @xmath154 @xmath50 .
furthermore , these authors find that , along extended sight lines at high latitude , the average number of absorbing caii clouds is 3.6 kpc@xmath18 .
this indicates that xte j1118 + 480 is located at a distance @xmath155 kpc if the line of sight to the source has average properties , an estimate which is consistent with those for other sxts . the actual distance
to xte j1118 + 480 could be somewhat different if this line - of - sight has properties deviating significantly from the mean . note that our estimate of the distance is in rough agreement with that of uemura et al .
( 2000 ) who use the 18.8 quiescent magnitude to derive @xmath156 kpc ( m - type secondary ) or @xmath157 kpc ( k secondary ) , and is consistent with the 0.8 kpc distance estimate of mcclintock et al .
( 2000 ) , assuming that the primary is a massive bh and that a third of the light in quiescence is provided by the m - dwarf secondary .
it is also of interest to note that the 21 cm emission from the leiden / dwingeloo survey mentioned above shows two peaks in brightness temperature which correspond closely to the ca ii components at @xmath119 and @xmath139 km s@xmath18 .
although the distance of the h i is not known , the scale height of h i in the milky way is @xmath158 kpc , and since the 21 cm emission arises from _ all _ the h i in the galaxy along the sightline , xte j1118 + 480 must be sufficiently far away so as to show the ca ii absorption from both h i components .
this again suggests a significant distance between us and xte j1118 + 480 .
an interesting feature of the spectra are the absorption troughs in @xmath159 and @xmath160 .
although rarely discussed ( see , however , soria et al . 2000 ) , these have been observed in several other transients with similar fwhm , including gro j1655 - 40 , a0620 - 00 , gs1124 - 68 , nova mus 91 , xte j2123 - 058 , xte j1859 + 226 , and especially gro j0422 + 32 which bears many similarities to xte j1118 + 480 ( see conclusion ) .
this type of line profile is also seen in dwarf novae ( dn ) and nova - like ( nl ) , systems which are analogous to low - mass x - ray binaries ( lmxbs ) , except for a primary which is a white dwarf . typically , the spectrum of quiescent dn display strong emission lines which are gradually replaced by absorption features during the rise to outburst and vice - versa during the decay ( e.g. szkody et al .
1990 ) .
schematically , the absorption is thought to arise from the optically - thick accretion disk whilst the emission is thought to result from photoionization in a chromosphere - like optically thin region above the disk .
a key feature is that the expected spectrum is inclination dependent since the absorption should disappear as the system gets closer to being edge - on due to strong line limb - darkening in the 2d disk ( see e.g. marsh & horne 1990 ; la dous 1989 ; wade & hubeny 1998 ) .
there is indeed such an observed dependence in both dn and nl with the highest inclination systems showing only emission lines ( la dous 1991 ) .
although the detected absorption makes it unlikely that the inclination of xte j1118 + 4800 is very high , this constraint is weak . for instance , gro j1655 - 40 also showed balmer absorption ( soria et al .
2000 ) but has a well - determined inclination of 70 .
in contrast with dn and nl , irradiation heating is much stronger in sxt where it usually dominates the heat balance in the outer disk regions ( hence the optical emission , van paradijs & mcclintock 1994 ) .
energy deposited by soft x - rays easily causes a thermal inversion in the top layers where the emission lines originate ( e.g. ko & kallman 1994 ) .
if the optically thin atmospheres of sxts are indeed more extended than in dn , one would expect balmer absorption not to be as common an occurence as in the non - magnetic cataclysmic variables .
the spectral models of x - ray irradiated disks computed by sakhibullin et al .
( 1998 ) show that softer irradiation leads to the disappearance of the absorption troughs . as the irradiation spectrum hardens , the x - ray photons deposit their energy in deeper disk layers , which do not contribute to the line emission .
inversely , any parameter change resulting in additional heating of the disk atmosphere leads to more emission .
any combination of a low inclination , low x - ray luminosity , hard x - ray spectrum or low fraction of reprocessed x - ray photons in the disk could explain the absorption troughs .
those sxts in which balmer absorption has been observed have probably been guilty of one or several of the above .
interestingly , in gro j0422 + 32 , the h@xmath15 and h@xmath16 lines evolved from absorption to emission within 3 days during the rise to one of the mini - outbursts ( callanan et al .
although the authors could not find any evidence for increased x - ray flux between 0.5 - 10 kev , the simultaneous rise in heii emission does suggest a larger flux of soft 0.05 - 0.3 kev photons which would have a strong effect on the chromosphere .
however , a hard x - ray upturn in gro j1655 - 40 was accompanied by a dramatic increase of h@xmath15 emission but also by a _ decrease _ of the heii @xmath1 flux ( shrader et al .
it seems difficult to identify the dominant culprit but we propose the low x - ray flux ( see next section ) and the hard x - ray spectrum of xte j1118 + 480 as prime suspects responsible for the absorption troughs observed .
a unique characteristic of this system is the large optical to x - ray ratio , as compared to other sxts .
the ratio for xte j1118 + 480 is not dissimilar ( garcia et al .
2000 ) to that found in accretion disk corona ( adc ) sources where the x - ray flux is partly hidden by the disk seen edge - on . yet
, xte j1118 + 4800 has not shown any of the typical behaviour associated with adc sources , namely x - ray dips or eclipses .
in addition , our crude ` fitting ' of the peak - to - peak velocity of the balmer lines and of the velocity map rather hints at intermediate inclinations . this high ratio might simply be due to the low absorption and proximity of the source . at the @xmath26 0.8 kpc distance implied by the caii absorption , the 2 - 10 kev x - ray flux ( @xmath161erg s@xmath18 @xmath50 , yamaoka et al .
2000 ) is such that @xmath162 erg s@xmath18 . by analogy with the low / hard state of cyg x-1 , fender et al .
( 2000 ) find the broad band x - ray luminosity is @xmath163erg s@xmath18 @xmath50 at 1 kpc , i.e. this outburst was weak for a sxt .
we compare the x - ray flux and optical magnitude @xmath164 by computing the quantity @xmath165 , where @xmath166 erg s@xmath18 is the eddington luminosity . with @xmath167 day , we find @xmath168 with @xmath169 m@xmath106 while @xmath170 for @xmath171 and @xmath172 . within the error bars ,
this is consistent with the correlation found for other sxts between @xmath173 and @xmath164 by van paradijs & mcclintock ( 1994 ) . in other words , the optical flux of xte j1118 + 480
is roughly in agreement with what would be expected if the disk is irradiation - dominated .
the agreement is less good for higher @xmath68 , the optical flux being larger than expected from the correlation .
the x - ray heated companion could contribute to the optical as in , e.g. , the persistent lmxb cyg x-2 .
in this case we might expect the heii emission to originate from the x - ray heated surface of the secondary .
we have argued this is unlikely .
a better candidate for this extra optical emission is synchrotron emission ( hynes et al .
2000b ) originating in a magnetic corona ( merloni et al .
2000 ) , an advection - dominated accretion flow ( adaf ; esin et al .
2000 ) and/or a jet ( @xmath26 equivalent to the first model with the base of the jet acting as the magnetic corona ; markoff et al .
xte j1118 + 480 probably has a powerful compact jet , with emission extending at least to the near - infrared ( fender et al .
both adaf and magnetic corona models can explain the high optical / x - ray ratio but the euv predictions may differ .
the model presented by merloni et al .
( 2000 ) has a large blackbody component peaking at @xmath174 hz due to hard x - ray reprocessing in the disk .
the soft x - ray observations of _ chandra _ do not seem to support this model ( mcclintock et al .
2000 ; esin et al .
2000 ) .
the detection of superhumps ( uemura 2000 ) could add several constraints to the system parameters . according to current models ,
superhumps appear when the disk can reach the 3:1 resonance radius , requiring small mass ratio @xmath175 ( warner 1995 ) . in principle
, the relative difference between the superhump period and the orbital period can give an even better constraint on @xmath74 ( e.g. odonoghue & charles 1996 ) .
such a small mass ratio is compatible with the constraint @xmath107 that we derived in 3.5 .
in dn and nl , the origin of the optical variation could lie in the enhanced viscous dissipation modulated on the superhump period .
however , these variations should be swamped in sxts where irradiation heating rather than viscous heating dominates the optical output .
haswell et al . ( 2000a ) recently argued that the small change in the accretion disk area on the superhump cycle , rather than the enhanced viscous dissipation , would give the desired effect for sxts .
the amplitude of the superhump would however decrease with larger inclinations .
the detection of superhumps in xte j1118 + 4800 , if confirmed , would therefore either mean the disk was not irradiation - dominated ( which is unlikely considering , e.g. , the correlated x - ray / near - uv behaviour reported by haswell et al .
2000c ) or that the inclination is low enough for the variations of the area of the disk to be observable .
the high galactic latitude of xte j1118 + 480 ( @xmath176 ) is in sharp contrast with the latitudes of other lmxbs which cluster next to the galactic plane .
whether or not xte j1118 + 480 is a halo object obviously depends on its actual distance . at our inferred @xmath177 kpc ,
xte j1118 + 480 lies @xmath178 kpc above the galactic plane .
the distribution of bh lmxbs around the galactic plane has a dispersion of @xmath179 kpc .
the larger dispersion of ns lmxbs ( @xmath180 kpc ) is probably due to larger kick velocities at birth ( white & van paradijs 1996 ) .
whether xte j1118 + 480 has a bh or a ns primary , its height above the galactic plane is not much larger than for other lmxbs .
for instance , white & van paradijs ( 1996 ) find @xmath181 kpc for the bh nova oph 1977 . a low metallicity environment would be supported by the low bowen to heii flux ratio @xmath182 ( table .
[ tab : lines ] ) that we find ( motch & pakull 1989 ) .
but the detection of n and the non - detection of c and o lines in the uv spectra has been used to argue that the matter has been cno processed in the companion ( haswell et al .
2000b ) and that , from an evolutionary point of view , @xmath183 m@xmath106 is preferred to allow for an evolved secondary .
considering that xte j1118 + 480 is not extremely far from the distribution of other lmxbs above the galactic plane and that cno processing might explain the low bowen / heii flux ratio , we conclude that there is no compelling evidence for a halo origin of xte j1118 + 480 .
let us first summarize the main conclusions of this work .
@xmath184 the optical spectrum of xte j1118 + 480 in outburst shows several variable broad and weak emission lines superposed on a @xmath185 continuum typical of an optically - thick accretion disk . the heii line shows a strong s - wave pattern which is consistent with the claimed 0.1708 day photometric period .
we see no other obvious periodic or long - term behaviour in the lines .
@xmath184 we find from the caii lines that the interstellar absorption to the source is low ( @xmath146 = 20.45 \pm 0.2 $ ] ) and that the presence of three intervening clouds suggests a distance of @xmath186 kpc to the source . since the low bowen to heii ratio could indicate cno - processed material ( haswell et al .
2000b ) rather than an intrinsically low metallicity , we conclude there is no strong evidence to support a halo origin for xte j1118 + 480 .
@xmath184 we estimate an upper limit for the 0.05 - 0.3 kev source flux of @xmath187 erg s@xmath18 @xmath50 . the dereddened euve spectrum ( using a column absorption @xmath143 \approx 20 $ ] ; hynes et al .
2000 ; mcclintock et al .
2000 ) appears compatible with this value .
this suggests that a significant fraction of the source flux is emitted in the euv / soft x - ray band .
given the very strong dependence of the slope of the dereddened euve spectrum on @xmath188 , it may be difficult to accomodate a higher value for @xmath188 as derived from caii absorption in 3.6 . @xmath184 the absorption in h@xmath45 , h@xmath16 and the balmer jump ( hynes et al .
2000b ) , originating from the optically - thick disk , suggests that x - ray heating of the atmosphere is not as strong as in other sxts , which may be consistent with the ( very ) low x - ray flux and hard x - ray spectrum .
this also disfavors very high inclinations for which limb - darkening would remove the absorption features .
@xmath184 the balmer lines have peak - to - peak velocities of @xmath189 km s@xmath18 . assuming this corresponds to the keplerian velocity of the outer disk , we use theoretical arguments on the disk size to estimate @xmath79 .
we find @xmath190 ( km s@xmath18)@xmath191 for @xmath80 .
@xmath184 using the reconstructed doppler tomogram , we tentatively identify the heii s - wave with the stream - disk interaction .
this gives us a second estimate of @xmath79 which , combined with the first one , restricts possible values for @xmath74 to @xmath192 .
this further implies that if the primary is a ns , the system should have a high inclination ( @xmath193 ) , while for a bh primary , an intermediate inclination is preferred ( 30@xmath19450 for @xmath195 ) .
superhumps ( uemura 2000 ) , if they require @xmath196 and low to intermediate inclinations ( haswell et al .
2000a ) , the absence of dips or eclipses in the x - ray lightcurve , the lack of high frequency qpos ( revnivtsev et al . 2000 ) and the broad ly@xmath15 line ( haswell et al .
2000b ) all favor an intermediate inclination system containing a bh .
all of the arguments above are , however , rather uncertain and model - dependent .
the assumptions made to interpret the peak - to - peak velocity and heii emission probably underestimate @xmath71 .
a larger @xmath71 would strengthen the case for a black hole primary .
we conclude by emphasizing the similarities between xte j1118 + 480 and the black hole lmxb gro j0422 + 32 in mini - outburst : both have orbital periods of @xmath197 hours , stayed in the low / hard x - ray state during outburst , had low x - ray luminosities , showed superhumps and had identical optical spectra ( including the absorption troughs ) . scaling the @xmath198 secondary of gro j0422 + 32 from @xmath199 kpc to the distance of xte j1118 + 480 ( @xmath200 kpc )
gives @xmath201 , consistent with the reported 18.8 quiescent magnitude .
the system parameters for gro j0422 + 32 ( taken from chen et al .
1997 ) are also within our allowed range for a coherent explanation of both the balmer line peak - to - peak velocity and the location of the heii emission in xte j1118 + 480 .
quiescent studies , which can constrain the spectral type of the secondary and @xmath71 , will shed more light on this interesting object .
the authors are grateful to mike garcia and jeff mcclintock for very useful discussions , to todd tripp for sharing his knowledge of the apo echelle spectrograph and to rob hynes for a very useful referee report .
we are indebted to ed turner for granting us some of his director s discretionary time at apo .
support for this work was provided by nasa through chandra postdoctoral fellowship grant number pf9 - 10006 awarded by the chandra x - ray center , which is operated by the smithsonian astrophysical observatory for nasa under contract nas8 - 39073 .
gd acknowledges support from the european commission through the tmr network ` accretion on to black holes , compact stars and protostars ' ( contract number erbfmrx - ct98 - 0195 ) and from the leids kerkhoven - bosscha fonds .
rsjk acknowledges support from nsf grant ast96 - 16901 and the princeton university research board .
ps acknowledges support from nasa grant nag 5 - 7278 .
callanan , p. j. , garcia , m. r. , mcclintock , j. e. , zhao , p. , remillard , r. a. , bailyn , c. d. , orosz , j. a. ; harmon , b. a. , paciesas , w. s. , 1995 , apj , 441 , 786 casares , j. , charles , p. a. , marsh , t. r. , 1995a , mnras , 277 , l45 casares , j. , marsh , t. r. , charles , p. a. , et al .
, 1995b , mnras , 274 , 565 cardelli , j. a. & wallerstein , g. , 1986 , apj , 302 , 492 .
chaty , s. , haswell , c. a. , smith , g. p. , smail , i. , & hynes , r. i. , 2000 , iau circ .
7394 chen , w. , shrader , c.r .
& livio , m. , 1997 , apj , 491 , 312 .
cohen , j.g . , 1975 , apj , 197 , 117 . cook , l. , patterson , j. , buczynski , d. , & fried , r. , 2000 , iau circ . 7397 dhawan , v. , pooley , g. g. , ogley , r. n. , & mirabel , i. f. , 2000 , iau circ .
7395 la dous , c. , 1991 , a&a , 252 , 100 la dous , c. , 1989 , a&a , 211 , 131 dubus , g. , lasota , j .-
hameury , j .-
m . , charles , p. a. , 1999 , mnras , 303 , 139 esin , a.a .
et al . , 2000 , preprint esin , a.a . ,
mcclintock , j. e. , narayan , r. , 1997 , apj , 489 , 865 esin , a.a . ,
narayan , r. , cui , w. , grove , j.e . &
zhang , s .- n . , 1998 ,
apj , 505 , 854 frank , j. , king , a. r. , & raine , d. j. 1992 , accretion power in astrophysics , 2nd ed .
, cambridge univ .
press , cambridge garcia , m. , brown , w. , pahre , m. , mcclintock , j. , callanan , p. , & garnavich , p. , 2000 , iau circ . 7392 fender , r. p. , hjellming , r. m. , tilanus , r. j. , pooley , g. g. , deane , j. r. , ogley , r. n. , spencer , r. e. , 2000 , mnras , submitted harlaftis , e. t. , charles , p. a. , horne , k. , 1997a , mnras , 285 , 673 harlaftis , e. t. , steeghs , d. , horne , k. , filippenko , a. v. , 1997b , aj , 114 , 1170 harlaftis , e. t. , horne , k. , filippenko , a. v. , 1996 , pasp , 108 , 762 haswell , c. a. , king , a. r. , murray , j. r. , charles , p. a. , 2000a , mnras , submitted ( astro - ph/0008367 ) haswell , c. a. , hynes , r. i. , & king , a. r. , 2000b , iau circ .
7407 haswell , c. a. , skillman , d. , patterson , j. , hynes , r. i. , & cui , w. , 2000c , iau circ .
7427 hummer , d. g. , storey , p. g. , 1987 , mnras , 224 , 801 hartmann , d. & burton , w.b . , 1997 , _ atlas of galactic neutral hydrogen _ ( cambridge : cup ) hynes , r. i. , charles , p. a. , haswell , c. a. , casares , j. , zurita , c. , 2000a , in _ proceedings of astro - tomography workshop _ , brussels , july 2000 , eds
. h. boffin , d. steeghs , springer - verlag lecture notes in physics ( astro - ph/0008145 ) hynes , r. i. , mauche , c. w. , haswell , c. a. , shrader , c. r. , cui , w. & chaty , s. , 2000b , apjl , 539 , l37 hynes , r. i. , haswell , c. a. , shrader , c. r. , chen , w. , horne , k. , harlaftis , e. t. , obrien , k. , hellier , c. , fender , r. p. , 1998 ,
mnras , 300 , 64 ko , y .- k . , kallman , t. r. , 1994 , apj , 431 , 273 kuulkers , e. , 2000 , a&a , submitted lubow , s. h. , shu , f. h. , 1975 , apj , 198 , 383 mauche , c. , hynes , r. , charles , p. & haswell , c. , 2000 , iauc 7401 markoff , s. , falcke , h. , fender , r. p. , apjl , submitted ( astro - ph/0010560 ) marsh t. r. , 1998 , in _ wild stars in the old west : proceedings of the 13th north american workshop on cvs and related objects _ , ed .
s. howell , e. kuulkers , c. woodward , asp conf .
series , vol .
137 , p.236 marsh , t. r. , horne , k. , 1990 , apj , 349 , 593 marsh , t. r. , horne , k. , 1988 , mnras , 235 , 269 marsh , t. r. , robinson , e. l. , wood , j. h. , 1994 , mnras , 266 , 137 mcclintock , j.e .
et al . , 2000 , preprint merloni , a. , di matteo , t. & fabian , a.c . , 2000 , mnras , submitted ( astro - ph/0006139 ) motch , c. , pakull , m. w. , 1989 , a&a , 214 , l1 odonoghue , d. , charles , p. a. , 1996 , mnras , 282 , 191 papaloizou , j. c. b. , pringle , j. e. , 1977 , mnras , 181 , 441 van paradijs , j. , mcclintock , j. e. , 1995 , in x - ray binaries , eds .
lewin , j. van paradijs & e.p.j .
van den heuvel , ( cambridge : cambridge university press ) , p. 58 .
van paradijs , j. , mcclintock , j. e. , 1994 , a&a , 290 , 133 patterson , j. 2000 , iau circ . ,
7412 patterson , j. , raymond , j.c .
, 1985 , apj , 292 , 550 pooley , g. g. , & waldram , e. m. , 2000 , iau circ .
, 7390 remillard , r. , morgan , e. , smith , d. , & smith , e. , 2000 , iau circ .
7389 revnivtsev , m. , sunyaev , r. , & borozdin , k. , 2000 , a&a , submitted ( astro - ph/0005212 ) sakhibullin , n. a. , suleimanov , v. f. , shimanskii , v. v. , suleimanova , s. l. , 1998 , astl , 24 , 22 seaton , m. j. , 1978 , mnras , 185 , 5p shrader , c.r . , wagner , r.m . ,
hjellming , r.m . ,
starrfield , s.g . , 1996 , a&as , 120 , 261 sembachs , k.r .
& danks , a.c .
, 1994 , a&a , 289 , 539 .
smak , j. , 1985 , aca , 35 , 351 smak , j. , 1981 , aca , 31 , 395 soria , r. , wu , k. , hunstead , r. w. , 2000 , apj , in press spitzer , l. , 1978 , physical processes in the interstellar medium ( new york : john wiley & sons ) spruit , h.c . , 1998 , astro - ph/9806141 szkody , p. , pich , f. , feinswog , l. , 1990 , apjss , 73 , 441 tanaka , y. & lewin w.h.g . , 1995 , in x - ray binaries , eds
lewin , j. van paradijs & e.p.j .
van den heuvel , ( cambridge : cambridge university press ) , p. 126 .
tanaka , y. , shibazaki , n. , 1996 , ara&a , 34 , 607 uemura , m. , 2000 , iau circ .
7418 uemura , m. , kato , t. , & yamaoka , h. , 2000 , iau circ .
7390 uemura , m. et al . , 2000 , pasj , in press ( astro - ph/0004245 ) vallerga , j.v . ,
vedder , p.w .
, craig , n. ; & welsh , b.y .
, 1993 , apj , 411 , 729 wade , r. a. , hubeny , i. , 1998 , apj , 509 , 350 warner , b. , 1995 , cataclysmic variables , ( cambridge : cambridge university press ) white , n. e. , van paradijs , j. , 1996 , apj , 473 , l25 white , n. e. , nagase , f. & parmar , a. n. , 1995 , in x - ray binaries , eds .
lewin , j. van paradijs & e.p.j .
van den heuvel , ( cambridge : cambridge university press ) , p. 1 .
wilson , c. a. & mccollough , m. l. , 2000 , iau circ .
7390 wood , k. s. et al . , 2000 ,
apjl , submitted ( astro - ph/0006234 ) yamaoka , k. , ueda , y. , dotani , t. , durouchoux , p. & rodriguez , j. , 2000 , iau circ .
7427 ccccc + date ( ut ) & epoch & instrument & wavelength coverage & exposures + & ( hjd-2400000 ) & & ( ) & + + + april 7 , 2000 & 51641.7 & dis lowres & 4000 - 9000 & @xmath202s + & & echelle & 3500 - 9800 & @xmath203s + @xmath204s + april 20 , 2000 & 51654.9 & dis hires & 4200 - 5000 / 5800 - 6800 & @xmath205s + april 23 , 2000 & 51657.8 & dis hires & 4200 - 5000 / 6200 - 7200 & @xmath206s + may 6 , 2000 & 51670.8 & dis hires & 4200 - 5000 / 6300 - 7300 & @xmath207s + may 12 , 2000 & 51676.8 & dis hires & 4200 - 5000 / 6300 - 7300 & @xmath206s + may 15 , 2000 & 51679.7 & dis hires & 4200 - 5000 / 6300 - 7300 & @xmath206s + may 28 , 2000 & 51692.8 & dis hires & 4200 - 5000 / 6300 - 7300 & @xmath206s
+ july 3 , 2000 & 51728.7 & dis hires & 4200 - 5000 / 6300 - 7300 & @xmath206s + july 4 , 2000 & 51729.7 & dis hires & 4350 - 5150 / 6050 - 7050 & @xmath208s + @xmath209s + + lcccc + line & @xmath210 ( ) & @xmath211 ( ) & ew ( ) & fwhm ( 1000 km s@xmath18 ) + + + h@xmath45 & 4340.5 & -2@xmath2124 & [email protected] & [email protected] + & 4340.5 & 15@xmath2125 & [email protected] & [email protected] + hei & 4471.5 & 1@xmath2125 & [email protected] & [email protected] + & 4471.5 & 4@xmath2125 & [email protected] & [email protected] + bowen blend & 4638 & 8@xmath21210 & [email protected] & [email protected] + heii & 4686.7 & 0@xmath2123 & [email protected] & [email protected] + h@xmath16 & 4861.3 & -1@xmath2125 & [email protected] & [email protected] + & 4861.3 & 13@xmath2126 & [email protected] & [email protected] + hei & 4921.9 & 2@xmath2125 & [email protected] & [email protected] + h@xmath15 & 6562.8 & 0@xmath2122 & [email protected] & [email protected] + hei & 6678.1 & -1@xmath2125 & [email protected] & [email protected] + note .
negative equivalent widths indicate emission lines , while positive values correspond to absorption troughs .
@xmath210 is the rest wavelength and @xmath213 is the central wavelength at which each feature is observed . | we report on optical spectroscopic observations of the x - ray transient xte j1118 + 480 covering the period from april 7 , 2000 to july 4 , 2000 .
the spectrum is characterized by weak , broad , double - peaked balmer and he lines on top of a blue continuum of slope @xmath0 , as expected for an optically - thick accretion disk .
the weak bowen blend seen in our spectra may indicate a low metallicity for the source .
the presence of a partial s - wave pattern in the he @xmath1 line appears consistent with the reported photometric period @xmath2 hr for xte j1118 + 480 . by using a combination of doppler mapping and various theoretical arguments , we constrain plausible orbital parameters for the system : a mass ratio @xmath3 , an inclination @xmath4 for a neutron star primary , or @xmath5 for a black hole primary with a mass between @xmath6 and @xmath7 .
caii absorption features observed at very high resolution constrain the interstellar hydrogen absorption column @xmath8 \approx 20.45 \pm 0.2 $ ] and the identification of three absorbing clouds indicate a distance to the source @xmath9 kpc , assuming the line - of - sight to xte j1118 + 480 has average high - latitude properties .
these results are discussed in the context of previous multiwavelength observations of this unusual system . |
chemically peculiar ( cp ) variables are spread between late - b and early - f spectral types , and thus contain various stars with effective temperatures greater than 6,500 k @xcite .
these variables are comprised mostly of ap and bp stars , which differ from other types having the same temperature by their abnormal chemical compositions and slow rotations .
the reason for the peculiarity is an under - abundance of solar - like elements , as well as an overabundance of both metal and rare - earth elements across their surfaces @xcite .
magnetic fields , radiative acceleration , and atomic diffusion determine the surface distribution of elements @xcite , and lead them to be present in the form of spots and rings on the surface . along with rotation , these non - uniformly distributed regions cause periodic variations in magnetic fields , line profile , and energy distribution , as well as in photometric brightness ( oblique - rotator model ) .
the periods of these variations are generally between a day and a week .
depending on the slow rotation , surface spot regions can remain stable for decades .
such a situation enables remarkably accurate calculations of surface distribution , rotation period , and rotational breaking mechanisms .
however , only very few of the cp stars discovered in our galaxy and others exhibit periodic variations , and less than one - tenth of these have been observed for scientific investigation . in order to study this type of stars , accurate observations are needed ( accuracy @xmath5 0.005 mag @xcite ) .
the high - precision instruments of the _ stereo _ satellite are a quite suitable , space - based source , since seasonal and four - year _ stereo _ observations provide a precision of @xmath6 and @xmath7 mmag , respectively .
the photometric variability of v473tau ( a0si , v = 7.26 mag ) was first detected by burke et al .
@xcite . they calculated the period of this variation as around 1.39(2 ) days , but this period value produced a light curve ( lc ) with a scattered maximum .
hence , rakosch and fiedler @xcite noted that their observations were more adaptable with a double period .
subsequently , maitzen @xcite derived a rotation period of 2.7795(1 ) days , which was indeed twice that of previous values . as a result of the double period , two minima and maxima having different levels
were formed in the lc ; this situation was explained in terms of the different chemical regions on the surface .
most importantly , this was a significant case since a double wave structure was not a common condition among si stars .
in a recent study , jerzykiewicz @xcite investigated rotation periods and found a value of 1.4068541(29 ) days in u , b , and v bands .
however , he could not completely determine the origin of the variabilities as he was unable to conclude whether the star was an oblique rotator or a g - mode pulsator .
the data were provided from the hi-1a instrument on - board the _ stereo - a _ satellite .
the hi-1a is capable of observing background stars with the magnitude of @xmath8 or brighter for a maximum of 20 days and a useful stellar photometer which covers the region around the ecliptic ( 20% of the sky ) with the field of view of @xmath9 .
the nominal exposure time of the camera is 40 seconds , and putting 30 exposures together on board , a 40-minute integrated cadence has been obtained to transmit for each hi-1 image @xcite .
therefore , the nyquist frequency of the data is around 18 c d@xmath2 .
lcs mostly affected by solar activities were cleaned with a @xmath10 order polynomial fit .
observation points greater than @xmath11 were clipped with a pipeline written in the interactive data language ( idl ) ( for a more detailed description of the data preparation , refer to sangaralingam & stevens @xcite and whittaket et al .
the lc of v473tau presented a sinusoidal characteristic due to spot modulation on the stellar surface .
therefore , all analyses were performed using the lomb - scargle ( ls ) algorithm since it is more sensitive to such variations . to determine a model of the sinusoidal lcs
, the levenberg - marquardt optimization method was applied , and the best fit was obtained after 5000 iterations . after the derivation of the model lc , random gaussian noise with the mean of zero and the sigma value , which was determined from the cleaned curve , was produced and added to the model .
this process was repeated 500 times .
the most accurate frequencies and their uncertainties were assessed using the monte - carlo simulation algorithm .
the results were compared to those derived from the phase dispersion minimization ( pdm ) method and period04 . to perform o - c calculations and to investigate period variabilities over years , the best extremum times were obtained from the seasonal lcs , and were put together with data from the literature .
l l l l r * v473tau * & * ls*&*period04*&*pdm*&*amp . *
+ & * ( c d@xmath2)*&*(c d@xmath2)*&*(c d@xmath2)*&*(mmag ) * + 2007 & 0.7104(8 ) & 0.7101(8 ) & 0.7120(16 ) & 8.63(25 ) + 2008 & 0.7101(6 ) & 0.7101(6 ) & 0.7118(15 ) & 10.93(24 ) + 2009 & 0.7116(7 ) & 0.7116(8 ) & 0.7157(15 ) & 8.62(24 ) + 2010 & 0.7128(7 ) & 0.7128(8 ) & 0.7123(20 ) & 9.20(25 ) + comb . &
0.710818(5 ) & 0.710818(7 ) & 0.711164(5 ) & 9.33(13 ) + [ tab : table1 ]
in this research , we obtained four consecutive years of data between 2007 and 2010 .
as reported by other researchers , all the lcs had explicit periodicity .
individual ls , pdm and period04 analyses of the seasonal curves showed a frequency at around 0.71 c d@xmath2 ( table [ tab : table1 ] ) .
furthermore , we detected the existence of another strong peak at approximately 1.40 c d@xmath2 ( 0.71 days ) on the ls periodogram ( figure [ fig : figure1 ] , top right ) . combining the four - year data , the precise rotation period of the star ( 0.710818 c d@xmath2 ) was determined with the help of the ls method , and the main lc was plotted based on this value ( figure [ fig : figure1 ] , top left ) .
accordingly , the folded lc was clearly formed by a maximum and a broad minimum .
the maximum was quite strong and had a peaked top .
moreover , there was a barely detectable bump in the middle of the minimum , which is probably the cause of the peak at 1.40 c d@xmath2 on the ls periodogram .
also , we produced a folded lc using the double _ stereo _ period since maitzen @xcite noted that his observations were compatible with the period value of 2.7795 days .
we derived a relatively clean lc with two minima and maxima .
however , since the consecutive structures appeared similar to each other , we assumed that the period value of 0.710818 c d@xmath2 was the full rotation period . to investigate a possible period variation ,
the best maximum times of the seasonal lcs were derived ( hjd24554583.4049 , 24554922.4465 , 24555274.1565 , 24554241.5599 ) , and these values were combined with the epochs from the literature . the archival epochs given as jd
were converted to hjd .
based on figure [ fig : figure1 ] ( bottom ) , we found out that the o - c diagram of the star exhibited a period decrease with the variation rate of around @xmath12 d y@xmath2 or @xmath1 s y@xmath2 ( parabolic line ) . with the help of the ls period and the _ stereo _ maximum time having the smallest error value , we determined the light elements as : @xmath13 since this star was a single rotating variable , such a period decrease might most likely be explained by an acceleration in rotation after a magnetic braking .
using the period and its variation rate , the acceleration time - scale of the star ( @xmath14 ) was determined as @xmath15 yr .
we also found the main sequence lifetime of the star ( @xmath16 ) as @xmath17 yr ( table 2 ) .
c c c c c c @xmath18 & @xmath19 & @xmath20 & @xmath21 & @xmath22 + ( day ) & ( s yr@xmath2 ) & ( s@xmath2 ) & ( yr ) & ( yr ) + 1.406830(10 ) & -0.11(3 ) & -2.86(68)@xmath23 & -1.11(63)@xmath24 & 9.26(1.25)@xmath25 + @xmath26l@xmath27 & @xmath28 & @xmath29 & @xmath30 & @xmath31 + & & ( m@xmath32 ) & ( r@xmath32 ) & ( km s@xmath2 ) + 1.64(15 ) & 4.045(11 ) & 2.59(14 ) & 1.80(32 ) & 65(12 ) + * * temperature , luminosity and mass values were adopted from wraight et al .
radius and rotational velocities were estimated from the ls period , temperature , and luminosity values .
[ tab : table2 ] the variation in period of v473tau ( @xmath20 in table 2 ) is 10 times greater than that of the most massive cp stars given by mikulasek et al . @xcite .
in addition , its acceleration is nearly three orders of magnitude shorter than the main sequence lifetime .
this , in turn , suggests that process of decrease in the period may be reversible .
if so , the length of the cycle is roughly calculated as 92(11 ) yr ( estimated by @xmath33 ; @xcite ) .
considering the fact that period variation processes may be reversible due to shorter acceleration time - scale than that of the main sequence lifetime , the rigid rotation hypothesis should be discarded and the differential rotation model should alternatively be discussed as expressed by stepien @xcite . if a differential rotation is indeed in question for this star , the acceleration in the rotation may be interpreted as a consequence of torsional oscillations produced by meridional circulations being in interaction with a magnetic field , and of rotational braking in outer layers caused by angular momentum loss via magnetically - confined stellar wind @xcite .
we acknowledge assistance from vino sangaralingam and gemma whittaker in the production of the data used in this study .
the stereo heliospheric imager was developed by a collaboration that included the rutherford appleton laboratory and the university of birmingham , both in the united kingdom , and the centre spatial de lige ( csl ) , belgium , and the us naval research laboratory ( nrl),washington dc , usa . the stereo / secchi project is an international consortium of the naval research laboratory ( usa ) , lockheed martin solar and astrophysics lab ( usa ) , nasa goddard space flight center ( usa ) , rutherford appleton laboratory ( uk ) , university of birmingham ( uk ) , max - planck - institut fr sonnen - systemforschung ( germany ) , centre spatial de lige ( belgium ) , institut doptique thorique et applique ( france ) and institut dastrophysique spatiale ( france ) . hubrig s. , et al . , 2005 ,
a&a , 440 , l37 mikulasek z. , szasz g. , krticka j. , zverko j. , ziznovsky j. , zejda m. , graf t. , 2009 , arxiv , arxiv:0905.2565 kochukhov o. , 2011 , iaus , 273 , 249 ] burke e. w. , jr .
, rolland w. w. , boy w. r. , 1970 , jrasc , 64 , 353 ] rakosch k. d. , fiedler w. , 1978 , a&as , 31 , 83 maitzen h. m. , 1977 , a&a , 60 , l29 ] jerzykiewicz m. , 2009 , aca , 59 , 307 eyles c. j. , et al .
, 2009 , soph , 254 , 387 sangaralingam v. , stevens i. r. , 2011 , mnras , 418 , 1325 ] whittaker , g. , sangaralingam , v. , and stevens , i. : 2012 , _ from interacting binaries to exoplanets : essential modeling tools _ * 282 * , 143 . mikulasek z. , krticka j. , jank j. , zejda m. , henry g. w. , paunzen e. , ziznovsky j. and zverko j. , 2014 , arxiv , arxiv:1310.6640 mikulek z. , krtika j. , henry g. w. , de villiers s. n. , paunzen e. , zejda m. , 2010 , a&a , 511 , l7 stpie k. , 1998 , a&a , 337 , 754 wraight k. t. , fossati l. , netopil m. , paunzen e. , rode - paunzen m. , bewsher d. , norton a. j. , white g. j. , 2012 , mnras , 420 , 757 | in this paper , the period evolution of the rotating chemically peculiar star v473tau ( a0si , v = 7.26 mag ) is investigated . even though the star has been observed for more than fifty years , for the first time four consecutive years of space - based data covering between 2007 and 2010
is presented .
the data is from the _ stereo _ satellite , and is combined with the archival results .
the analysis shows that the rotation period of v473tau is @xmath0 days , and has slightly decreased with the variation rate of @xmath1 s yr@xmath2 over time .
also , the acceleration timescale of the star is found to be around @xmath3 yr , shorter than its main sequence lifetime ( @xmath4 yr ) .
this indicates that the process of decrease in period might be reversible .
on this basis , it can be suggested that v473tau has a possible magnetic breaking and a differential rotation , which cause a variation in the movement of inertia , and hence the observed period change . |
irreversible adsorption of particles at solid and liquid interfaces is of major significance for many fields such as medicine and material sciences as well as pharmaceutical and cosmetic industries e.g. @xcite .
for example adsorption of some proteins plays crucial role in blood coagulation , fouling of contact lenses , and plaque formation .
the process is also used in membrane filtration units .
additionally , controlled adsorption is fundamental for efficient chromatographic separation and purification , gel electrophoresis as well as biosensing and immunological assay performance .
adsorption layers formed on flat , homogeneous surface typically exhibit random structure without any signs of order . however , a number of latest studies on adsorption focuses on deposition on specially prepared , patterned surfaces @xcite .
it is due to the possibility of gaining better control over adsorption process and the properties of resulting mono- and multilayers . in such experiments ,
adsorbed particles take on order of the base . on the other hand ,
anisotropic particles can form orientationally ordered layers even on flat , homogeneous surfaces .
ordering can originate from non - uniform alignment of particles in a bulk phase @xcite or appear spontaneously in a form of ordered domains .
the second is the subject of investigation in our study . as far as we know ,
such ordering hasnt been investigated before .
although local ordering has been reported in earlier numerical studies on dimer adsorption @xcite , not even basic properties of orientationally ordered domains have been further investigated . here
, we focus on domain sizes and their spatial distribution for different range of repulsive interaction between adsorbed particles .
adsorbed particle is modeled by two identical touching spheres of radius @xmath0 ( see fig.[fig : model ] ) .
this model was chosen because it is the simlest numerical model of anisotropic particle .
moreover , it has been carefully studied and basic properties of the adsorbed layers are well known @xcite .
it can be also used as a coarse - grained model for some biomolecules @xcite .
electrostatic interaction between two dimers originates from electrostatic potentials generated by individual spheres . in typical substrate
this interaction is approximated by an exponentially decaying yukawa potential known also as screened electrostatic potential @xcite : @xmath1 & \mbox{for $ l_e > 0 $ } , \\ 0 & \mbox{for $ l_e = 0 $ } , \end{array } \right.\ ] ] where @xmath2 and @xmath3 are particle diameter and charge , respectively ; and @xmath4 is distance between individual spheres from different particles .
@xmath5 is a constant parameter , @xmath6 is a vacuum permittivity and @xmath7 characterizes dielectric properties of a solvent .
range of the interaction is controlled by parameter @xmath8 , known also as the debye screening length .
it is a key parameter for electrostatic interaction in electrolytes and can be calculated by @xcite : @xmath9 where @xmath10 is the boltzmann constant , @xmath11 is temperature , @xmath12 is elementary charge , and @xmath13 is ionic strength of electrolyte solution .
adsorption layers were generated using random sequential adsorption algorithm introduced by feder @xcite .
it is based on repeating the following steps : i : : a new virtual dimer is created on a collector .
its position and orientation are selected randomly according to uniform probability distribution ; ii.a : : the virtual particle undergoes overlapping test with its nearest neighbors ; ii.b : : the total electrostatic potential @xmath14 between the virtual molecule and previously adsorbed dimers is calculated using eq.([uel ] ) ; iii : : if there is no overlap the virtual particle is added to the existing layer with a probability @xmath15 ( * ? ? ?
* ; * ? ? ?
* ) . otherwise the virtual dimer is removed and abandoned .
the whole procedure is repeated for a specified number of times , which guarantees close approach to the jamming limit .
the simulation of adsorption process was performed for a square collector of @xmath16-side size and was stopped after @xmath17 iterations . for each set of parameters , 20 to 100 independent simulations were performed .
parameters of electrostatic potential ( [ uel ] ) were chosen based on @xcite to describe typical experimental conditions of water solutions .
therefore , the thermal energy @xmath18 at a room temperature acts as energy unit and @xmath19 .
the @xmath20 provides length scale typical to mid sized bio - molecules , and @xmath21 .
particle is treated as a domain member when it is close to other domain components and its orientation does not differ much from the mean domain orientation .
however , specific threshold values of distance and angle may have significant impact on domain size and shape .
moreover , mean distance between particles in a layer depends on electrostatic repulsion range @xmath8 .
it means that the same distance threshold value for all @xmath8 will prevent domains to be classified as such when interaction range is large . on the other hand
, such `` soft '' interaction affects the effective anisotropy of a particle which changes the tendency for domain formation .
fig.[fig : ratio ] shows ratio of particles classified as domain members in respect to specific values of distance and angle difference used in classification process .
.,width=302 ] for the purpose of the present study , the concept of domain is defined according to refs.@xcite .
particle is domain member when it fulfills both of the following conditions : i : : its angle does not differ from mean domain orientation more than @xmath22 ; ii : : its monomer - to - monomer distance to the closest domain component is not larger than @xmath23 , where @xmath24 is characteristic size of an interacting dimer .
distance is measured between the centres of monomers .
parameter @xmath24 depends on @xmath8 . for hard core interaction ( @xmath25 ) , @xmath26 .
with the growth of @xmath8 parameter @xmath24 increases , to compensate lower coverage ratio @xmath27 of a layer : @xmath28 for convenience , @xmath24 values used in this study is collected in tab.[tab : sigma ] .
.parameter @xmath24 ( [ eq : sigma ] ) for different electrostatic interaction range @xmath8 .
values of maximal random coverage ratios @xmath27 were taken from @xcite . [ cols="^,^,^",options="header " , ] [ tab : sigma ] using the above definition of a cluster , typically , @xmath29 to @xmath30 of dimers are classified as domain members .
adsorption of dimers not interacting electrostatically ( @xmath25 ) provides reference level for further investigations .
the first important observation is that appearance of orientationally ordered dimers clusters is typical for adsorbed layers generated by rsa algorithm ( see fig.[fig : clusters ] ) .
.,title="fig:",width=151 ] .,title="fig:",width=151 ] the distribution of domain sizes is presented in fig.[fig : hist00 ] . )
interacting dimer layers .
dots are measured data whereas solid line is an exponential fit : p(n ) = @xmath31 .
insets shows fitted exponentials to domain size distribution for partially filled layers.,width=226 ] its exponential decay suggests that there is constant probability of classifying particle as a member of orientationally ordered cluster .
the probability read from plot is equal to @xmath32 .
this value is much higher than @xmath33 , which is the probability of random selection of particle orientation within @xmath34 $ ] interval . on the other hand , a single cluster borders on more than one dimer , which increases the probability of finding nearby parallely aligned particle .
therefore , the above results do not answer the question whether domain formation is a merely statistical effect or rather intrinsic property of elongated particle adsorption process . to check it , we analysed cluster statistic before the maximal random coverage has been reached . for result comparison purposes ,
domain classification criteria had to be changed due to lower particle density according to ( [ eq : sigma ] ) .
for example , @xmath35 for half of particles inside a layer .
results presented in fig.[fig : hist00 ] inset show that denser layers are characterized by larger probability of parallel alignment .
for example , after adsorption of only @xmath36 of the particles , the probability of classifying particle as a domain component is @xmath37 , which is significantly smaller than the value for maximally covered layers .
therefore , it can be concluded that random sequential adsorption of anisotropic particles leads to formation of orientationally ordered clusters even in environment with no specific direction .
anisotropic molecules characterised by soft interactions also form orientationally ordered clusters . moreover , the domain size distribution here is exponential , too .
the probability of attaching particle to cluster decreases with interaction range ; however , it has incomprehensible high value for @xmath38 , which requires further study ( see fig.[fig : a_le ] ) . for both : hard core and electrostatically interacting dimers.,width=226 ] besides the one result , obtained exponents are similar to the ones for hard - core interacting dimers at earlier stage of adsorption ( @xmath39 ) .
it suggests that the main factor affecting formation of orientationally ordered clusters is the density of particles inside layers .
the result is not as entirely expected , as soft interaction generated by screened electrostatic potential effectively decreases anisotropy of dimers , as reported in @xcite . on the other hand , soft interaction increases distance , at which shape anisotropy
can be noticed by other particles .
as shown , both effects compensate each other , at least for @xmath40 .
dimer adsorption on homogeneous flat surfaces was investigated numerically using random sequential adsorption ( rsa ) algorithm .
formed layers contain orientationally ordered clusters of particles .
their size distribution is exponential .
the exponent depends mainly on particle density inside a layer but not on the character of interaction between particles , at least when interaction range is small .
99 m. ciesla , j. barbasz , j. stat .
p03015 ( 2012 ) .
m. ciesla , j. barbasz , arxiv:1210.8098 ( 2012 ) .
a. rafati , a. boussahel , k.m .
shakesheff , a.g . shard ,
roberts , x. chen , d.j .
scurr , s. rigby - singleton , p. whiteside , m.r .
alexander , m.c .
davies , j. contr .
* 162 * 321 ( 2012 ) . c. galli , m. collaud coen , r. hauert , v.l .
katanaev , m.p .
wymann , p. groning , l. schlapbach , surf .
, * 474 * l180 ( 2001 ) .
browne , c.v .
lubarsky , m.r .
davidson , r.h .
bradley , surf .
, * 553 * 155 ( 2004 ) .
bae , b.k .
lee , w.h .
lee , j.w .
choi , biosens .
bioelectron . , * 21 * 103 ( 2005 )
cho , h. kong , t.b .
oh , k. cho , soft matter , * 8 * 11801 ( 2012 ) . t.f .
keller , j. schonfelder , j. reichert , n. tuccitto , a. licciardello , g.m.l .
messina , g. marletta , k.d .
jandt , acs nano , * 5 4 * 3120 ( 2011 ) .
j. zemla , j. rysz , a. budkowski , k. awsiuk , soft matter , * 8 * 5550 ( 2012 ) .
m. ciesla , j. barbasz , arxiv:1208.0167 ( 2012 ) .
v. tozzini , current opinion in structural biol . * 15 * 144 ( 2005 ) .
b. senger , j .- c .
voegel , p. schaaf , colloids and surfaces a : physic- ochem .
aspects * 165 * 255 ( 2000 ) .
m. trulsson , j. forsman , t. akesson , b. jonsson , langmuir bf 25(11 ) 6106 ( 2009 ) .
debye , e. huckel , phys . z.
* 24 * 185 ( 1923 ) .
z. adamczyk , _ particles at interfaces : interactions , deposition , structure _ , elsevier / academic press : amsterdam , 2006 .
z adamczyk , m zembala , b siwek , p warszynski , j. colloid interface sci . 140 ( 1990 ) 123 .
oberholzer , n.j .
wagner , a.m. lenhoff , j. chem .
( 1997 ) 9157 .
j. feder , j. theor .
87 * 237 ( 1980 ) .
s. fujiwara , t. sato , phys .
* 80 * 5 991 ( 1998 )
. k. esselink , p.a.j .
hilbers , b.w.h .
van beest , j. chem .
* 101 * , 9033 ( 1990 ) . | structure of monolayers built during adsorption process is strongly related to the properties of adsorbed particles .
the most important factor is their shape .
for example , adsorption of elongated molecules on patterned surfaces may produce certain orientational order inside a covering layer .
this study , however , focuses on random adsorption of dimers on flat , homogeneous surfaces .
it has been observed that despite the lack of global orientational ordering , adsorbed dimers may form local , orientationally ordered structures @xcite .
our investigations focus on the dependence between domain size distribution and environmental parameters such as ionic strength , which affects the range of electrostatic interaction between molecules . |
while our knowledge about the universe has improved over the last decades with the advent of new observational data , there are several dark sides of the universe that have not been so far described from a fundamental point of view : what caused the initial inflationary era of the universe ? what is the origin of dark matter ? what is the fundamental cause of the current acceleration of the universe ? even though nowadays none of the previous issues have a satisfactory answer , a parallel approach , that can shed some light on the dark sides of the universe , is a phenomenological one or a model building strategy .
a good example in this regard is the inflationary paradigm @xcite , where an inflaton field ( or several scalar fields ) induce the initial acceleration of the universe .
such a field would leave tracks on the evolution of the universe for example through the primordial power spectrum of the scalar perturbations @xcite , extremely useful to constrain the inflaton field .
it is precisely such an approach that we will follow in this paper .
the main goal of this paper is to obtain a phenomenologically consistent model for the early universe ( inflationary and radiation dominated epochs ) by properly modifying the generalised chaplygin gas ( gcg ) @xcite .
a first attempt in this direction has been recently carried out in @xcite ( see also @xcite ) where a new scenario for the early universe was proposed .
such a scenario provides a smooth transition between an early de sitter - like phase and a subsequent radiation dominated era . in that model
@xcite , the matter content was given by a type of generalized chaplygin gas for the early universe , with an underlying scalar field description .
here , we give a more _ realistic _ model where the early inflationary phase of the universe is more general than a de sitter - like universe .
we will show how the spectrum of the present model is not as red as the one presented in @xcite and therefore more consistent with observations .
more precisely , rather than having a de sitter - like phase in the past , we will consider an early `` quintessence '' inflationary phase .
this phase will be connected to a radiation dominated phase at later time .
the model can be described through a scalar field or a chaplygin gas inspired model .
we will then analyze the possible imprints of such a gas in the primordial power spectrum of scalar perturbations and the power spectrum of the stochastic background of gravitational waves .
the paper is organized as follows . in section
ii , we will present the model which is based on a properly modified chaplygin gas that interpolates between a quintessence inflationary era and a radiation dominated period . in section iii , we will show how this kind of gas can be modeled by a minimally coupled scalar field . in section iv
, we constrain our model observationally and obtain the full spectrum of scalar perturbations numerically . in section
v , we obtain the spectrum of gravitational waves using the method of bogoliubov coefficient .
finally , in section vi we summarize our results and conclude .
we generalize the model presented in @xcite by considering an inflationary period corresponding to a `` quintessence '' like behavior ( described by a power law expansion ) and followed by a radiation dominated epoch .
the matter content of the universe can then be modeled _ la _ chaplygin gas as @xmath0 where @xmath1 are constants being @xmath2 positive .
the matter content is not interacting with any other fluid and therefore its energy density is conserved : @xmath3 where the dot stands for derivative with respect to the cosmic time and @xmath4 is the pressure of the fluid .
if we substitute the energy density ( [ rho ] ) in the conservation equation ( [ conservation ] ) , we obtain @xmath5 the above equation of state can be rewritten as @xmath6 this equation shows clearly that we recover the model discussed by one of us in @xcite for @xmath7 .
we would like to highlight that the equations of state ( [ p])-([p2 ] ) have been previously analyzed in @xcite to study a possible interplay between dark matter and dark energy and therefore in a scenario completely different to the one we study here .
as the inflationary period takes place before the radiation dominated one , we deduce that the inequality @xmath8 must hold at early time ; i.e. at small scale factors
. this will be the case as long as @xmath9 the inequality ( [ inequality1 ] ) implies @xmath10 , where @xmath11}.\ ] ] the scale factor @xmath12 represents the `` border line '' between two different regimes .
more precisely @xmath13 there are two more conditions that should be imposed on @xmath14 and @xmath15 to have a model that interpolates between an early inflationary phase of the type of quintessence and a radiation dominated phase at later time : ( i ) the energy density ( [ rhoinf ] ) must induce a period of inflation and ( ii ) such a period of inflation should not induce a super - inflationary expansion ; i.e. @xmath16 , and therefore a super - accelerating phase of the universe where the energy density ( [ rhoinf ] ) corresponds to phantom matter , with an energy density that grows as the universe expands . by combining these two ansatz with the inequality ( [ inequality2 ] )
, we can easily deduce that the set of allowed values of @xmath14 and @xmath15 satisfy is not that surprising ( see for example @xcite ) .
indeed , it has been shown in @xcite that type ia supernovae allow for negative @xmath14 under the standard generalized chaplygin gas . ]
@xmath17 a perfect fluid with an equation of state ( [ p ] ) or ( [ p2 ] ) describe an early inflationary period as long as the conditions ( [ conditions ] ) are satisfied .
the universe will exit the inflationary epoch when the scale factor reaches the value @xmath18^{1/[1+\beta-4(1+\alpha)]}.\ ] ] the last condition can be obtained by imposing the condition @xmath19 .
again this is in agreement with the results of @xcite for @xmath7 . before concluding this section ,
let us analyze more closely what sort of inflation does the universe undergo in its initial stages .
the friedmann equation with the matter content ( [ rho ] ) is too difficult to be integrated analytically so we use an approximation where the energy density can be written as in eq .
( [ rhoinf ] ) .
then the scale factor can be approximated as @xmath20}\sqrt{\frac{\kappa^2}{3}}t\right]^{2(1+\alpha)/(1+\beta ) } , \label{powerlaw}\ ] ] where @xmath21 is the cosmic time and @xmath22 where @xmath23 is the gravitational constant . as we can see from the previous equation the inflationary era of the universe
is given by a power law expansion .
in sections iv and v , we will see that the entire evolution of the universe can be described by the appropriately modified chaplygin gas ( [ p2 ] ) and a subsequent @xmath24cdm expansion , that is @xmath25 where @xmath26 is the current scale factor , @xmath27 , @xmath28 and @xmath29 are the current energy densities corresponding to radiation , matter ( cold dark matter and baryonic matter ) and dark energy ( modelled through a cosmological constant ) , respectively . on the other hand , at the radiation dominated epoch the energy densities ( [ energydensities ] ) are equal , consequently @xmath30 the parameter @xmath31 is related to the scale of inflation ( see eqs .
( [ rhoinf ] ) and ( [ vapowerlow ] ) ) . , against @xmath32 , where @xmath32 corresponds to a power of the scale factor @xmath33 , @xmath34 , with @xmath35 .
the values @xmath36 and @xmath37 correspond to the moments when the pivot scale @xmath38 exists the horizon and the inflation ends , respectively.,width=302 ]
the inflationary dynamics of the model presented in the previous section can be described through a minimally coupled scalar field , @xmath39 , with a potential , @xmath40 , whose energy density and pressure read @xmath41 in the previous equations the prime stands for derivative with respect to the conformal time .
this scalar field can be mapped to the perfect fluid with equations of state ( [ p ] ) or ( [ p2 ] ) ; i.e. @xmath42 and @xmath43 . through this mapping
we can write down @xmath39 and @xmath44 as a function of the scale factor ( see fig .
[ phixa ] ) : @xmath45\right\},\nonumber\\ \label{phia}\\ v(a)&=&a^{1/(1+\alpha)}\left(\frac{a}{b}\right)^{-(1+\beta)/[q(1+\alpha ) ] } x^{-(1+\beta)/[q(1+\alpha ) ] } \nonumber\\ & \ , & ( 1+x)^{1/(1+\alpha)}\left[\frac13-\frac{q}{6(1+\alpha)}\frac{1}{1+x}\right],\label{va}\end{aligned}\ ] ] where @xmath34 and @xmath35 .
the scalar field starts with a negative value and it rolls down the potential as the universe inflates ( cf . fig .
[ v1 ] ) . against @xmath39 ( see eqs .
( [ phia])-([va ] ) ) , where @xmath46}$ ] .
the values @xmath47 and @xmath48 correspond to the moments when the pivot scale @xmath38 exists the horizon and the end of inflation , respectively.,width=302 ] we can also consider the behavior of the equation of state parameter @xmath49 , @xmath50 , for the scalar field as a function of the scale factor ( see fig . [ w ] ) . at early times , the kinetic energy is negligible with respect to the potential , and therefore @xmath51 .
afterwards , the parameter @xmath49 increase gradually with time .
the universe stops inflating when @xmath49 reaches the value @xmath52 and becomes radiation dominant when @xmath49 finally reaches @xmath53 . for the modified gcg model with equation of state ( [ p ] ) or ( [ p2 ] ) .
the dashed red line indicates the end of inflation .
we set @xmath54 as an example here.,width=302 ] at very early times where the scale factor is very small , it can be shown that the inflationary expansion follows a power law .
indeed , the scalar field and the potential can be approximated by @xmath55 a^{1/(1+\alpha)}a^{-(1+\beta)/(1+\alpha)},\\ \phi_0&=&\frac{1}{q\kappa}\left[\sqrt{\frac{1+\beta}{1+\alpha}}\ln\left(\frac{qb}{4|1+\alpha|a}\right)\right.\nonumber\\ & \ , & \left.+4\tanh^{-1}\sqrt{\frac{1+\beta}{4(1+\alpha)}}\,\,\right],\end{aligned}\ ] ] and therefore @xmath56a^{1/(1+\alpha)}\exp\left[-\kappa\sqrt{\frac{1+\beta}{1+\alpha}}\left(\phi-\phi_0\right)\right].\nonumber\\ \label{vapowerlow}\end{aligned}\ ] ] at much later times where the scale factor becomes large , the universe is radiation dominant .
it can be easily proved that for this period : @xmath57.\end{aligned}\ ] ] consequently , we obtain @xmath58 .
\label{powerlawpotential}\ ] ] since we know the form of the potential @xmath40 for this modified gcg model , we are able to find the evolution of the slow - roll parameters @xmath59 and @xmath60 , where @xmath61 the slow - roll approximation is valid when the conditions @xmath62 are satisfied , which means the potential energy dominates over the kinetic term during the inflation era . in fig .
[ slowroll ] , we show the behavior of these functions in terms of the scale factor @xmath33 . and the blue one to @xmath63 .
the dashed red line locates the time when inflation ended .
we also notice that the slow - roll conditions are no longer satisfied when the field is close to the end of inflation ( @xmath64 ) .
we set @xmath54 as an example here.,width=283 ]
inflation not only solve some of the shortcomings present in the big bang theory but also generates density perturbations that seeds the structure of the present universe .
those density perturbations have been constrained through observations of the cosmic microwave background ( cmb ) . in this section
, we will constrain the model introduced in the previous section by using the measurements of wmap7 @xcite for the power spectrum of the comoving curvature perturbations , @xmath65 , and its index , @xmath66 , where @xmath67 these measurements correspond to a pivot scale @xmath68 @xcite . once the parameters of the model have been constrained
we will obtain the full spectrum of the scalar perturbations . at very early times , the model introduced in sect .
ii induces a power law expansion as shown clearly in eq .
( [ powerlaw ] ) ( see also eq .
( [ vapowerlow ] ) and fig .
[ v1 ] ) . on the other hand ,
the slow - roll approximation is valid for an inflationary power law expansion as the one we are considering ( see for example figs .
[ w ] and [ slowroll ] ) . in this regime ,
the power spectrum ( for the comoving curvature perturbations ) and the spectral index can be expressed as ( see for example @xcite ) @xmath69 at the horizon exit ; i.e. at @xmath70 . throughout the paper a dot stands for a derivative with respect to the cosmic time . for a power law expansion ,
i.e. , @xmath71 where @xmath72 is the conformal time , the spectral index reduces to @xmath73 . for the modified chaplygin gas we are analyzing @xmath74 . with the previous inputs , we can proceed as follows : ( i )
the parameter @xmath75 is fixed by the current amount of radiation in the universe as stated in eq .
( [ bdef ] ) , ( ii ) for a given parameter @xmath14 , the parameter @xmath15 is fully determined by the measurement of @xmath76 and ( iii ) the parameter @xmath31 is fixed such that @xmath65 at the pivot scale @xmath68 . before moving forwards
, we would like to highlight that the mode @xmath68 exits the horizon well inside the inflationary era where both the power law expansion and the slow - roll approximation are valid for the model ( [ p2 ] ) .
we found out that the inflationary scale , @xmath77 , is almost of constant .
the tensor power spectrum to the scalar power spectrum ratio in the slow - roll approximation can be expressed through the slow - roll parameter @xmath78 @xcite : @xmath79 for power - law inflation in the slow - roll approximation , the spectral index is @xmath73 , as we already mentioned , while @xmath80 . because this gcg model behaves like a slow - roll power - law inflation when the pivot @xmath81 exits the horizon , we can use these approximations to evaluate @xmath82 at that time .
using once more wmap7 data ( @xmath83 at the scale @xmath84 ) we can find @xmath85 .
we obtain @xmath86 at the pivot scale @xmath81 , which is in agreement with the wmap7 constraints .
we obtain the full spectrum of the gravitational waves on section v. at later times close to the end of inflation or when larger modes @xmath87 exit the horizon , the slow - roll conditions are not fulfilled .
this is a simple consequence of the behavior of the gcg model that deviates from a power - law inflation at that time .
thus we can not use the results in eqs.([pssr ] ) and ( [ nssr ] ) when inflation approaches @xmath88 where it halts . in this regime
, we will rather calculate the spectrum numerically .
the primordial density perturbation can be directly mapped to the comoving curvature perturbation , the last remains constant on large scale after exiting the hubble horizon , as long as the perturbations are adiabatic .
this is precisely the case , as we have a unique degree of freedom corresponding to the generalised chaplygin gas .
the comoving curvature perturbation , @xmath89 , is determined by the fluctuations of the generalised chaplygin gas .
the corresponding power spectrum for the field @xmath90 is @xcite @xmath91 where @xmath92 .
the field @xmath90 satisfies in the fourier space the equation @xcite @xmath93 if the slow - roll conditions are satisfied , the variation of the field @xmath39 and the hubble parameter @xmath94 is much slower than that of the scale factor @xmath33 .
therefore , the following approximation can be used : @xmath95 , at the lowest order of the slow - roll approximation .
consequently , the solutions of ( [ seof ] ) read in this case @xcite @xmath96 where the bunch - davies vacuum has been imposed for modes well inside the horizon ( @xmath97 ) .
we remind that the parameter @xmath72 stands for the conformal time . now , if we consider the super - horizon scale @xmath98 and substitute the approximation @xmath99 into eq .
( [ nusol ] ) , the solution becomes @xmath100 the corresponding power spectrum in this slow - roll approximation is given in eq .
( [ pssr ] ) .
these results remain valid on the next order of the slow - roll as shown for example in @xcite and therefore can be applied for an inflationary power law expansion as it is the case in the model we are analyzing at very early time . in order to obtain the full power spectrum @xmath101
, we use numerical methods instead of applying eq .
( [ pssr ] ) for the reasons stated in the previous subsection .
therefore , we first solve the differential equation ( [ seof ] ) , and find the solution @xmath102 calculated at the time when the mode @xmath87 exits the horizon ; i.e. @xmath103 . then through eq .
( [ powerspectrum ] ) we obtain the power spectrum @xmath104 .
we proceed as follows .
we separate eq .
( [ seof ] ) into two first order differential equations : @xmath105 where we set @xmath106 . in order to solve the previous set of differential equations ,
we take the following actions : * it is easier to make a change of variable from the conformal time to the scale factor .
notice that @xmath107 can be fully determined in terms of the scale factor as @xmath108 and the scalar field @xmath39 is fully determined as a function of the scale factor ( see eq .
( [ phia ] ) ) . * in addition , we have simply to use the set of values for the parameters of our model that have been deduced by imposing observational constraints on the model as discussed in the previous subsection . *
last but not least , we need to impose a set of boundary conditions . when the wavelength of a given mode @xmath87 is much smaller than the hubble radius @xmath109 , the effect of curvature can be neglected .
therefore , the result reduces to that of a flat minkowski spacetime ( when @xmath109 ) .
so , the initial condition is @xmath110 which corresponds to the minkowski vacuum at very early time .
we change the variable from the conformal time @xmath72 to the scale factor @xmath33 through the relation @xmath111\sqrt{\frac{8\pi g}{3}}a^{1/[2(1+\alpha)]}\tau\right\}^{1/\left[\frac{1+\beta}{2(1+\alpha)}-1\right]}. % \quad(\mbox{power - law inflation } ) , \label{atau}\ ] ] following a similar procedure , we can see if the spectral index , @xmath76 ( defined in eq .
( [ defns ] ) ) , depends on the scale .
our results are shown in figs .
[ ps ] and [ ns ] . against @xmath87 for six different values of @xmath14 .
the dashed black line is the pure power law inflation , and the vertical dashed red line locates the pivot @xmath84 .
we can see that all these lines merge when small @xmath87 ; i.e. large scale , exits the horizon .
the grey , violet , red , orange , green and blue curve correspond respectively to @xmath112,width=340 ] against @xmath87 . the red dashed line locates the pivot scale @xmath84 .
the spectral index @xmath76 is approximately a constant at early time .
the grey , violet , red , orange , green and blue curve correspond respectively to @xmath112,width=309 ] for lower modes , i.e. , those that entered the horizon very recently , the power spectrum is independent of the parameter as explained in the previous subsection . ]
it is only for modes satisfying @xmath113 that we can start to see some dependence of @xmath114 on @xmath14 .
as should be expected , the power spectrum has a constant slope ; i.e. , a constant spectral index , for lower @xmath87 when the modified chaplygin gas induces a power - law expansion .
indeed , the results for power - law matches very well those of our model for these modes ( cf . figs . [ ps ] and [ ns ] ) .
however , the spectral index of the primordial power spectrum is not scale independent as fig .
[ ns ] clearly shows .
the deviation of @xmath76 from the scale independence starts when the slow - roll condition ceases to be valid and therefore approximately a bite before the end of inflation . before concluding this section
, we recall that the current cmb observations cover roughly the range of scale @xmath115 to @xmath116 @xcite . in this observable region
, it is hard to separate a modified gcg model for different values of @xmath14 or even from a power - law inflation .
could currently running observations or future missions help us in this regard ?
for example , it is expected that the maximum scale @xmath117 in cmb observations from the planck mission would reach @xmath118 .
this scale @xmath119 could be translated into a @xmath120 scale .
even though the relation between @xmath117 and @xmath87 is not a one - to - one relationship , we can still find the main contribution for each mode : @xmath121 @xcite , where @xmath122 is the elapsed conformal time since the last scattering surface until the present .
here we use the concordance @xmath24cdm model to calculate @xmath123 numerically .
we find that the maximum scale corresponding to @xmath118 is @xmath124 .
consequently and unfortunately , the answer to the question raised in this paragraph is negative , as is clear from figs .
[ ps ] and [ ns ] .
we will next show that by looking at the high frequency spectrum of the gravitational waves , it is possible to separate a modified gcg model for different values of @xmath14
. those frequencies might be within the reach of future gravitational - wave detectors such as bbo and decigo @xcite .
the early inflationary era not only leaves imprints on the scalar cosmological perturbations but also creates a fossil of gravitational waves . in this section
, we will analyze the possible imprints in the power spectrum of the stochastic background of gravitational waves , for the model presented in section ii . in this case the background evolution of the universe , till nowadays , is described by the matter content in eq .
( [ energydensities ] ) .
this analysis is quite important as it can shed some light on the inflationary scenario behind the early accelerating phase of the universe and its transition to the subsequent radiation era . in order to obtain the spectrum of the gravitational waves ( gws )
, we will use the method of bogoliubov coefficients . to our knowledge , this method was first developed in @xcite and later applied in @xcite .
bogoliubov coefficients describe how the vacuum changes as the universe expands . in particular ,
one of these coefficients , which we will denote @xmath125 , gives the number of gravitons created .
furthermore , it can be shown that the dimensionless relative logarithmic energy spectrum of the gravitational waves , @xmath126 , at the present time reads @xcite : @xmath127 the parameter @xmath128 is the energy density of gws and @xmath129 the respective angular frequency ; @xmath130 and @xmath94 are the critical density of the universe and hubble parameter , respectively , evaluated at the present time . in summary ,
the present value of the bogoliubov coefficient , @xmath125 , determines the power spectrum of the gravitational waves .
this parameter can be determined in terms of two continuous functions @xmath131 such that @xmath132 , where @xmath131 fulfil @xmath133 the coefficient @xmath125 gives the number of gravitons , @xmath134 , for each mode @xmath87 created during the evolution of the universe , where @xmath135 . .
the dashed line corresponds to @xmath136 , and the solid line corresponds to @xmath137 for a modified gcg model.,width=313 ] to calculate the gws spectrum ( [ spectrum ] ) , we have to solve the differential equations ( [ ode ] ) numerically and use appropriate initial conditions for @xmath138 and @xmath139 .
we proceed as follows : * the expression for @xmath137 can be obtained from the friedmann equation and the conservation law : @xmath140 and therefore can be written in terms of the scale factor , @xmath33 , in this modified gcg model as follows , @xmath141\frac{a}{a^{1+\beta}}\times&\\ \left(\frac{a}{a^{1+\beta}}+\frac{b}{a^{4(1+\alpha)}}\right)^{-\alpha/(1+\alpha)}\ , , & \mbox{early time}\\ \frac{\kappa^2}{6}a^2\left[\rho_{m0}\left(\frac{a_0}{a}\right)^3 + 4\rho_\lambda\right]\ , , & \mbox{late time } \end{array}\right.\ ] ] * at very early time , the model we are analyzing behaves like a pure power - law inflation . in that case
, the set of differential equations ( [ ode ] ) have an analytical solutions @xcite , which we will use as initial conditions for @xmath142 and @xmath143 in our numerical integration : @xmath144 .
\label{bcgw2}\end{aligned}\ ] ] before proceeding , let us highlight that the general solution of eq .
( [ ode ] ) for a power - law inflation includes the first and the second kind of hankel functions @xcite .
but here we adopt just hankel function of the first kind because the solution should reduce to the minkowski spacetime solution ( [ minkowski ] ) at very early time when @xmath97 @xcite . *
similarly to the scalar perturbations , it is simpler to solve the set of equations ( [ ode ] ) in terms of the scale factor . with this in mind and in order to apply the boundary conditions ( [ bcgw1 ] ) and ( [ bcgw2 ] )
, we use the relation between the conformal time and the scale factor stated in eq .
( [ atau ] ) . *
an important issue is the range of integration of eqs .
( [ ode ] ) . in principle
, the integration must be done from a very early time when the mode is well inside the horizon on the inflationary era to the present time .
however , there is a way of reducing the computing time without affecting in practice the results : for a given mode @xmath87 , we start integrating from 0.01@xmath145 until 100@xmath146 , where @xmath145 and @xmath146 correspond to the scale factor where the mode @xmath87 exits the horizon and reenters the horizon , respectively .
our numerical integration will be carried out for modes @xmath87 ranging from the maximum of the potential @xmath147 until the minimum one .
the last one corresponding roughly to the modes that are reentering currently the horizon .
the methods is schematically shown in fig .
[ integration ] . * as can be noticed from fig .
[ integration ] , there is an intersection point in the potential @xmath147 which takes place roughly at @xmath148 .
given that we have set the current scale factor to @xmath149 , the intersection point corresponds roughly to the time of big bang nucleosynthesis ( bbn ) . on our model , that point corresponds to the transition between the modified generalized chaplygin gas and the @xmath24cdm model . for modes that enters the horizon after the bbn , we will have to integrate eqs .
( [ ode ] ) twice as we will be dealing with two different potentials as shown in eq .
( [ potentialgw ] ) . for the second integration
, we use the final values obtained in the first integration as the initial values for the functions @xmath150 and @xmath151 .
* the integration will be performed assuming : @xmath152 , @xmath153 , @xmath149 and @xmath154 km / s / mpc .
in addition , the parameters of the model are fixed as explained in section ii . in this subsection
we present our results of the power spectrum of the gravitational waves for the model presented in section ii . as we explained before
, the bogoliubov coefficient , @xmath125 , gives the number of graviton created in each period , @xmath155 , and therefore we can calculate the time evolution of @xmath156 during the expansion of the universe in this gcg model . in fig .
[ betasqure ] we present an example of our results where we set @xmath54 .
the plot corresponds to six different wave numbers @xmath87 : @xmath157 , @xmath158 , @xmath159 , @xmath160 , @xmath161 , and @xmath162 , or different frequencies @xmath163 . .
the integration has been carried out for six different values of frequencies .
we can see that for each mode @xmath87 , the gravitons have been created mostly during the inflationary era .
for this plot we fixed @xmath54.,width=302 ] similarly we can obtain the spectrum of the gravitational waves as shown in fig .
[ gwspectrum ] .
this figure is quite enlightening .
it shows that for large frequencies the larger @xmath164 is , the larger is the dimensionless logarithmic energy spectrum of the gws . on the other hand ,
the larger @xmath164 is , the more the spectrum is shifted towards larger frequencies .
this is a simple consequence of the fact that a larger @xmath164 implies a larger maximum of the potential @xmath147 . in summary ,
an increase in @xmath164 implies two things : an upward shift and a rightward shift of the spectrum .
it is also worthy of notice that the plateau of the spectrum merges at middle / low frequencies for different values of @xmath14 .
this is not surprising as the energy scale of inflation , @xmath165 , is almost fixed and independent of @xmath14 in our model , where @xmath166 ( cf . the previous section ) and it is precisely the value of @xmath165 that shifts vertically such a plateau @xcite .
the parameter @xmath165 is fixed mainly by the measurements of @xmath114 and @xmath76 .
the fact that @xmath76 is much more constrained by the latest wmap7 data than the previous one , wmap5 , explains the changes in the plateau of our fig .
[ gwspectrum ] and fig . 4 of ref .
@xcite , where on the last mentioned figure different values of @xmath76 , or equivalently different values of the power law universal expansion @xmath85 ( in terms of the conformal time ) are assumed .
therefore , the spectrum of gws at very low frequencies is insensitive to the value of @xmath14 , because for larger frequencies the spectrum already merged for different @xmath14 s whereas for smaller frequencies the universe is described by the concordance model @xmath24cdm ( at time the modes reenters the horizon ) , which is independent of the parameter @xmath14 . against the frequency @xmath129 for different value of @xmath14 in this gcg model :
the blue line refers to @xmath54 , the red one refers to @xmath167 , and the grey one to @xmath168.,width=340 ] there are some observational constraints on the upper limit of the energy spectrum @xmath169 @xcite : * constraint from cmb : + @xmath170 .
* constraint from timing observations of millisecond pulsars : + @xmath171 . * constraint from doppler tracking of the cassini spacecraft : + @xmath172 .
* constraint from ligo : + @xmath173 , + for a few hundred @xmath174 .
* constraint from bbn : + @xmath175 , + where @xmath176 , where @xmath177 .
all these constraints are fulfilled by the power spectrum shown in fig .
[ gwspectrum ] .
this model is a clear example of how the detection of gws can be extremely helpful to distinguish among several inflation models and a complementary tool to the study of primordial spectrum of the scalar perturbations . indeed , while our model is indistinguishable for any value of @xmath14 at low @xmath87 and @xmath129 , on the spectrum of scalar and tensorial perturbations , respectively , it should be distinguishable at high @xmath87 and @xmath129 .
most importantly , the high frequency regime is within the reach of future gravitational wave detectors such as bbo and decigo @xcite .
we present a model that attempts to fuse the inflationary era and the subsequent radiation dominated era under a unified framework so as to provide a smooth transition between the two .
the model is based on a modification of the generalized chaplygin gas . more precisely , it interpolates between a `` quintessence - like '' , or power law , expansion and a subsequent radiation dominated universe .
such a gas fulfills an equation of state ( [ p ] ) or ( [ p2 ] ) which has been previously used in the literature as a means to explain a possible interplay between the dark sectors of the universe @xcite .
our attempt is therefore in a different context .
we have shown how such a model has an underlying scalar field description , where the scalar field starts rolling down a potential hill where the extreme slow - roll approximation is valid until a bite before the end of inflation .
we have obtained the full power spectrum of the scalar perturbations and constrained the model using the latest wmap7 data . from our analysis
it turns out that the model is indistinguishable from a power - law expansion at low wave numbers @xmath87 ( cf .
[ ps ] and [ ns ] ) even if we vary the only degree of freedom present on the model , the parameter @xmath14 .
notice in this regard that the rest of the parameters of the model are fixed by observations as explained on sect .
we have completed our analysis by obtaining the spectrum of the gravitational waves ( cf . fig .
[ gwspectrum ] ) where we have used the method of bogoliubov coefficients @xcite . it turns out that our model is within the reach of future gravitational - wave detectors like bbo and decigo @xcite .
most importantly , for those frequencies it appears possible to distinguish this modified gcg model among different values of @xmath14 , while this seems unattainable using the power spectrum of the scalar perturbation with the present or near future observations .
last but not least , we have shown as well how the scalar perturbation spectrum of the present model is not as red as the one presented in @xcite for an alternative gcg model for the early universe . in comparison ,
the model presented here is more consistent with observations .
the authors are grateful to alfredo b. henriques for a careful reading of the manuscript .
is supported by the portuguese agency fundao para a cincia e tecnologia through the fellowship sfrh / bpd/26542/2006 .
she also wishes to acknowledge the hospitality of lecospa center at the national taiwan university during the completion of part of this work .
p.c . and y.w.l .
are supported by taiwan national science council under project no .
nsc 97 - 2112-m-002 - 026-my3 and by taiwan s national center for theoretical sciences ( ncts ) .
is in addition supported by us department of energy under contract no .
de - ac03 - 76sf00515 .
a. h. guth , phys .
d * 23 * , 347 ( 1981 ) ; a. a. starobinsky , phys . lett .
b * 91 * , 99 ( 1980 ) ; a. j. albrecht and p. j. steinhardt , phys .
lett . * 48 * , 1220 ( 1982 ) ; a. d. linde , phys .
b * 108 * , 389 ( 1982 ) .
d. langlois , lect .
notes phys .
* 800 * , 1 ( 2010 ) [ arxiv:1001.5259 [ astro-ph.co ] ]
. b. a. bassett , s. tsujikawa and d. wands , rev .
phys . *
78 * , 537 ( 2006 ) [ arxiv : astro - ph/0507632 ] .
j. e. lidsey , a. r. liddle , e. w. kolb , e. j. copeland , t. barreiro and m. abney , rev .
phys . * 69 * , 373 ( 1997 ) [ arxiv : astro - ph/9508078 ] . a. y. kamenshchik , u. moschella and v. pasquier , phys .
b * 511 * , 265 ( 2001 ) [ arxiv : gr - qc/0103004 ] ; n. bili , g. b. tupper and r. d. viollier , phys .
b * 535 * , 17 ( 2002 ) [ arxiv : astro - ph/0111325 ] ; m. c. bento , o. bertolami and a. a. sen , phys .
d * 66 * , 043507 ( 2002 ) [ arxiv : gr - qc/0202064 ] ; p. m. s and a. b. henriques , phys .
d * 77 * , 064002 ( 2008 ) [ arxiv:0712.2697 [ astro - ph ] ] .
p. m. s and a. b. henriques , gen .
grav . *
41 * , 2345 ( 2009 ) [ arxiv:0804.3278 [ astro - ph ] ] .
m. abramowitz and i stegun , _ handbook of mathematical functions _
( dover , 1980 ) . | we propose a phenomenological model for the early universe where there is a smooth transition between an early `` quintessence '' phase and a radiation dominated era .
the matter content is modelled by an appropriately modified chaplygin gas for the early universe .
we constrain the model observationally by mapping the primordial power spectrum of the scalar perturbations to the latest data of wmap7 .
we compute as well the spectrum of the primordial gravitational waves as would be measured today .
we show that the high frequencies region of the spectrum depends on the free parameter of the model and most importantly this region of the spectrum can be within the reach of future gravitational waves detectors . |
in material science , fractography is concerned with the description of fracture surfaces in solids and is routinely used to determine the cause of failure in engineering structures @xcite .
various types of crack growth ( fatigue , stress corrosion , cracking ... ) produce different characteristic features on the surface , which in turn can be used to identify the failure mode and direction .
fractography is one of the most used experimental techniques to recover some aspects of crack dynamics , and is thus a major tool to develop and evaluate theoretical models of crack growth behavior .
fractography in two dimensional and three dimensional materials is fundamentally different .
a broken surface in three dimensions is the trace of a front line singularity and unless a full dynamical measurement is available , it is not possible to reconstruct its propagation history @xcite .
an extreme example is the family of systems in which a crack front is confined to move along a weak plane @xcite and where the post - mortem surface is simply a flat surface that can not reveal any information on the actual advance of the crack front .
in contrast , a crack path in two dimensions is simply the trace left by the propagation of a crack tip and thus can be thought of as the world - line of a moving singularity .
this means that a post - mortem fractographic analysis can fully recover the crack propagation history .
this shows that the 2d problem provides a good framework with many advantages to decipher the crack tip dynamics , and is therefore our focus here . for slowly propagating cracks in homogeneous materials ,
fracture surfaces are smooth and a fundamental question that arises in that context is the stability of the propagating crack with respect to a prescribed path .
a stability analysis of two - dimensional cracks propagating in homogeneous materials based on linear elastic fracture mechanics ( lefm ) was performed by cotterell and rice @xcite , and yields the famous @xmath0-criterion .
this criterion states that if the quantity called the @xmath0-stress ( see eq .
( [ eq : sif ] ) below for a definition ) is positive the crack path becomes unstable , whereas if @xmath1 the path is stable .
experimentally and theoretically , the instability predicted by the @xmath0-criterion has been proven to be a necessary condition but not a sufficient one @xcite .
the stability analysis of cotterell and rice being incomplete already in homogeneous media can not be expected to describe correctly crack propagation in heterogeneous media where the path of the fracture is generally rough .
actually , for such materials , fracture surfaces are claimed to exhibit fractal ( or self - affine ) properties @xcite .
the self - affinity of a @xmath2-dimensional surface is fully characterized by @xmath3 exponents @xcite . since the dynamics of cracks in heterogeneous media is a rich field encompassing a wide range of physical phenomena , it is important to distinguish between three different exponents : the one describing roughness in the direction perpendicular to the crack propagation @xcite , the second one describing the roughness in the direction of the propagation ( the so called `` out - of - the - plane '' roughness , which is the subject of this paper ) and the third one describing the in - plane roughness of the crack front during its propagation through the material @xcite . in some cases
these exponents are related @xcite but generically they are independent
. there are some experimental measurements of the roughness exponent @xmath4 of two - dimensional or quasi two - dimensional cracks ( in the appendix some methods of measurements of the `` out - of - the - plane '' roughness exponent are described ) . for berea sandstone ( @xmath5 @xcite ) ; for concrete ( @xmath6 @xcite ) ; for paper ( @xmath7 @xcite , @xmath8 @xcite and @xmath9 @xcite ) ; for wood ( @xmath10 @xcite ) and for a system of drinking straws ( @xmath11 @xcite ) .
as can be seen , all the values vary between @xmath12 and @xmath13 and thus suggest non - universal behaviour .
in particular note that the measured exponents are all larger than @xmath14 , a value that corresponds to the roughness exponent of an uncorrelated random walk @xcite . in this work
we aim at a thorough study of crack propagation in 2d disordered materials .
a natural question is whether the roughness of the broken surface is related to an instability mechanism of the crack tip propagation . in the following
we will derive an equation of motion for a crack propagating in disordered medium that would allow us to study both its stability and roughness properties .
the paper is organized as follows : we start by recalling the stability analysis la cotterell and rice
. then we present the formulation of a stochastic model that takes into account the material heterogeneity and uses results regarding kinked cracks , and extend the @xmath0-criterion to heterogeneous materials .
we then specialize to the case @xmath15 and discuss the roughness of the resulting crack surfaces .
thanks to an exact result of the model in that limit @xcite , we are able to obtain analytically the form of the power spectrum of the paths , and offer an alternative interpretation of experimental results .
we conclude by discussing the implication of our result on the methodology of self - similarity analysis by suggesting a new measurement bias that has not been considered previously @xcite .
the key ingredient that allows a general discussion of cracks in a brittle material is the fact that the static stress field in the vicinity of the crack tip has the following universal expansion @xmath16 where @xmath17 are polar coordinates with @xmath18 located at the crack tip , and @xmath19 are known functions describing the angular variations of the stress field components @xcite . in this expansion ,
@xmath20 ( @xmath21 ) and @xmath0 are the stress intensity factors ( sifs ) and the nonsingular @xmath0-stress respectively .
this singular behaviour of the stress field justifies the expectation that the crack - tip dynamics could be formulated in terms of the sifs and the @xmath0-stress alone . in a 2d material ,
well - established criteria for quasi - static crack propagation are the griffith energy criterion @xcite and the principle of local symmetry ( pls ) @xcite .
this is expressed by the following equations of motion @xmath22 \ ; , \label{eq : grif}\\ k_2 & = & 0 \qquad \qquad \qquad \qquad \qquad \qquad [ { \rm pls } ] \ ; , \label{eq : pls}\end{aligned}\ ] ] where @xmath23 is the lam shear coefficient and @xmath24 is the fracture energy .
( [ eq : grif ] ) states that in order to induce crack propagation , the energy release rate @xmath25 must be large enough to create new crack surfaces .
( [ eq : pls ] ) imposes the symmetry of the stress field in the vicinity of the crack tip , such that it is locally under a pure opening mode .
therefore , the crack path is mainly selected by the pls while eq .
( [ eq : grif ] ) controls the intensity of the loading necessary to allow propagation .
other propagation criteria have been proposed in the literature @xcite , notably the maximum energy release rate criterion @xcite which states that the crack extends in the direction that maximizes the rate of energy release .
however , the pls has been shown to be the only self - consistent one @xcite .
a linear stability analysis of quasi - static two - dimensional crack propagation in homogeneous materials based on these concepts has been performed by cotterell and rice @xcite , which gave rise to the @xmath0-criterion .
this criterion states that for @xmath26 , a tensile crack propagation becomes unstable with respect to small perturbations around the straight path .
otherwise the straight crack propagation is stable .
experimentally , the @xmath0-criterion is known to hold , at least for cases with @xmath27 , that is when the prediction is that cracks become unstable @xcite . however
, even when @xmath1 the crack path can become unstable in some situations ( for example , the thermal crack problem in which the crack path exhibits an oscillatory instability @xcite ) .
in addition to the @xmath0-criterion , cotterell and rice predicted that for a semi - infinite straight crack experiencing a sudden local shear perturbation the subsequent crack path @xmath28 scales as @xmath29 in the stable regime , while @xmath30 in the unstable case .
however , the result @xmath31 is only marginally stable ( i.e. @xmath32 for large @xmath33 s ) @xcite , thus reflecting a limited aspect of stability .
this situation calls for a revision of the stability properties of slow cracks , especially in heterogeneous materials .
based on these observations we propose an equation describing the propagation of a crack in a disordered medium that allows to predict its path and study its stability . our model is based on a description where all the relevant information is encoded in the sif s @xmath34 and in the @xmath0-stress .
the crack propagation criteria used are the griffith energy balance ( [ eq : grif ] ) and the principle of local symmetry ( [ eq : pls ] ) . the physical picture of a propagating crack in a disordered material in the current formulation
is summarized in fig .
[ fig : curved ] .
it assumes that the crack tip propagates smoothly until it encounters a heterogeneity that changes locally the fracture energy and induces a local shear perturbation . as a result
, the crack forms a kink at a prescribed angle depending on the local perturbation induced by the heterogeneity . in order to calculate this angle
it is necessary first to introduce some results regarding kinked cracks
. corresponds to location @xmath35 of the @xmath36 heterogeneity.,width=302 ] consider an elastic body containing a straight crack with a kinked curved extension of length @xmath37 and kink angle @xmath38 . using standard assumptions related on the scaling of the stress field in the vicinity of the crack tip
it is shown that the shape of the local crack extension should be given by @xcite @xmath39 where @xmath40 is the slope of the kink and the curvature parameter @xmath41 quantifies the curved extension of the kink .
moreover , it is shown that the static sifs at the crack tip after kinking @xmath42 @xmath43 are related to the sifs before kinking @xmath20 and to the @xmath0-stress via @xcite @xmath44\sqrt{s}+o(s)\ ; , \label{eq : exp}\ ] ] where @xmath45 , @xmath46 and @xmath47 are universal functions in the sense that they do not depend on the geometry of the body nor on the applied loading .
they depend only on the kink angle @xmath38 and were computed in @xcite .
note that eq .
( [ eq : exp ] ) shows that in general @xmath48 unless a special symmetry sets it to zero .
therefore , applying the principle of local symmetry means that the expansion ( [ eq : exp ] ) should vanish order by order in @xmath37 . in the presence of a small shear loading ( @xmath49 ) the extension of the initial straight crack
must therefore satisfy @xmath50 where the expansions of the functions @xmath45 , @xmath51 and @xmath47 for small angles have been used @xcite .
( [ eq : kink0 ] ) fixes the kink angle that develops due to the presence of shear perturbations , while eq .
( [ eq : kink1 ] ) determines the subsequent curvature of the crack path . in order to use eq .
( [ eq : kink0 ] ) , one needs to know the sifs @xmath52 just before kinking for an arbitrary broken surface @xmath53 ( with @xmath54 ) . for pure opening loading conditions and using a perturbation analysis around a straight crack ( i.e. a crack parallel to the @xmath33-axis whose
tip coincides with the curved crack located at @xmath55 ) , it can be shown that to first order in @xmath56 one can write @xmath57 and @xmath58 as functionals of @xmath28 through @xcite @xmath59 where @xmath60 and @xmath61 are the stress intensity factors and the @xmath0-stress component of a straight crack located at @xmath55 .
also , under pure tensile loading one readily has @xcite @xmath62 note that @xmath63 is proportional to @xmath64 , as expected , but also depends on the geometry of the problem via @xmath28 .
the superscript @xmath65 refers to quantities corresponding to the configuration of a centered straight crack ( i.e. one that is located at @xmath66 ) .
in addition , we will assume , as in @xcite , that @xmath67 is independent of @xmath33 , arguing that its variation in space does not modify qualitatively the results .
however , a variation in the stress intensity factor @xmath68 should induce a variation in the @xmath0-stress . indeed , lefm insures that for the same conditions under which eq .
( [ eq : k2 ] ) is valid , one has @xmath69 .
notice that eq .
( [ eq : k2star ] ) reveals an additional source of bias in the stability analysis of cotterell and rice , since the linear perturbation performed in @xcite ignores the term proportional to @xmath28 in the expression of @xmath70 . at this point
we introduce the heterogeneities in our model .
the source of heterogeneities can be either variations of the elastic moduli in the material , or from residual stresses that were introduced for example by welding @xcite , or during the machining of the material .
since the stress field is tensorial , these heterogeneities should affect both @xmath71 and @xmath70 independently .
the local fluctuations in the toughness denoted by @xmath72 have a finite mean @xmath73 .
however , the local shear fluctuations , denoted by @xmath74 must have a vanishing average because of pls .
assuming that the crack advance between heterogeneities obeys the principle of local symmetry and using eqs .
( [ eq : kink0],[eq : k2],[eq : k2star ] ) and the discussion above , one concludes that the local kinking angle is determined by @xmath75 + o\left ( \delta k_\ell^2 , \delta k_\ell^2 \right ) \label{eq : dtheta},\end{aligned}\ ] ] where @xmath76 is a length - scale that depends on the geometry of the configuration .
for example , it is proportional to the width of the strip in the case of a finite strip geometry - a configuration that is often adopted in experiments .
the stage is set now to write the equation governing the crack path evolution .
since we are dealing with linear perturbations , we will also neglect the curvature parameter @xmath41 introduced in eq .
( [ eq : shape ] ) , and assume that the crack extension after kinking is always straight .
this assumption is justified when @xmath15 or when the distances between successive kinking events are small ( which is equivalent to high density of heterogeneities ) .
it leads to the configuration depicted in fig .
[ fig : model ] , from which one can easily read the equation @xmath77 . in the limit of small equal intervals
@xmath78 between successive heterogeneities , one has @xmath79 , then @xmath80 and eq . ( [ eq : dtheta ] ) leads to @xmath81 where the indexes @xmath82 have been replaced by the position @xmath33 through the passage to the continuum limit .
( [ eq : motion ] ) reveals two noise terms that can be redefined by @xmath83 also , let us use the geometrical scale of the configuration @xmath76 as a unit length and define the constant @xmath84 then , eq .
( [ eq : motion ] ) becomes the following dimensionless stochastic equation @xmath85 note that by choosing the length scale @xmath76 , the total extension of the crack is not given .
also , @xmath86 can be either positive or negative depending on the sign of the @xmath0-stress .
the discontinuous nature of crack propagation in a disordered material imposes a detailed discrete microscopic description of the influence of heterogeneities .
the resulting stochastic integro - differential equation of the crack path should be derived as the continuum limit of the discrete model .
this approach is different from previous pure continuum modeling @xcite that implicitly assumes smoothness of the paths and _ one _ source of noise that is introduced _ a posteriori_. in opposite , eq .
( [ eq : start ] ) shows that our approach leads to derivatives of _ two _ noise terms , one of which is multiplicative and the other is additive , without imposing them _ a priori_. the properties of the noise terms are prescribed by the original distribution of heterogeneities in the material that may exhibit long - range correlations as well as anisotropy . although such features may be important @xcite and in order to remain general , we assume short range correlations and thus model the noise terms @xmath87 as independent gaussian white noises @xmath88 note that @xmath89 and @xmath90 that enter eq .
( [ eq : start ] ) are conserved random terms ( i.e. , derivatives of white noises ) modeling the fluctuations in the local toughness and the local shear respectively . in the following
, we will show that the simple scenario of uncorrelated disorder already offers a rich spectrum of results . including additional features in the disorder , such as long - range power law correlations ,
could lead to richer phenomena @xcite and is left as a possible extension to the present analysis . to be consistent with the derivation of the model one needs both the noise amplitudes @xmath91 and @xmath92 to be small .
however , from eq .
( [ eq : start ] ) one can see that varying the amplitude @xmath92 is equivalent to multiplying @xmath28 by a constant , i.e. to fixing the overall scale of the height fluctuations , which does not influence the roughness of the curve .
since the scaling properties are not affected , the value of @xmath92 will not be reported in the following and @xmath91 will be the only pertinent noise parameter . as a result @xmath28
will be presented in arbitrary units .
regarding the @xmath0-stress , one expects @xmath93 to be of order @xmath94 in the framework of lefm .
as a first observation , if the material is homogeneous or weakly disordered , one has @xmath95 and the solution of eq .
( [ eq : start ] ) is simply @xmath96 .
the addition of suitable initial conditions allows recovering the zero order solution @xmath97 corresponding to a centered straight crack path . eq . ( [ eq : start ] ) should be understood as resulting from a perturbation analysis of the crack trajectory around the solution in the absence of heterogeneities that is selected by the pls .
this should be contrasted with the stability analysis of a straight crack in a homogeneous material with respect to other solutions that satisfy also the pls .
an example of such a situation is the thermal crack problem @xcite where oscillatory crack paths exist in addition to the centered straight one and become more stable than the straight configuration at a given well defined threshold .
the crack propagation there is always smooth and is very different from the stability encountered in disordered materials , which is due to the large density of heterogeneities that induces discontinuous propagation via linear segments between the heterogeneities . in order to study crack paths that result from eq .
( [ eq : start ] ) , we start with a numerical integration of it .
the initial condition will be always chosen to be a straight semi - infinite crack , @xmath97 for @xmath98 . in a discretized form , eq .
( [ eq : start ] ) becomes@xmath99 for @xmath100 , using @xmath101 as the uniform distance between heterogeneities and @xmath102 as initial conditions . eq .
( [ eq : start - dis ] ) is a discretized version of eq .
( [ eq : start ] ) that corresponds to the it prescription @xcite and was chosen by the discrete manner by which eq .
( [ eq : start ] ) was derived .
also , the quantities @xmath103 are taken as independent random numbers , equally - distributed in the segment @xmath104 $ ] , and thus with variance @xmath105 .
the averages of @xmath103 are not important since only derivatives of the noise terms appear in the model . in order to be consistent with length normalisation
, we pick a @xmath106 lattice of size @xmath107 with @xmath108 sites , @xmath108 being also the number of heterogeneities in the interval @xmath109 $ ] .
thus @xmath110 corresponds to the density of heterogeneities and a small @xmath101 probes the regime of highly disordered materials .
[ fig : zoom ] shows an example of a crack grown using eq .
( [ eq : start - dis ] ) .
the inset shows a zoom into a small part of the path , which may be suggestive of self - similar properties to the naked eye .
however , before a thorough study of this aspect , a stability analysis la cotterell and rice should be performed . , @xmath111 and @xmath112 .
[ fig : zoom],width=377 ] as mentioned above , the classical @xmath0-criterion of cotterell and rice @xcite states that straight tensile crack propagation in homogeneous materials become unstable when @xmath27 . we therefore simulated crack paths with various values of @xmath0 in order to test this criterion within our model .
we first consider the case with only one shear perturbation at @xmath113 , after which the local toughness is @xmath114 and only @xmath115 is allowed to fluctuate .
the results are presented in fig .
[ fig : t - crit]a .
essentially , we recover the @xmath0-criterion , namely an instability occurs for @xmath116 ( or equivalently @xmath27 ) .
it turns out that by adding the shear perturbations ( i.e. , @xmath117 ) , the same scenario is recovered ( see fig .
[ fig : t - crit]b ) .
one noticeable difference is that for positive values of @xmath0 , the divergence of the path seems to accelerate due to the presence of the shear perturbations .
still , paths that do not experience shear perturbations ( @xmath1 and @xmath118 ) will not destabilize in their presence .
it should be mentioned that within the approach of cotterell & rice it is not possible to follow more than one kink , as would certainly be the case in a disordered material where cracks propagate via many consecutive kinking events . in summary ,
our results confirm the @xmath0-criterion for homogeneous materials and extends it to disordered systems .
it is shown that straight crack propagation is unstable for @xmath27 and stable elsewhere .
moreover , in the stable case @xmath119 the marginal stability has been cured by the suppressing the square root behaviour @xmath120 predicted in @xcite .
-stress and @xmath92 with @xmath121 , @xmath122 .
( a ) @xmath118 , i.e. @xmath123 ) and ( b ) @xmath124 .
note that for the cases @xmath125 , the range of the @xmath56-axis is wider emphasising the exponential increase of the amplitude of the oscillations in those cases .
[ fig : t - crit],title="fig:",width=340 ] -stress and @xmath92 with @xmath121 , @xmath122 .
( a ) @xmath118 , i.e. @xmath123 ) and ( b ) @xmath124 . note that for the cases @xmath125 , the range of the @xmath56-axis is wider emphasising the exponential increase of the amplitude of the oscillations in those cases .
[ fig : t - crit],title="fig:",width=340 ]
in view of these results , from now on we will restrict our study to stable paths - as we are interested in crack roughness .
we first focus on the case @xmath15 , since it is simple enough to allow for definite numerical and analytical results , and at the same time contains the necessary complexity .
this claim is based on scaling arguments - power counting @xcite which is also supported by a numerical study . to put it simpler , as long as the crack path is stable , the presence of the @xmath0-stress does not change dramatically the shape of the crack paths . at the end of the paper
, we will come back to this point and show how a non zero @xmath0-stress influences the result . for the case of vanishing @xmath0-stress ( @xmath112 in eq .
( [ eq : start ] ) ) , exact results have been obtained previously @xcite . here , these results will be summarised briefly and extended .
one technical difference is that in this work lengths are scaled using the width @xmath76 and not using the length @xmath107 ; as a result the dimensionless parameter @xmath126 defined in @xcite is set to @xmath94 here .
essentially , our analysis is divided into two parts .
the idea is to identify first the possible responses of the crack path to one shear perturbation , and only then to generalize to a superposition of many shear perturbations .
qualitatively , in reaction to a single shear perturbation , a straight crack deviates from its former direction by an angle which is proportional to the strength of the perturbation and then starts to relax to its original form .
interestingly , it is found that the crack path can relax in two different ways : either by decaying exponentially ( inset of fig .
[ fig : eta1]a ) or by decaying exponentially while oscillating ( inset of fig .
[ fig : eta1]b ) . this behaviour can be understood from the study of the logarithmic derivative of the crack path , @xmath127 , which becomes stationary during the relaxation .
a fokker - planck equation is derived for @xmath128 , along similar lines to those described in @xcite , and the effect of the toughness fluctuations , @xmath89 , on average , can be reproduced by an effective deterministic evolution .
this result allows to derive an effective ( coarse grained ) simple equation of motion in the presence of a single shear perturbation ( and @xmath112 ) , namely @xcite @xmath129 \label{eq : h - c } \ , \ ] ] where @xmath130 . comparing this with eq .
( [ eq : start - dis ] ) one concludes that the averaged equation ( [ eq : h - c ] ) is obtained from the full one ( in a non - trivial way explained in @xcite ) by simply replacing the noise term @xmath89 with a constant , @xmath131 , that is always negative ( even though @xmath89 is equally positive and negative ) , and proportional to its variance .
this nontrivial result shows that the effect of the local toughness fluctuations is not so dramatic on the shape of the crack , apart from its constant variance and of course apart from setting a relevant scale for the energy that has to be invested in making the crack grow .
however , one should be careful with the interpretation of eq .
( [ eq : h - c ] ) , as it describes only mean quantities , which does not imply that each realization behaves exactly the same .
now , one can easily solve eq .
( [ eq : h - c ] ) with the initial conditions @xmath132 ( i.e. one shear perturbation at the origin only ) . for @xmath133
we get @xmath134 this result shows why two kinds of responses to an initial perturbation are possible . the solutions of eq .
( [ eq : h - c ] ) can exhibit either an exponential decay or damped oscillations depending on the sign of @xmath135 .
when @xmath136 , @xmath28 simply decays exponentially , while for @xmath137 , the hyperbolic sine becomes an oscillating function , and thus we find an oscillatory relaxation . since traditionally , noisy data are analysed in fourier space by looking for example at the power spectrum , it would be interesting to obtain an analytical expression for it as well @xmath138 where @xmath139 is the fourier component of @xmath28 . in figs . [
fig : eta1]a-[fig : eta1]b below we compare the result of the averaged power spectrum over @xmath140 simulated paths ( all with the same parameters but different realizations of the noise ) for the two cases @xmath141 ( damped oscillations ) and @xmath142 ( i.e. , simple exponential decay ) . as can be seen
the theoretical curve is in very good agreement with the numerical result over many decades . and @xmath118 .
the insets show examples of such paths , while the main figures show the power - spectra @xmath143 averaged over @xmath140 realizations of the noise @xmath115 .
dashed curves are the corresponding theoretical curves . ( a )
the case @xmath141 : results produced using @xmath144 , @xmath145 with initial conditions @xmath146 .
( b ) the case @xmath142 : results produced using @xmath144 , @xmath147 , with initial conditions @xmath146 .
[ fig : eta1],title="fig:",width=302 ] and @xmath118 .
the insets show examples of such paths , while the main figures show the power - spectra @xmath143 averaged over @xmath140 realizations of the noise @xmath115 .
dashed curves are the corresponding theoretical curves .
( a ) the case @xmath141 : results produced using @xmath144 , @xmath145 with initial conditions @xmath146 .
( b ) the case @xmath142 : results produced using @xmath144 , @xmath147 , with initial conditions @xmath146 .
[ fig : eta1],title="fig:",width=302 ] a natural step forward is to study crack propagation in a regime where there are many shear perturbations .
this of course amounts to retaining the additive noise @xmath90 in eq .
( [ eq : start ] ) . unlike the fluctuations in the local toughness ,
the shear perturbations can not be modeled by a constant .
this term seems crucial for creating the random patterns that are observed for fracture surfaces in nature .
interestingly , varying the various parameters results in rather different patterns , as shown in fig .
[ fig:0.6]a-[fig:0.8]a .
moreover , when analyzed using the real space methods , such as the min - max or the rms method ( see appendix ) , one can produce various values of roughness exponent @xmath4 which depend on the parameters of the model .
[ fig:0.6]b-[fig:0.8]b show the analysis the two crack paths and show that one can obtain , for example , values of roughness exponent @xmath7 and @xmath148 that can be found in literature @xcite . and @xmath149 .
( b ) results of a min - max and rms analysis that yields @xmath150 over more than @xmath151 decades .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.6],width=302 ] and @xmath149 .
( b ) results of a min - max and rms analysis that yields @xmath150 over more than @xmath151 decades .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.6],title="fig:",width=302 ] and @xmath149 .
( b ) results of a min - max and rms analysis that yields @xmath150 over more than @xmath151 decades .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.6],title="fig:",width=302 ] and @xmath152 .
( b ) results of a min - max and rms analysis that yields @xmath153 over a decade and a half .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.8],width=302 ] and @xmath152 .
( b ) results of a min - max and rms analysis that yields @xmath153 over a decade and a half .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.8],title="fig:",width=302 ] and @xmath152 .
( b ) results of a min - max and rms analysis that yields @xmath153 over a decade and a half .
( c ) the power spectrum of the crack path and a theoretical prediction for it based on the parameters of the model averaged over @xmath140 realizations .
[ fig:0.8],title="fig:",width=302 ] however , when looking at the power - spectrum of each crack path there is no longer a simple scaling picture , as in the min - max or rms plots . for both cracks , the power spectrum always begins with a plateau for small values of @xmath154 , but for larger values of @xmath154 there is nt an easy way to determine a slope , which seems to vary over different values .
this phenomenon is traditionally interpreted as a crossover between different regimes characterized by different roughness exponents in the analysis of experimental data @xcite . in view of the analytical results which we obtained in the previous section we argue for a different and simpler scenario .
the starting point is eq .
( [ eq : start ] ) with @xmath112 , supplemented with the initial condition @xmath155 and @xmath156 . in this equation @xmath90 is just a nonhomogeneous term .
this means that once we have a solution for the homogeneous equation , we can build a special solution that solves the nonhomogeneous part . recalling that @xmath157 and @xmath158 are independent random variables
, we conclude that , as before , averaging over realizations of the local toughness fluctuations amounts to replacing @xmath89 by @xmath131 .
this yields @xmath129 + \eta_2'(x ) \label{eq : h2-c } \ .\ ] ] using the initial conditions @xmath159 , the solution of this equation is given by @xmath160 where @xmath161 is the solution to the homogeneous problem given by @xmath162 the power spectrum of this solution yields the expression @xcite @xmath163 this expression is different from the expression given by eq .
( [ eq : h - cps ] ) as there is a @xmath164 term in numerator , which implies a tail of @xmath165 in the spectrum .
the coefficients in eq .
( [ eq : h2-cps ] ) are determined from the fourier transform of a stationary signal after cutting out the transient regime , that is @xmath166 .
this leads to a system with effective random initial conditions at @xmath167 . since
@xmath168 is chosen in the steady regime the statistics of @xmath169 and @xmath170 are known , and an analytical expression of the power spectrum can be obtained . figs . [ fig:0.6]c-[fig:0.8]c shows that despite eq .
( [ eq : h2-cps ] ) agrees very well with the simulations data for the power spectrum , one might be tempted to fit it with a power law ansatz . however , apart from a tail @xmath171 at large @xmath154 s , which yields a roughness exponent @xmath172 at small length scales , eq .
( [ eq : h2-cps ] ) tells us that there is no self - affine behaviour of fracture surfaces at intermediate length scales , a simple crossover is taking place .
the result contained in eq .
( [ eq : h2-sol ] ) allows us to derive the full probability distribution function of @xmath173 as defined in eq .
( [ eq : delh ] ) , which is becoming a popular measure for self - affinity @xcite . using eq .
( [ eq : h2-sol ] ) and some simple manipulations one can rewrite @xmath173 as @xmath174dx } \label{eq : delh2 } \ , .\ ] ] here , the left most point has been pushed to @xmath175 in order to ensure stationarity .
then , the required pdf is formally given by @xmath176dx } } \right ) } \right\rangle_{\eta_2 } \label{eq : pdf1 } \ , .\ ] ] using the fourier representation of the delta distribution @xmath177 , eq .
( [ eq : pdf1 ] ) becomes @xmath178dx } } } \right\rangle _ { \eta _ 2 } dq } \label{eq : pdf2 } \ , .\ ] ] since the term in the exponent is a linear combination of independent random terms one gets @xmath179dx } } } \right\rangle = e^{-\frac{1}{2}q^2 \left\langle { \left [ { \int _ { - \infty } ^0 { h'_0 \left ( { - x } \right)\left [ { \eta _ 2 \left ( { x + \delta x } \right ) - \eta _ 2 \left ( x \right ) } \right]dx } } \right]^2 } \right\rangle } \label{eq : pdf3 } \ , , \ ] ] namely a gaussian , and one just needs to calculate its variance @xmath180dx } } \right]^2 \right\rangle_{\eta_2 } \nonumber \\ & & = \frac{d_2}{c}\left[1 - \exp\left\ { - c\delta x { \textstyle{{4 - c\cosh \left [ { \frac{1}{2}\sqrt { c(c - 4 ) } \delta x } \right ] + \sqrt { c(c - 4 ) } \sinh \left [ { \frac{1}{2}\sqrt { c(c - 4 ) } \delta x } \right ] } \over { c - 4 } } } \right\ } \right ] \label{eq : pdf4 } \ , , \end{aligned}\ ] ] and finally @xmath181 ^ 2 } \label{eq : pdf5 } \ , .\ ] ] recall that for the underlying shape @xmath28 to be self affine the pdf must obey two properties ( see eq .
( [ scaling4 ] ) ) : first , it should have the same form for all the scales @xmath182 .
this property is explicitly obeyed by the derived pdf ( [ eq : pdf5 ] ) .
the second requirement is that the rms @xmath183 scales as @xmath184 .
this property is not obeyed here , since @xmath185 meaning that there is a slow crossover from a square - root behaviour , for small scales , to a constant , for large scales , and strictly speaking the path @xmath28 is not self - affine .
sampled from a crack path simulated using @xmath186 and @xmath149 over @xmath187 time points on a semi - log scale . for each @xmath182 a pure gaussian
is plotted as a guide to the eye .
note that the various distributions are shifted logarithmically horizontally for visual clarity .
( b ) a comparison of the moments @xmath188 to those expected for a gaussian distribution @xmath189 for @xmath190 . the ratio @xmath191 is presented and the dashed lines mark a @xmath192 deviation interval .
[ fig:0.6pdf],title="fig:",width=302 ] sampled from a crack path simulated using @xmath186 and @xmath149 over @xmath187 time points on a semi - log scale . for each @xmath182 a pure gaussian
is plotted as a guide to the eye .
note that the various distributions are shifted logarithmically horizontally for visual clarity .
( b ) a comparison of the moments @xmath188 to those expected for a gaussian distribution @xmath189 for @xmath190 . the ratio @xmath191 is presented and the dashed lines mark a @xmath192 deviation interval .
[ fig:0.6pdf],title="fig:",width=302 ] in order to check these theoretical predictions , a propagating crack has been simulated over a very long interval , using the parameters presented in fig .
[ fig:0.6 ] , in order to produce the pdf for various values of @xmath182 .
as can be seen , the height distribution seems to exhibit gaussian statistics as predicted . in order to verify this more quantitatively
, we compare the moments of @xmath193 normalized by the @xmath194 moment , namely @xmath195 to those obtained by the gaussian distribution , namely @xmath196 up to @xmath197 order .
the results are presented in fig .
[ fig:0.6pdf]b , and supports the gaussianity of the distributions .
finally , we compare the width of distribution @xmath198 to the one given by eq .
( [ eq : pdf4 ] ) . as can be seen in fig .
[ fig:0.6pdfsigma ] , although the theoretical prediction captures the form and has the right order of magnitude , it clearly deviates from the result of the simulation .
in fact , it seems that the prefactor in eq .
( [ eq : pdf4 ] ) ( i.e. @xmath199 ) underestimates the measured one , such that by tuning this prefactor one can reproduce the right behaviour over the whole range .
a possible reason for this difference is due to discretization and finite - size scaling .
another reason could be the fact that we derive the pdf by first averaging over @xmath115 ( and thus obtaining the effective equation ( [ eq : h2-c ] ) ) , and only then averaging over @xmath200 , while in reality these two noisy terms fluctuate simultaneously at the same scale .
since the pdf is a sensitive probe this delicate issue is pronounced . at any rate ,
the simulated @xmath198 confirms the statement that there is a crossover from a square - root behaviour to a constant , and thus no real self affinity exists .
obtained from the simulation ( with the parameters defined in fig .
[ fig:0.6pdf ] ) to the one calculated in eq .
( [ eq : pdf4 ] ) .
[ fig:0.6pdfsigma],width=302 ] let us now study of the effect of the @xmath0-stress term in eq .
( [ eq : start ] ) .
still , we consider only the stable regime , that is when @xmath201 . in real experimental situations ,
one expects @xmath202 and most physical systems can be well described by the case @xmath112 .
indeed our analysis and numerical results show that eq .
( [ eq : start ] ) with @xmath202 exhibits scaling behaviour that is very close to the case @xmath112 . nevertheless , in order to demonstrate the impact of a non - zero @xmath0-stress
, a large value of @xmath86 is used .
[ fig : rought ] shows results of crack paths grown with @xmath203 and the corresponding power spectrum . , @xmath204 and @xmath203 .
( b ) rms and min - max curves , the fit is just a guide to the eye ( c ) the power spectrum of the crack path over @xmath205 realizations and a theoretical prediction based on the heuristic approximation given by eq .
( [ eq : h3-cps ] ) .
[ fig : rought],width=302 ] , @xmath204 and @xmath203 .
( b ) rms and min - max curves , the fit is just a guide to the eye ( c ) the power spectrum of the crack path over @xmath205 realizations and a theoretical prediction based on the heuristic approximation given by eq .
( [ eq : h3-cps ] ) .
[ fig : rought],title="fig:",width=302 ] , @xmath204 and @xmath203 .
( b ) rms and min - max curves , the fit is just a guide to the eye ( c ) the power spectrum of the crack path over @xmath205 realizations and a theoretical prediction based on the heuristic approximation given by eq .
( [ eq : h3-cps ] ) .
[ fig : rought],title="fig:",width=302 ] there are no analytical results for @xmath206 .
however , one expects a simplification in the spirit of the previous sections , i.e that averaging over @xmath115 would yield an effective langevin equation of the form @xmath207 + \eta_2'(x ) \label{eq : h3-c } \ , \ ] ] where @xmath208 and @xmath209 are renormalized deterministic prefactors , that would in general depend on @xmath210 and @xmath86 .
the spectrum of the solution averaged over @xmath200 would then be given by @xmath211 by comparing with the numerical results we find evidence indicating that one needs to take @xmath212 , where @xmath213 is defined similarly to the case @xmath112 . using this modified form ,
the agreement with the simulation is good ( see fig . [
fig : rought]c ) . note that in the expression ( [ eq : h3-c ] ) the term related to @xmath86 is not dominant for small @xmath154 s neither for large @xmath154 s which means that it does not modify the shape of the spectrum in a dramatic way .
this is consistent with the power - counting argument mentioned before .
the probability distribution functions for height increments can be easily computed numerically .
[ fig : pdft ] reports the same details as for the case @xmath15 .
[ fig : pdft]a-[fig : pdft]b show deviations from gaussianity at large scales meaning that the pdf of the height differences are not self - similar .
since the first requirement for self - affinity is violated , it is clear that an analysis of the scaling of the variance is biased , and the results it yields should be taken prudently .
[ fig : pdft]c reports such an analysis , i.e. @xmath198 , and interestingly shows that the deviations from gaussianity are sufficient to produce similar artifacts to those seen in the previous sections , namely a seemingly power law behaviour for small values of @xmath182 .
we verified that the power spectrum of the crack paths does not manifest the same bias , but rather reproduces a @xmath171 behaviour at the tail as expected from eq .
( [ eq : h3-cps ] ) ( implying @xmath172 ) . anyway , the fact that the crack paths becomes flat at the large scales is independent of the method used . , @xmath204 and @xmath203 on a semi - log scale . for each @xmath182
we plot in addition a pure gaussian as a guide to the eye .
note that we shifted the various distributions logarithmically for visual clarity .
( b ) a comparison of the moments @xmath188 to those expected for a gaussian distribution @xmath189 for @xmath190 , revealing some deviations from gaussianity .
the dashed lines marks a @xmath192 deviation interval .
( c ) the rms @xmath198 obtained from the simulation on a log - log scale .
the first part of the curve that seems to follow a power law is fitted and an exponent close to @xmath214 is obtained .
[ fig : pdft],title="fig:",width=302 ] , @xmath204 and @xmath203 on a semi - log scale . for each @xmath182
we plot in addition a pure gaussian as a guide to the eye .
note that we shifted the various distributions logarithmically for visual clarity .
( b ) a comparison of the moments @xmath188 to those expected for a gaussian distribution @xmath189 for @xmath190 , revealing some deviations from gaussianity .
the dashed lines marks a @xmath192 deviation interval .
( c ) the rms @xmath198 obtained from the simulation on a log - log scale .
the first part of the curve that seems to follow a power law is fitted and an exponent close to @xmath214 is obtained .
[ fig : pdft],title="fig:",width=302 ] , @xmath204 and @xmath203 on a semi - log scale . for each @xmath182
we plot in addition a pure gaussian as a guide to the eye .
note that we shifted the various distributions logarithmically for visual clarity .
( b ) a comparison of the moments @xmath188 to those expected for a gaussian distribution @xmath189 for @xmath190 , revealing some deviations from gaussianity .
the dashed lines marks a @xmath192 deviation interval .
( c ) the rms @xmath198 obtained from the simulation on a log - log scale .
the first part of the curve that seems to follow a power law is fitted and an exponent close to @xmath214 is obtained .
[ fig : pdft],width=302 ]
so far we discussed the stability of crack paths in heterogeneous media , and the possible shapes they can take .
it turned out that in the stable regime ( @xmath215 ) one can get many possible types of patterns depending on the parameters of the model , including random patterns that seem to resemble self similar shapes .
actually , the analysis whose results are summarized in figs .
[ fig:0.6]b-[fig:0.8]b supports this observation and suggests that not only can one produce self - similar shapes but also a large family of such shapes that seem to span a wide range of different roughness exponents . however , the analysis of the corresponding power spectra of these crack paths yields a different conclusion .
a power spectrum of the form ( [ eq : h2-cps ] ) or ( [ eq : h3-cps ] ) means that strictly speaking the shape is not self - affine . to be more precise
, it implies the existence of a self - affine structure on a small scale described by a roughness of @xmath216 ( deduced from the large @xmath154 behaviour of the spectrum , i.e. @xmath165 ) which is super - imposed on a decaying function or on damped oscillations at larger scales , and thus the spectra crosses over to a flat behaviour for small @xmath154 s ( possibly with a peak at some particular @xmath217 as in fig .
[ fig : eta1]b ) . in this context ,
an attempt to fit a straight line to the power spectrum for intermediate values of @xmath154 might yield a seemingly reasonable fit for one / two decades but is certainly unjustified .
so how does one settle the difference between the results that come form the real space and the fourier space approaches ?
the answer to this question is not restricted to crack surfaces and is related to a general discussion of reliability of self - affine measurements .
a starting point is the work of schmittbuhl _
@xcite that reviews various methods to extract the roughness exponent from measured or simulated profiles . in @xcite , the authors compare between various methods with respect to different artifacts that can appear in the data , such as misorientation or signal amplification .
interestingly , they came up with a useful sensitivity assessment of each method with respect to the biases .
however , they did not discuss the case when a self - affine structure is imposed on an oscillating background , or more generally on a bias which is not translation invariant .
for this case , we claim that the real - space methods ( such as min - max and rms ) are highly vulnerable , while the power - spectrum is , naturally , very robust .
an extreme example is given below in fig .
[ fig : sin ] , where a pure sinusoidal path , which is clearly not a self - affine profile , is analyzed using the min - max and the rms methods .
the resulting curves misleadingly reveal @xmath218 decades of self - affine behaviour ( actually , an arbitrary number of decades can be devised easily ) , with a roughness exponent @xmath219 . in contrast , a fourier analysis of this profile gives essentially a delta function localized at the wavelength of the oscillations .
this artifact has not been discussed in @xcite nor in other reviews of existing methods for measuring the roughness exponent such as @xcite .
decades with a roughness exponent @xmath220 .
[ fig : sin],width=302 ] back to our model , this extreme example clearly favors fourier based analysis over real - space approaches .
moreover , it can also explain why by varying the parameters of the model and by using a real - space analysis , we could obtain a whole range of @xmath4 s between @xmath14 ( which is the roughness exponent of a simple random walk ) and @xmath94 .
this possibility is of course excluded when looking at the power spectrum where we have clear predictions for the its form .
finally , let us discuss the approach of examining self - affinity through an analysis of the whole pdf of height differences @xmath221 . in principle , this method allows to check in a very precise way the self - affinity of crack paths and yields the roughness exponent . however , this method is not easy to implement since it needs a large amount of data that is typically much more than the amount of data that a usual experiment can yield . in our simulations
we could use relatively long cracks of @xmath187 pixels with a reasonable effort .
even with this large amount of raw data it was not easy to get rid of artifacts in the pdf such as over estimates of the roughness exponent and detection of deviations from gaussianity , while the power spectrum could perform better already with smaller samples .
this means that while in principle the pdf method is superior to other ones , when discussing real samples which are always of a limited precision and resolution , other methods such as the analysis of the power spectrum ( resulting only from a @xmath151-point statistics ) are usually better .
in this paper , we studied the stability and roughness of slow cracks in 2d disordered materials with respect to the disorder .
our approach relies on a solid ground in both mechanics and statistical physics . after proposing an equation of motion of a crack tip in a 2d disordered material ,
we first generalize the well known @xmath0-criterion predicted to disordered materials and then describe the roughening of crack paths . using this equation of motion
, we observe numerically various possible patterns , including oscillating , decaying and rough paths .
we analyze the rough cracks using commonly used techniques . by using real - space methods we are able to obtain a whole range of roughness exponents , while the power - spectrum of the paths does not support these findings .
thanks to an exact result @xcite we are able to predict the power spectrum analytically for @xmath15 ( eq .
( [ eq : h2-cps ] ) ) , and approximately for @xmath222 ( eq .
( [ eq : h3-cps ] ) ) .
these analytical results suggest that the shapes are not self - affine , but rather flat objects on the large scale and random - walk like objects ( with @xmath172 ) at small scales .
we conclude that in such situations real - space methods are very vulnerable ( as they mix the scales and can yield seemingly self - affinity with almost any @xmath4 between @xmath14 and @xmath94 ) , while the fourier - space approach is much more appropriate and thus preferred .
the last point is not only relevant to the analysis of cracks , but applies to analysis of self - affine shapes where an oscillating background is likely to exist .
we hope that this will contribute to the general discussion of reliability of self - affine measurements . from an experimental point of view , it could be interesting to use expressions like eqs .
( [ eq : h2-cps ] ) and ( [ eq : h3-cps ] ) to fit the experimental data of 2d rough cracks . an important prediction implied by these expressions
is that the scale at which the crossover to a flat shape occurs is roughly @xmath223 , that is a geometric mean of a geometric length - scale and a disorder length - scale ( which is the density of disorder @xmath224 times the amplitude of the toughness fluctuations @xmath91 ) .
this suggests that by varying the width of the strip in which the crack propagates and/or tuning the density and amplitude of the heterogeneities one can change the crossover scale .
another interesting prediction is that by varying the sign of the combination @xmath135 from positive to negative ( for example by reducing the density of heterogeneities ) one can switch between a pure exponential decay of the shear perturbations to an oscillatory one . in the presence of many shear perturbations ,
this property can be easily observed as a peak appearing in the power spectrum , as the transition to an oscillatory response occurs .
it could be the case that some generalization of this effect is responsible to the transition reported in @xcite from oscillatory to rough cracks .
it could be that an extension of the approach developed here , together with the recent results of @xcite to account for thermal effects , provide a proper explanation for this transition .
an open question , not developed in this work , is how does the material micro - structure affect the roughness of the cracks paths . here
, we considered only short - ranged disorder , modeled by @xmath225-correlated noise for both kinds of disorder @xmath91 and @xmath92 ( corresponding to toughness and shear fluctuations respectively ) .
it is clear that considering off - lattice heterogeneities and taking into account the different correlation properties of @xmath115 and @xmath200 would make the paths and their resulting power spectra more complex - a well known phenomenon in stochastic systems @xcite .
furthermore , long - range power law correlations that are known to exist in certain materials such as quasi - crystals @xcite , porous materials @xcite and others @xcite could yield long range correlations in the toughness / shear fluctuations , and thus lead to yet richer phenomena . actually , a recent discrete numerical model of fracture has shown that long - range correlations in the disorder and its anisotropy can lead to non - universal scaling exponents @xcite .
these results may change dramatically future approaches to problems of crack propagation in disordered materials .
finally , a fundamental open question is whether this work could help to understand the enigma of crack roughness in 3d samples .
it is hoped that the in - plane roughness , mentioned above , and studied experimentally @xcite , may provide an important starting point .
more precisely , modeling the simplified 3d problem @xcite may be combined with the out - of - the - plane fluctuations ( namely the current work ) to yield a full 3d theory .
various elements needed in that direction can be found in the literature @xcite but such a theory is still far from being formulated .
we thank b. derrida , a. boudaoud , k. j. mly and s. santucci for fruitful discussions .
for any random surface , one of the most studied quantities to characterize its geometry is the roughness exponent .
for one dimensional crack paths embedded in two dimensional materials one single exponent @xmath4 is needed .
this exponent is also known as the hurst exponent .
there are many methods to measure self - affine exponents @xcite and we will shortly review here four that are commonly used .
let us parameterize the path using the function @xmath28 with @xmath226 . using the variable bandwidth methods @xcite
, the roughness exponent can be related to the scaling of the width in the @xmath56-direction as a function of a window size in the @xmath33-direction .
more specifically one expects @xmath227 .
what is left is to specify a way to define the width @xmath228 .
a possible definition is given by the standard deviation ( or rms ) of the height profile @xmath229 ^ 2 dx } \sim \left(\delta x\right)^{\zeta } \label{scaling1}\ .\ ] ] we will refer to it as the rms ( root mean square ) method .
another variable bandwidth method is the min - max method defined by @xmath230 a more elaborated way to extract the roughness exponent , which allows to test the self - affinity at the same time @xcite , is based on examining directly the probability distribution function ( pdf ) of the discrete gradient @xmath231 rather than just looking at its second moment as in the rms method . in order to implement this method , one needs to plot the ( properly normalized ) pdf of @xmath228 for every @xmath182 .
if the shape is self - affine two conditions should be met .
first , the pdfs emanating from different values of @xmath182 , namely @xmath232 , should collapse on to a single curve .
second , the normalization factor of the various pdfs should scale as @xmath233 .
these requirements can be summarized by @xmath234 where @xmath235 is the scale factor .
an important remark is that the master curve for the pdf @xmath236 does not have to be gaussian ( although it can be ) in order to imply self - affinity , it just has to be the same across all the different scales @xmath182 .
although this is a very rigorous and precise method that allows both testing for self - affinity and determining the roughness exponent , it is not always possible to implement it in experimental systems as it requires precise data that spans many orders of magnitude .
the last method that we will describe is not based on real - space measurements , but rather on the fourier transform of the path , denoted here as @xmath139 , where @xmath154 is the wave - number .
more precisely , the power - spectrum @xmath237 of a self - affine path is expected to scale as @xmath238 99 d. hull , _ fractography _ ( cambridge university press , cambridge , 1999 ) . k. ravi - chandar and b. yang , j. mech .
solids * 45 * , 591 ( 1997 ) . c. guerra , j. scheibert , d. bonamy , and d. dalmas , proc .
usa * 109 * , 390 ( 2012 ) .
p. daguier , e. bouchaud , and g. lapasset , `` roughness of a crack front pinned by microstructural obstacles '' , europhys .
lett . * 31 * , 367 ( 1995 ) .
j. schmittbuhl and k. j. mly , `` direct observation of a self - affine crack propagation '' , phys .
* 78 * , 3888 ( 1997 ) .
a. delaplace , j. schmittbuhl , and k. j. mly , `` high resolution description of a crack front in a heterogeneous plexiglas block '' , phys .
e 60 , 1337 ( 1999 ) .
k. j. mly , s. santucci , j. schmittbuhl , and r. toussaint , `` local waiting time fluctuations along a randomly pinned crack front '' , phys .
96 , 045501 ( 2006 ) .
j. chopin , a. prevost , a. boudaoud , and m. adda - bedia , `` crack front dynamics across a single heterogeneity '' , phys .
107 , 144301 ( 2011 ) .
b. cotterell and j. r. rice , `` slightly curved or kinked cracks '' , int .
* 16 * , 155 ( 1980 ) .
j. c. radon , p. s. leevers , and l. e. culver , `` fracture toughness of pmma under biaxial stress '' , in fracture 1977 , vol 3 , university of waterloo press ( 1977 ) 1113 - 1118 .
a. yuse and m. sano , `` transition between crack patterns in quenched glass plates '' , nature ( london ) * 362 * , 329 ( 1993 ) .
m. marder , `` instability of a crack in a heated strip '' , phys .
e * 49 * , 51 ( 1994 ) .
m. adda - bedia and y. pomeau , `` crack instabilities of a heated glass strip '' , phys .
e * 52 * , 4105 ( 1995 ) .
f. corson , m. adda - bedia , h. henry and e. katzav , `` thermal fracture as a framework for quasi - static crack propagation '' , int . j. fract . * 158 * , 1 ( 2009 ) .
r. d. deegan , s. chheda , l. patel , m. marder , h. l. swinney , j. kim and a. de lozanne , `` wavy and rough cracks in silicon '' , phys .
e * 67 * , 066209 ( 2003 ) .
t. menouillard , and t. belytschko , `` analysis and computations of oscillating crack propagation in a heated strip '' , int .
* 167 * , 57 ( 2011 ) .
b. b. mandelbrot , d. e. passoja and a. j. paullay , `` fractal character of fracture surfaces of metals '' , nature * 308 * , 721 ( 1984 ) .
e. bouchaud , g. lapasset and j. plans , `` fractal dimension of fracture surfaces : a universal value ? '' , europhys . lett .
* 13 * , 73 ( 1990 ) .
k. j. mly , a. hansen , e. l. hinrichsen and s. roux , `` experimental measurements of the roughness of brittle cracks '' , phys .
lett . * 68 * , 213 ( 1992 ) . m. marder , `` roughing it '' , science * 277 * , 647 ( 1997 ) .
j. schmittbuhl , j .-
vilotte and s. roux , `` reliability of self - affine measurements '' , phys .
e * 51 * , 131 ( 1995 ) . v.
y. milman , n. a. stelmashenko and r. blumenfeld , `` fracture surfaces : a critical review of fractal studies and a novel morphological analysis of scanning tunneling microscopy measurements '' , prog .
sci . * 38 * , 425 ( 1994 ) .
h. bakke and a. hansen , `` accuracy of roughness exponent measurement methods '' , phys .
e * 76 * , 031136 ( 2007 ) .
l. ponson , d. bonamy , and e. bouchaud , `` two - dimensional scaling properties of experimental fracture surfaces '' , phys .
lett . * 96 * , 035506 ( 2006 ) .
e. katzav , and m. schwartz , `` exponent inequalities in dynamical systems '' , phys .
* 107 * , 125701 ( 2011 ) .
e. katzav , and m. schwartz , `` dynamical inequality in growth models '' , epl * 95 * , 66003 ( 2011 ) .
f. plourabou , k. w. winkler , l. petitjean , j .-
hulin , and s. roux , `` experimental study of fracture surface roughness on rocks with crack velocity '' , phys .
e * 53 * , 277 ( 1996 ) .
a. s. balankin , o. susarrey , r. g. paredes , l. morales , d. samayoa , and j. m. lpez , `` intrinsically anomalous roughness of admissible crack traces in concrete '' , phys .
e * 72 * , 065101(r ) ( 2005 ) .
l. i. salminen , m. j. alava , and k. j. niskanen , `` analysis of long crack lines in paper webs '' , eur .
j. b * 32 * , 369 ( 2003 ) .
e. bouchbinder , i. procaccia , s. santucci , and l. vanel , `` fracture surfaces as multiscaling graphs '' , phys .
lett . * 96 * , 055509 ( 2006 ) .
n. mallick , p .-
cortet , s. santucci , s. g. roux , and l. vanel , `` discrepancy between sub - critical and fast rupture roughness : a cumulant analysis '' , phys .
lett . * 98 * , 255502 ( 2007 ) .
t. engy , k. j. mly , a. j. hansen , and s. roux s , `` roughness of two - dimensional cracks in wood '' , phys .
lett . * 73 * , 834 ( 1994 ) . c. poirier , m. ammi , d. bideau , and j. p. troadec , `` experimental study of the geometrical effects in the localization of deformation '' , phys .
. lett . * 68 * , 216 ( 1992 ) .
barabasi and h. e. stanley , _ fractal concepts in surface growth _
( cambridge university press , cambridge , 1995 ) .
e. katzav , m. adda - bedia , and b. derrida , `` fracture surfaces of heterogeneous materials : a 2d solvable model '' , epl * 78 * , 46006 ( 2007 )
. k. b. broberg , _ cracks and fracture _ ( academic press , london , 1999 ) .
j. b. leblond , _ mcanique de la rupture fragile et ductile _
( hermes science publications , 2003 ) .
r. v. goldstein and r. l. salganik , `` brittle fracture of solids with arbitrary cracks '' , int .
j. fracture * 10 * , 507 ( 1974 ) .
m. adda - bedia , r. arias , m. ben amar , and f. lund , `` generalized griffith criterion for dynamic fracture and the stability of crack motion at high velocities '' , phys .
e * 60 * , 2366 ( 1999 ) . j. a. hodgdon and j. p. sethna , `` derivation of a general three - dimensional crack - propagation law : a generalization of the principle of local symmetry '' , phys .
b * 47 * , 4831 ( 1993 ) .
i. afek et al . , `` void formation and roughening in slow fracture '' , phys . rev .
e * 71 * , 066127 ( 2005 ) .
k.p . mroz and z. mroz , `` on crack path evolution rules '' engineering fracture mechanics * 77 * , 1781 ( 2010 ) .
g. erdogan and g.c .
sih , `` on the crack extension in plates under plane loading and transverse shear '' , asme j. basic engineering * 85 * , 519 ( 1963 ) . v.
hakim and a. karma , `` crack path prediction in anisotropic brittle materials '' , phys .
* 95 * , 235501 ( 2005 ) .
v. hakim and a. karma , `` laws of crack motion and phase - field models of fracture '' , j. mech . phys .
57 * , 342 ( 2009 ) .
m. amestoy and j. b. leblond , `` crack paths in plane situations - ii .
detailed form of the expansion of the stress intensity factors '' , int .
j. solids structures * 29 * , 465 ( 1992 ) .
j. b. leblond , `` the stress intensity factor at the tip of a kinked and curved crack '' , int .
j. solids struct .
* 25 * , 1311 ( 1989 ) .
a. b. movchan , h. gao , and j. r. willis , `` on perturbations of plane cracks '' , int . j. solids struct .
* 35 * , 3419 ( 1998 ) .
y. sumi , `` a second order perturbation solution of a non - collinear crack and its application to crack path prediction of brittle fracture in weldment '' , naval architecture and ocean engineering * 28 * , 143 ( 1992 ) .
s. ramanathan , d. ertas , and d. s. fisher , `` quasistatic crack propagation in heterogeneous media '' , phys .
lett . * 79 * , 873 ( 1997 ) .
m. ansari - rad , s.m .
vaez allaei , and m. sahimi , `` nonuniversality of roughness exponent of quasistatic fracture surfaces '' , phys .
e * 85 * , 021121 ( 2012 ) .
j. zinn - justin , _ quantum field theory and critical phenomena ( 3 ed . ) _ , ( oxford university press , oxford , 1996 ) .
b. derrida and e. gardner , `` lyapunov exponent of the one dimensional anderson model : weak disorder expansions '' , j. physique * 45 * , 1283 ( 1984 ) .
d. bonamy , l. ponson , s. prades , e. bouchaud , and c. guillot , `` scaling exponents for fracture surfaces in homogeneous glass and glassy ceramics '' , phys .
lett . * 97 * , 135504 ( 2006 ) . s. santucci , k. j. mly , a. delaplace , j. mathiesen , a. hansen , j. .
h. bakke , j. schmittbuhl , l. vanel , and p. ray , `` statistics of fracture surfaces '' , phys .
e * 75 * , 016104 ( 2007 ) .
l. ponson , h. auradou , m. pessel , d. lazarus , and j. p. hulin , `` failure mechanisms and surface roughness statistics of fractured fontainebleau sandstone '' , phys . rev .
e * 76 * , 036108 ( 2007 ) .
t. horst , k. reincke , s. ilisch , g. heinrich and w. grellmann , `` fracture surface statistics of filled elastomers '' , phys .
e * 80 * , 046120 ( 2009 ) .
d. bonamy , `` intermittency and roughening in the failure of brittle heterogeneous materials '' , j. phys .
* 42 * , 214014 ( 2009 ) .
e. katzav and m. schwartz , `` self - consistent expansion for the kardar - parisi - zhang equation with correlated noise '' , phys .
e * 60 * , 5677 ( 1999 ) .
e. katzav and m. schwartz , `` kardar - parisi - zhang equation with temporally correlated noise : a self - consistent approach '' , phys .
e * 70 * , 011601 ( 2004 ) .
r. lifshitz , `` quasicrystals : a matter of definition '' , foundations of physics * 33 * , 1703 ( 2003 ) .
e. katzav , s. f. edwards and m. schwartz , `` structure below the growing surface '' , europhys . lett . * 75 * , 29 ( 2006 ) .
d. r. askeland and p. p. phule , _ the science and engineering of materials ( 5 ed . ) _ ( thomson , 2006 )
. h. gao and j. r. rice .
, `` first - order perturbation analysis of crack trapping by arrays of obstacles '' , j. appl . mech . * 56 * , 828 ( 1989 ) .
s. ramanathan and d. s. fisher , `` dynamics and instabilities of planar tensile cracks in heterogeneous media '' , phys .
lett . * 79 * , 877 ( 1997 ) .
s. ramanathan and d. s. fisher , `` onset of propagation of planar cracks in heterogeneous media '' , phys .
b * 58 * , 6026 ( 1998 ) . m. adda - bedia , e. katzav and d. vandembroucq , `` second - order variation in elastic fields of a tensile planar crack with a curved front '' , phys
e * 73 * 035106 ( 2006 ) .
e. katzav and m. adda - bedia , `` roughness of tensile crack fronts in heterogeneous materials '' , europhys . lett .
* 76 * ( 2006 ) 450 - 456 .
e. katzav , m. adda - bedia , m. ben amar , and a. boudaoud , `` roughness of moving elastic lines - crack and wetting fronts '' , phys .
e * 76 * , 051601 ( 2007 ) .
n. pindra , v. lazarus and j. b. leblond , `` geometrical disorder of the fronts of a tunnel - crack propagating in shear in some heterogeneous medium '' , j. mech .
. sol . * 58 * , 281 ( 2010 ) .
m. vasoya , j. b. leblond , and l. ponson , `` a geometrically nonlinear analysis of coplanar crack propagation in some heterogeneous medium '' , int .
j. sol . and struc .
* 50 * , 371 ( 2013 ) .
r. c. ball and h. larralde , `` linear stability analysis of planar straight cracks propagating quasistatically under type i loading '' , int .
* 71 * , 365 - 377 ( 1995 ) .
a. a. al - falou , h. larralde and r. c. ball , `` effect of t - stresses on the path of a three - dimensional crack propagating quasistatically under type i loading '' , int .
j. solids structures * 34 * , 569 - 580 ( 1997 ) . | we study the stability and roughness of propagating cracks in heterogeneous brittle two - dimensional elastic materials .
we begin by deriving an equation of motion describing the dynamics of such a crack in the framework of linear elastic fracture mechanics , based on the griffith criterion and the principle of local symmetry .
this result allows us to extend the stability analysis of cotterell and rice to disordered materials . in the stable regime
we find stochastic crack paths .
using tools of statistical physics we obtain the power spectrum of these paths and their probability distribution function , and conclude they do not exhibit self - affinity .
we show that a real - space fractal analysis of these paths can lead to the wrong conclusion that the paths are self - affine . to complete the picture
, we unravel the systematic bias in such real - space methods , and thus contribute to the general discussion of reliability of self - affine measurements . |
theory suggests that supernovae and/or winds from massive stars existing in clusters confined to the disk of the galaxy produce bubbles in the interstellar medium which expand , becoming superbubbles , and burst out of the galactic disk producing collimated outflows into the galactic halo ( norman & ikeuchi 1989 ; tenorio - tagle & bodenheimer 1988 ) .
the superbubbles , which create cavities in the atomic hydrogen gas in the disk of the galaxy , contain a plasma of hot gas , high energy photons , and cosmic rays ( crs ) . when a bubble s wall bursts at high latitudes the structure is termed a `` chimney '' since the cavity and its walls act as a conduit for the outflow of this plasma towards the halo of the galaxy ( see e.g. de avillez & berry 2001 ) .
in addition , the magnetic field that is oriented parallel to the disk of the galaxy is `` frozen '' into the plasma and is thus `` pulled ''
upwards with the shell material as the bubble expands .
consequently , it is predicted that these conduits should have magnetic field lines that run tangential to the shell ( norman & ikeuchi 1989 ; basu , johnstone & martin 1999 , hereafter bjm99 ) .
an unusual , cone shaped void observed in neutral hydrogen ( normandeau , taylor & dewdney 1996 , hereafter ntd96 ) , which has the star cluster ocl 352 at its base , is one of the few observationally identified galactic chimney candidates and has come to be known as the `` w4 chimney '' .
it was proposed that the region has been evacuated of by the winds of 9 o - type stars that are part of the open cluster ocl 352 . in this case
the stellar wind scenario is favoured over the supernovae scenario since the cluster seems too young to have had any supernovae occur and the energy output of the winds appears to be sufficient to account for the void ( ntd96 ) .
estimates of the age of ocl 352 range from 1.32.5 myr ( dennison , topasna & simonetti 1997 , hereafter dts97 , and references therein ) to 3.74.3 myr ( ntd96 ) .
we adopt a distance to the region of 2.35 kpc .
the `` chimney '' region is located above ocl 352 and is a distinct component from the ionized `` w4 loop '' , which is located below ocl 352 ( normandeau , taylor & dewdney 1997 ) .
we consider only the structure that is above the star cluster and above the ridge of ionized gas that defines the upper latitude structure of the w4 hii region .
the hi images ( see ntd96 and figures [ fig : hi-43.40_general ] and [ fig : hi-43.40_faint ] ) make it clear that the regions above and below the cluster , i.e. above and below the upper ionized ridge of w4 , are different in nature : the upper region has been mostly cleared of hi , hence the possibility that it is a chimney or a superbubble .
the lower region has not been cleared , at least not at the same velocities .
the interpretation of terebey et al .
( 2003 ) and bjm99 that the superbubble structure includes the w4 loop is in contrast to this picture . while it is likely that both the w4 loop and the structure above w4 are powered by the same stars
, they do not form one dynamical entity .
the use of the term `` chimney '' provides a metaphor only for the part of the structure above w4 since a chimney not only has walls but it also has a fire at its base . in this case
the fire is the ocl 352 cluster . to avoid unwieldy nomenclature , rather than referring to the structure as the w4 superbubble / chimney , we dub it g134.4 + 3.85 .
the new name , however , should not be understood as implying that g134.4 + 3.85 has a different origin from w4 .
indeed we will argue that both objects are thermal emission regions excited by the open cluster ocl 352 . in the radio rgime , at galactic latitudes of less than + 4
, the structure is reminiscent of norman & ikeuchi s ( 1989 ) conventional chimney model , i.e. a broken shell allowing crs and uv photons to escape to higher latitudes .
the contrasting analysis of dts97 , using narrow - band data , is that w4 is an `` apparently closed superbubble '' .
however the dts97 data suffer from substantial vignetting at the field edges which could lead the eye to see a bubble shape where none exists .
this makes corroboration from other datasets desirable .
additionally , the dts97 data do not include velocity information , which means that the observed emission could be from unrelated sources along the line of sight .
bjm99 considered both the dynamics ( kompaneets modelling ) and the ionization of the superbubble based on the data from dts97 . of particular interest here
is the conclusion of bjm99 that an open geometry in images and a closed one in data showing ionized gas are not mutually exclusive
. the superbubble s shell could be sufficiently thin at high latitudes that , while it closes the shell and prevents streaming of gas towards higher latitudes , it does not hold back the ionizing radiation which then obliterates the at higher latitudes .
this scenario is in fact likely according to these authors . in this paper
we present new observations of radio continuum and hi emission , extending coverage from @xmath2 , the limits of earlier data , to @xmath3 .
we analyze the spectral line emission as well as the 1420 mhz and 408 mhz radio continuum emission from this region for evidence supporting either the chimney or the superbubble picture . for the 1420 mhz continuum data ,
full polarization information is available .
the data acquisition and processing are described in [ sec : data ] .
image mosaics and false colour morphological analysis are presented in [
sec : mosaics ] .
the emission is addressed in [ sec : hi ] .
the polarization data are discussed in
[ sec : polar ] .
the inter - relationship between emission components is discussed in [ sec : inter ] and the magnetic field of the region is explored , using our polarization data , in
[ sec : best ] .
our analysis creates a picture of a _ fragmented _ superbubble , g134.4 + 3.85 , above w4 that may be in the process of evolving into a chimney possibly due to magnetic effects . while we use two separate names we believe that both `` objects '' are the outcome of the ionizing flux from ocl 352 , and are therefore in some sense just one object . in [ sec : discuss ] we discuss this picture , superbubble formation scenarios , and the superbubble s connection to the galactic halo .
the paper is summarized in [ sec : conclu ] .
landecker et al . ( 2000 ) provide a complete description of the dominion radio astrophysical observatory ( drao ) synthesis telescope .
the telescope is an east - west array of seven antennae with diameters @xmath49 m. of these seven , four are fixed and three are movable allowing sampling of baselines from 12.86 m to 604.29 m in steps of 4.29 m. at @xmath5 cm , this corresponds to coverage on all spatial scales from @xmath41 arcmin to 56 arcmin . the data collected consist of several components : spectral line emission from atomic hydrogen at @xmath6 cm , continuum emission at 1420 mhz with full polarization information , and continuum emission 408 mhz .
the drao synthesis telescope recently completed phase i of the canadian galactic plane survey ( cgps ; taylor et al .
the cgps is a relatively high resolution ( @xmath41 at 1420 mhz ; @xmath43.7 at 408 mhz ) dataset covering 73.1 in longitude , from @xmath7 to @xmath8 along the galactic plane of the milky way .
the latitude extent covers @xmath9 .
three higher latitude fields , on an extension of the cgps grid , were acquired for the present study . at 1420 mhz
the field of view of the synthesis telescope has a full width at half maximum ( fwhm ) of 107.2 . at 408 mhz ,
the fwhm of the field of view is 332.1 .
the survey region has been observed on a regular hexagonal grid with 112 between field centers .
the 1420 mhz and 408 mhz mosaics used in the present study consist of 11 cgps fields plus the additional 3 fields observed at latitudes greater than + 5.6 having field centers 135.47 , + 5.85 ; 134.53 , + 7.46 ; and 133.60 , + 5.85 .
the polarization mosaics contain the 14 fields described above plus an additional 4 cgps fields , which reveal more of the `` ambient '' galactic medium surrounding the w4 region .
mosaics consist of the pilot project data ( normandeau , taylor & dewdney 1997 ) plus 3 additional cgps fields .
since the minimum spacing of the drao synthesis telescope is 12.86 m , data from single - antenna telescopes are added to regain missing large scale structure . at 1420 mhz ,
single - antenna continuum , total - intensity data are obtained from the 100 m effelsberg telescope ( reich et al . 1990 , 1997 ) for latitudes up to + 4 , and from the stockert 25-meter northern sky survey ( reich 1982 , reich & reich 1986 ) for higher latitudes .
the effelsberg telescope has a resolution of 9 , whereas the stockert telescope has a resolution of 35 .
the effelsberg data were absolutely calibrated using the stockert survey data , which ensures consistency amongst the contributions from the short - spacing data .
short - spacing information is only added to the 1420 mhz total intensity continuum maps ( stokes i ) and not to the stokes q or stokes u polarization maps . at 408 mhz
, short - spacing data are obtained from the all sky map of haslam et al.(1982 ) with a resolution of 51 .
for the spectral line data , short - spacing information was collected using the drao 26-m telescope .
the radio continuum emission at 1420 mhz is observed in four discrete 7.5 mhz bands centered at @xmath10 mhz and @xmath11 mhz from the spectral line frequency of 1420.406 mhz .
this leaves a 5 mhz gap at the central frequency , which is sufficient to avoid contamination from the spectral line .
the central 1 mhz of the gap is used for the spectral line observations .
the four continuum bands are combined when producing the images used in this paper ( stokes i , stokes q and stokes u ) .
taylor et al . ( 2003 ) indicate that there is an uncertainty in the absolute scale of the 408 mhz data of up to @xmath415% due to uncertain calibration data .
the internal precision of the data is unaffected by this error .
the processing of all cgps fields and the addition of short - spacing data in all fields were carried out by members of the cgps data processing team using custom software developed at drao ( willis 1999 ; taylor et al . 2003 ) .
processing of the additional , high - latitude fields and the mosaicking of fields were carried out by the authors using the drao software , with the assistance of the cgps data processing team .
the higher latitude data were processed , calibrated , and supplemented with short - spacing data such that they matched the cgps survey data .
mosaics showing the spectral line emission at velocities of interest are shown in figures [ fig : hi-43.40_general ] , [ fig : hi-43.40_faint ] , [ fig : hi-40 ] , and [ fig : hi-48 ] .
the mosaics of continuum emission at 408 mhz and 1420 mhz , stokes i , q and u are presented in figures [ fig:408 ] through [ fig : qu ] .
the grating rings that remain visible in the hi and polarization data are due to the high intensity sources w3 and 3c 58 .
the 1420 mhz data were convolved to match the lower resolution ( @xmath43.7 ) of the 408 mhz data .
the polarization data were convolved to a resolution of 5 to improve the signal to noise ratio and the stokes q and u images were combined to produce a polarized intensity image ( @xmath12 ) and a polarization angle map ( @xmath13 ) ( see figure [ fig : pipang ] ) .
measurements of stokes i intensities and pi for 56 regions of interest are tabulated in west ( 2003 ) .
[ [ sec : hi ] ] figure [ fig : hi-43.40_general ] shows the data at @xmath14 .
the eastern wall of the superbubble is clearly visible up to + 5.5 , at which point it curves slightly inward and disappears .
the western wall is only well defined up to + 3.4 . in the new data set presented here ,
there is no evidence of a cap at high latitudes in the . at lower latitudes , the cavity spans 5 channels and lsr velocities 38.46 to 45.05 .
we consider the region defined by @xmath15 , @xmath16 , the latitudes associated with the shell described by dts97 .
we assume that the cap would have a velocity width of 5 channels ( 8 ) , the velocity width of the cavity . averaging the signal over 5 channels
, we could have detected a cap of column density @xmath17 @xmath18 .
using 5 channels reduces @xmath19 from the value of the rms noise in a single channel , which is 3.3 k or @xmath20 @xmath18 .
thus if such a cap had been detected its column density would be at the 6@xmath19 level .
the observed column density in the top of the ionized bubble is @xmath21 @xmath18 . by comparison ,
the column density of electrons in the dts97 shell is @xmath22 @xmath23 . for this latter estimate we converted emission measure into column density using an average of dts97 s measurements of 4 points in the upper shell , t = @xmath24
k , their emission measure conversion of 1r = 2.8 pc @xmath25 , and their shell thickness estimate of 10 to 20 pc .
if there was neutral material associated with this shell at a density comparable to that of the ionized material , we would not have been able to detect it .
there is a small filament of pointing upward , away from the plane ( figure [ fig : hi-43.40_faint ] ) .
this low level feature is present in four channels of the mosaic , from 41.76 to 46.70 . a line through this linear feature passes through ocl 352 .
we will argue in [ sec : chimney ] that this is material within the superbubble which remains neutral , protected from the ionizing radiation by shadowing .
this faint arc is well below the latitude of the cap claimed by dts97 and therefore below the upper boundary of the best fit kompaneets model found by bjm99 .
furthermore , the upper tip of the vertical filament is above the ionized cap suggested by the dts97 data .
it is noticeable that the ocl 352 cluster is at the base of the cone , as was described by ntd96 , whereas the kompaneets model by bjm99 , which was fit primarily to the h@xmath26 image , shows a bubble extending to lower latitudes , to the base of the ionized gas loop which forms the lower half of w4 .
this conical shape is not unique - something similar is seen at the base of the aquila supershell ( maciejewski et al .
the w4 cone extends approximately 200 pc upward from the ocl 352 cluster , based on the well - defined upper longitude wall .
this is comparable to the cone at the base of the aquila supershell which extends roughly 175 pc for an assumed distance of 3.3 kpc .
the most surprising result from the modeling by bjm99 is the implication of a very small scale height ( @xmath27 ) , namely 25 pc .
this value was obtained by matching the aspect ratio of the superbubble as seen in h@xmath26 using the kompaneets model and assuming a distance of 2.35 kpc . using the data from the 26-m telescope , which extends further up in latitude , almost to + 10 , and fitting an exponential to the decay in column density ,
a scale height of 140 @xmath28 40 pc is found for the in the vicinity of the g134.4 + 3.85 ( c.f . figure [ fig : nvsz ] ) .
a variety of different velocity intervals were used to test the robustness of this result , as well as different data sets ( 26-m data , cgps pilot project data , and cgps data ) , and all yield a scale height of approximately 140 pc , which is significantly larger than the 25 pc required by the model of bjm99 .
this value compares well with the values of 135 pc quoted by lockman , hobbs & shull ( 1986 ) .
apart from the contradiction with the prediction by bjm99 , this result is not particularly surprising : scale heights for the neutral medium are in the 100 - 200 pc range .
clearly , however , although the fit of the kompaneets model to the shape of the w4 superbubble is very good , it does not address all of the physical processes occurring in this vicinity .
in particular , magnetic fields may have contributed to the collimation ( komljenovic et al .
1999 , de avillez & breitschwerdt 2005 ( see [ sec : time ] ) ) .
it should be noted that the scale height evaluated here used data from all longitudes covered by the available 26-m data , from 124.4to 144.4 with varying coverage in latitude .
however , if only longitudes greater than those corresponding to g134.4 + 3.85 are used , the profile is greatly different for latitudes below + 6 , particularly between + 1.5 and + 4.0 where the high longitude data show a plateau instead of a decrease .
the difficulty lies in deciding which profile is representative of the medium into which the superbubble expanded .
the effect of including g134.4 + 3.85 itself in the longitude range is minor compared to the effect of what appears to be a partial shell at lower longitudes which may be associated with 3c 58 .
these difficulties generate the uncertainty of @xmath28 40 pc associated with the 140 pc scale height used in this paper .
gray et al .
( 1999 ) presented polarization data from the low latitudes ( @xmath29 ) of this region .
the high electron density combined with the magnetic field in the w3/w4/w5 region act as a `` faraday screen '' , causing the polarized , background galactic synchrotron radiation passing through the plasma to be rotated by the process of faraday rotation .
these effects are observationally recognized as depolarized regions that follow the contours of the regions .
depolarization occurs when the polarization angle varies by a significant amount between vectors contributing to a single measurement .
the polarization angle can differ between adjacent lines of sight sampled by a single telescope beam ( beamwidth depolarization ) or vary along the path length ( depth depolarization ) . a third depolarization mechanism ,
bandwidth depolarization , becomes important if the rotation measure ( rm ) is very high at the observing wavelength , and so the angle may change by a large amount across the observing band .
bandwidth depolarization is not significant in this study as typical rms in the region are on the order of @xmath30 and to produce significant depolarization over the 35 mhz observing band of these observations , a rm of @xmath41400 would be required ( brown 2002 ) . in fig .
[ fig : qu ] and fig .
[ fig : pipang ] we see that the depolarization extends without interruption from w4 up the `` walls '' of g134.4 + 3.85 .
this is seen in stokes q and u and in polarized intensity .
the continuity of depolarization from w4 to the walls of the superbubble provides strong evidence that the superbubble is connected to the hii region ; this is not a chance superposition . a temperature spectral - index map ( west 2003 )
also suggests that w4 and g134.4 + 3.85 are a single coherent structure .
an unusual wishbone - shaped feature is visible in both stokes q and stokes u as well as in polarized intensity and polarization angle maps .
this feature has no counterpart in stokes i , total - intensity .
the feature appears above g134.4 + 3.85 with approximate center coordinates ( 134.9 , + 7.1667 ) ( see fig .
[ fig : qu ] and fig .
[ fig : pipang ] ) .
the morphology of the object suggests a shell - like structure .
a local steepening of the spectral index is observed south of the wishbone ( west 2003 ) .
a bright , comma - shaped knot of strong polarized intensity with approximate center coordinates of ( 134.5 , + 2.58 ) is particularly prominent in polarized intensity ( fig .
[ fig : pipang ] ) but invisible in total - intensity .
it is @xmath40.5 wide and @xmath40.2 high and has a peak polarized intensity of 0.46 k. the object appears near the base of the prominent v - shaped filaments which are visible in both and 60 @xmath31 m iras dust emission .
however , there is no clear morphological evidence of an association .
the simbad database was searched for catalogue objects which might be coincident with the comma - shaped knot on the plane of the sky .
a prominent object in this region is an o8 star , bd+62 424 with coordinates 134.53 , + 2.46 .
there have been two , distinctly different heliocentric radial velocity measurements for this star : abt , levy , & gandet ( 1972 ) measured a value of 42.5 ( 40.2 lsr ) while fehrenbach et al .
( 1996 ) measured a value of 9 ( 6.7 lsr ) .
the measurement by abt et al .
is consistent with the chimney velocities as observed in the spectral line .
however , the radial velocity measurement of fehrenbach et al . ( 1996 ) yields a kinematic distance of @xmath4800 pc ( brand & blitz 1993 ) implying that this star is a foreground object .
the high polarized intensity of this knot suggests that it may be closer than the w4 region as it suffers less depolarization than the surrounding region .
to aid in comparison of the various wavelengths , a colour image combining the datasets is shown in figure [ fig : colour ] .
this image illustrates that there is correlation of radio continuum and emission , but that the iras emission seems uncorrelated .
the regions w3 , w4 and w5 , the supernova remnants ( snrs ) hb3 and 3c 58 , as well as the spiral galaxy maffei 2 are readily seen .
the superbubble , identified by dts97 , is much more apparent in these combined colour images than in the individual images . in fig.[fig : labelled ] we outline an egg - shaped bubble and suggest its three - dimensional structure .
a feature that seems to be a continuous branch of emission coming off the main shape of g134.4 + 3.85 has been labeled the fork . the new data presented in this paper show strong evidence of radio continuum emission above ocl 352 , associated with the w4 loop below the star cluster , and extending up to a latitude of @xmath32 .
the emission seen at 1420 mhz and 408 mhz generally confirms the dataset of dts97 , and reveals g134.4 + 3.85 to have a striking , continuous egg - shaped structure .
in addition , the polarization data reveal a depolarized region consistent with the shell - like structure and indicate that thermally emitting material is depolarizing the background synchrotron emission ( see [ sec : polar ] and [ sec : best ] ) . on the east wall
we see clearly that the ionized material ( whose position is judged from the depolarization feature in figures [ fig : qu ] and [ fig : pipang ] ) lies inside the neutral material ( position judged from figure [ fig : hi-43.40_general ] ) .
the relative temperature spectral index is flatter within the region outlined by the depolarized shell and the value of the index indicates that this emission is thermal ( west 2003 ) .
although the data suggest that g134.4 + 3.85 has a top ( we have already noted in [ sec : hi ] that our observations do not have sufficient sensitivity to show atomic gas connected with the ionized material at the top of the shell ) , the shape does not have a uniform intensity around the perimeter .
there are regions of lower intensity that suggest breaks in the continuous structure .
two main breaks appear in the structure , one near the top and one on the lower longitude side near @xmath33 .
these are seen at both radio frequencies and in with closely coincident positions .
the center coordinates and approximate widths for the two breaks are listed in table [ tbl : breaks ] . at 1420 mhz and in , there is emission , nicknamed the fork , that branches off the main structure ( figure [ fig : contours ] ) .
a faint arc of emission extends up from the high longitude side of the fork and appears to arc westward back to g134.4 + 3.85 .
this arc of emission is coincident with the top portion of the polarization wishbone .
the lower portion of the fork exhibits depolarization similar to that of the lower portion of the wall , suggesting that the fork is associated with g134.4 + 3.85 . however it is unclear whether the continuing arc of emission is associated with g134.4 + 3.85 .
a hand tracing of the regions of higher emission including the fork and the apparent breaks is presented in figure [ fig : contours ] .
there is a distinctive region of very low continuum emission intensity inside the walls of g134.4 + 3.85 above @xmath34 .
it extends up to + 4.8@xmath35 and spans @xmath36 .
figure [ fig : ivsb_ivsl ] shows plots of the 1420 mhz data at fixed longitude ( @xmath37 ) and fixed latitude ( @xmath38 ) . in both plots this region of very low emission ( @xmath39 k ) is evident ; in figure [ fig : ivsb_ivsl]a the area runs from @xmath40 and in figure [ fig : ivsb_ivsl]b the area extends from @xmath36 .
the region has lower emission than even the regions exterior to g134.4 + 3.85 .
this low emission region is most obvious at 1420 mhz but is also apparent in both 408 mhz and .
there appears to be 60 @xmath31 m emission in this region from dust that one could reasonably expect to find in the foreground along the line of sight .
the low emission implies either that the region is evacuated of plasma that would emit at these wavelengths or that , in the case of the data , uv photons contained within are being absorbed by dust . in the former case
, there would be lower plasma column density along the lines of sight passing through this region producing a lower observed intensity . in this low emission region , there are two horizontally oriented filaments positioned at @xmath41 which were first identified by ntd96 .
these are either foreground or background features , although , judging by their radial velocity , quite likely to be in the vicinity of g134.4 + 3.85 .
they are unliekly to be within it , given the evidence that ionization extends to much higher latitudes .
additionally , a ridge of and 1420 mhz emission appears at @xmath42 to + 3.5 .
the ridge at 1420 mhz is visible in figure [ fig : stokesi ] .
coincident with the 1420 mhz ridge is a flattening of the temperature spectral index and below this ridge but above dts97 point `` o '' ( west 2003 ) the spectral index steepens , indicating a shift between these two features ( west 2003 ) .
this supports the picture of a lower region containing electrons and ionizing radiation ( at @xmath43 ) with an evacuated region above the ridge ( at @xmath44 ) . to summarize , although the structure appears open in the images , data showing the ionized gas , along with polarization and temperature spectral index maps , delineate an egg - shaped bubble with a shell that appears to be broken in at least two locations .
the temperature spectral index map and data suggest that radiation may possibly be constrained to the lower part of the bubble by a ridge that is evident in and 1420 mhz data .
rotation measures of extragalactic sources seen through the bubble were measured in an attempt to determine the magnetic field in the shell @xcite but the uncertainties are large and the number of sources is small , allowing @xmath45 to take any value between 1 and 24@xmath31 g . since rotation measure analysis does not constrain the magnetic field well , we use the following approach based on depolarization in the walls of the shell .
we noted above that the walls of g134.4 + 3.85 are seen in our data as regions of low polarization relative to the surroundings of the superbubble and that they join smoothly to the depolarization region coinciding with w4 @xcite . from the observed degree of depolarization in the walls
we can derive an estimate of the magnetic field in the shell of the superbubble .
we use the electron density derived from the measurements of dts97 , and base our calculation on an assumed geometry of the walls .
the electron density is derived from the flux measurements of dts97 measured at four locations along the upper arc ( @xmath46 ) of the shell ( dts97 s points labeled h , i , j , and k ; see figure [ fig : contours ] ) .
we estimate the shell thickness , @xmath47 , from the projected intensity to be between 10 and 20 pc .
assuming a temperature of 10@xmath48 k , and using 1 r ( rayleigh ) as equivalent to an emission measure of em = 2.8 , and dts97 s result that the emission measure of the shell @xmath49 , then @xmath50 is calculated to be 0.44 @xmath28 0.19 for an assumed shell thickness of 10 pc , or 0.31 @xmath28 0.13 for an assumed shell thickness of 20 pc .
the range of possible electron density is approximately @xmath51 . for the following geometrical argument
, we assume that the superbubble is cylindrical , with its axis perpendicular to the galactic plane ( and therefore perpendicular to the line of sight ) .
its cross - section is circular , with outer diameter 80 pc and wall thickness between 10 and 20 pc .
the line of sight passes through the cylinder of ionized material in which the electron density is assumed to take the values derived above , and the polarized synchrotron background emission will suffer faraday rotation on passage through this material if a magnetic field is present and has a line - of - sight component . the path length through the wall of the cylinder for adjacent lines of sight will differ , and
so the amount of faraday rotation will differ .
if the rotation is sufficiently different for adjacent paths within the telescope beam , there will be significant beam depolarization .
the data presented in figures [ fig : qu ] and [ fig : pipang ] , where the angular resolution is 5 , indicate that the depolarization amounts to @xmath450% , implying that the faraday rotation on lines through the shell separated by 5 differ by about 60@xmath52 . at the distance of g134.4
+ 3.85 ( 2.35 kpc ) 5corresponds to a physical distance of 3.5 pc .
a simple geometrical calculation shows that two paths through the center of the wall of the cylinder , separated by a distance 3.5 pc transverse to the line of sight , will differ in length through the wall by @xmath53 pc if the wall thickness is 10 pc and by @xmath54 pc if the wall thickness is 20 pc .
the faraday rotation is @xmath55 where @xmath56 is the wavelength , @xmath50 is the electron density , @xmath57 is the line - of - sight component of the magnetic field , and @xmath58 is the path length . inserting the values @xmath59 , @xmath60 , with values for @xmath61 and @xmath58 derived above
, we find @xmath62 .
the field derived for the `` average '' parameters ( @xmath63 and @xmath64 pc ) is @xmath65 .
the range of possible values of electron density in the walls of the superbubble , @xmath51 is 10 to 30 times the ambient electron density in the vicinity of w4 , @xmath40.02 ( taylor & cordes 1993 ) .
this implies that the material in the walls of g134.4 + 3.85 has been compressed by this factor .
@xmath45 at this longitude is expected to be @xmath66 , and @xmath67 is about @xmath68 ( beck 2001 ) . if @xmath69 varies as @xmath70
, then we expect @xmath71 in the walls of the bubble to be between 3.1 and @xmath72 , compatible with the above field estimate . however , we need to consider the likelihood that magnetic field and electron density will be uniform , as is assumed in our field calculation . in the q ,
u , and pi images ( figures [ fig : qu ] and [ fig : pipang ] ) the surroundings of g134.4 + 3.85 present a mottled appearance that is strongly suggestive of a non - uniform magneto - ionic medium . if we think of the medium as divided into `` cells '' in which the field and electron density are more or less constant , then the size of the cells appears to be about 20 in the surroundings . with a compression factor of 10 to 30
, the linear dimension of the cells will shrink by the square root of the compression factor to about 4 .
beam depolarization , which would arise from averaging across many cells , is then unlikely to be significant in the 5 data , and the geometrical effect we have proposed seems to be an adequate description of the processes leading to the observed depolarization .
nevertheless , there is still a problem .
at 1 , the full resolution of the observations , we would expect to see much less depolarization , both when the geometrical effect is dominant and when the bubble walls contain many cells , and we should not see the bubble walls as significant depolarization features .
however , at 1resolution the eastern wall of the bubble is still detectable as a depolarization feature ( although the western wall is not ) .
there must be some cells in the eastern wall as small as 1 , but we can not place a precise value on the extent of this small - scale structure because the signal - to - noise ratio at 1 is very low .
the level of polarized intensity in the eastern wall relative to that of the surroundings is very roughly 1.6 , although the uncertainty in this ratio is so high that we can not claim that it is higher or lower than the value of 2.2 @xmath28 0.3 measured from the data at 5resolution .
in conclusion we note that bjm99 propose the presence of a swept - up magnetic field , tangential to the shell of the w4 chimney / superbubble , that potentially has a stabilizing effect on the structure ( by suppressing rayleigh - taylor instabilities that tend to break up the shell ) .
kolmjenovic et al .
( 1999 ) use magnetohydrodynamic modeling to argue that a _ vertical _ magnetic field of a few microgauss is required to explain a bubble structure that is so highly collimated , and to provide the required stabilizing effect .
our depolarization observations indicate that a line - of - sight component of 3 to 5 @xmath73 exists in the shell . however , this measurement does not exclude the possibility of a stronger vertical component , giving a total field of 4 to 7 @xmath73 or larger .
at our adopted distance of 2.35 kpc , the projected ellipse of g134.4 + 3.85 , which measures @xmath44 wide by @xmath46high , has physical width @xmath4164 pc and height @xmath4246 pc .
it has risen to almost twice the scale height of the , @xmath4140 pc , determined in [ sec : scale_height ] , and at that level the density is only 13% of its mid - plane value ( figure [ fig : nvsz ] ) .
nevertheless , the bubble has not reached the halo , considered to begin at a height of @xmath4500 pc @xcite .
two - dimensional numerical simulations ( e.g. maclow , mccray & norman 1989 ) and analytic theoretical explorations ( e.g. ferrara & tolstoy 2000 ) indicate that at a distance away from the plane of about 3 scale heights a bubble will accelerate and break out into the halo , creating a chimney .
not only has g134.4 + 3.85 not reached this height , it may not have the high levels of input wind ( or supernova ) energy to do so .
most treatments of a single superbubble evolving in a uniform environment indicate that substantially more driving energy is required for breakout .
for example , maclow , mccray , & norman ( 1989 ) show that a bubble driven by a mechanical luminosity of @xmath74 erg s@xmath75 , more than five times the output of ocl 352 , will break out into the halo .
tomisaka ( 1998 ) describes the evolution of a bubble driven by @xmath76 erg s@xmath75 , equal to the mechanical luminosity of ocl 352 , and show that it is unlikely to break out , and , furthermore , is stabilized by the ambient magnetic field .
mcclure - griffiths et al . ( 2003 ) describe the galactic chimney gsh 277 + 00 + 36 , perhaps a counter - example to g134.4 + 3.85 .
this is clearly a chimney , very empty inside , with very complex hi structures in the walls , interpreted as hydrodynamic instabilities . in this case
the gas flow within the chimney has had a major impact on the neutral gas forming its walls .
we see no such features in g134.4 + 3.85 .
furthermore , gsh 277 + 00 + 36 reaches to a height of 1 kpc above the galactic plane , at least four times as high as g134.4 + 3.85 .
g134.4 + 3.85 is either driven by a smaller mechanical energy than gsh 277 + 00 + 36 , or is at a much earlier stage of evolution .
thus the data we have presented seem at first glance to support a picture in which g134.4 + 3.85 is a superbubble , not a chimney .
the structure of ionized gas tapers at the top , suggesting convergence or closure . the 1420 mhz and data clearly show this converging top , and , to a lesser degree , so do the 408 mhz data .
there is no cap of neutral material at the upper levels in our data . nor
is there any detectable neutral shell surrounding the ionized one ; rather g134.4 + 3.85 appears as a cavity in the surrounding medium .
the images do , however , show the diverging conical shape exhibited by chimney models ( norman & ikeuchi 1989 ) .
these observations can be understood in terms of the picture presented by bjm99 . in the upper reaches of the bubble
the ambient density is dropping rapidly and the ionization structure is unbounded . all neutral gas has been ionized . in this picture ,
any neutral gas found within the conical void must be protected from uv photons by shadowing .
the prime example is the pair of filaments forming a * v * at @xmath77 .
a second possible instance , found in our observations , is the vertical filament extending from @xmath4(134.15 , + 6.083 ) to @xmath4(134.175 , + 7.25 ) and with a width @xmath40.25 ( see figures [ fig : hi-43.40_faint ] and [ fig : labelled ] ) , which appears in several velocity channels from 38.46 to 45.05 , velocities which are consistent with g134.4 + 3.85 velocities .
these features are linear , or made up of straight - line elements , and are oriented in line with the position of ocl 352 .
although the ionized shell appears closed , our observations have also revealed two gaps ; details are given in table [ tbl : breaks ] .
these regions of low emission may well be breaks in the shell .
one break is at the top of the bubble , at @xmath78 , and has an apparent width of 20 pc .
the ambient density is lowest at the top of the shell , and this is exactly where one might expect the shell to be disrupted .
the other break is at lower latitude , and appears to be larger , nearly 50 pc wide . in this break
we see neither neutral or ionized material .
we interpret these breaks as evidence that the superbubble is fragmenting , and turn to an existing theoretical treatment of superbubble breakout and disruption to see what might be learned from it .
we base our discussion on the recent work of de avillez & breitschwerdt ( 2005 ) , who have made high - resolution , three - dimensional simulations of the magnetized ism .
their calculations trace in fine detail the evolution of such a medium driven by supernova explosions at the galactic rate .
while they start with a simple , uniform medium , their simulations follow the evolution into the more realistic situation of a medium that becomes highly processed as the result of stellar activity .
the morphology of g134.4 + 3.85 is quite similar to some superbubbles shown in their figure 2 .
a bubble has developed , centered at @xmath79kpc , @xmath80kpc , in which the higher density skin , rather than forming a closed shell , is open at the top and can be described as filamentary or fragmented .
given its density , this skin can be identified as cool gas if the region is in pressure equilibrium .
interior to this skin is a more continuous lower density lining which could be identified with emission from warmer ionized gas .
this bubble has protrusions of even hotter gas at both the top and side which would appear as breaks in ionized gas observations , remarkably similar to the breaks seen in our continuum data for g134.4 + 3.85 .
the dimensions of the bubble are approximately 200 pc laterally and 600 pc vertically , somewhat larger than g134.4 + 3.85 because of this height , the model bubble can be considered to have reached the galactic halo at @xmath81 pc .
it has been shown with 2-d simulations ( e.g. maclow , mccray & norman ( 1989 ) ) , that , at this altitude of a few scale heights , rayleigh - taylor instabilities ( in which hot gas pushes into cooler , denser gas ) will occur and cause fragmentation of the shell .
however , instabilities , rayleigh - taylor or other types , can occur at earlier stages , according to the 3-d simulations .
indeed the dense skins of smaller bubbles in the de avillez and breitschwerdt simulation are also filamentary , indicating that this is so .
although a magnetic field in the shell can stabilize it against rayleigh - taylor instabilities , de avillez and breitschwerdt mention that mhd instabilities can occur which will mix the magnetized shell with the ambient medium and cause the escape of hot gas .
the field in the skin of their model bubble is a few @xmath31 g or less , and rayleigh - taylor instabilities would not be suppressed by this field .
the ambient field around g134.4 + 3.85 is only a few @xmath31 g , but , due to compression , the ionized lining of g134.4 + 3.85 could have a magnetic field strength as high as 7 @xmath31 g ( [ sec : best ] ) . thus rayleigh - taylor instabilities may be reduced and mhd instabilities , instead , may be the cause of both breaks apparent in g134.4 + 3.85 .
the simulations need to be examined in detail in order to assess whether this is the case ; de avillez has begun such an investigation in relation to g134.4 + 3.85 ( private communication ) . in summary , g134.4 + 3.85
has now risen into the lower density levels of the galactic disk , but not yet into the halo .
its ionized shell is apparently closed indicating that it is a superbubble . at high latitudes
, the distribution retains the conical chimney structure , but that is because the ionized shell is transparent to ionizing radiation .
gaps seen in the shell may well be the first signs of fragmentation of the superbubble and indicate that hot gas is escaping . if so , g134.4 + 3.85 is on its way to becoming a chimney .
the process is likely to accelerate as supernova explosions among the stars of ocl 352 inject energy into the bubble .
w4 and its immediate neighbour , w3 ( ic 1795 ) , together comprise a complex , highly processed sector of the galaxy .
many authors have suggested that multiple epochs of star formation have occurred in the region ( lada et al .
1978 ; thronson et al . 1980 ; carpenter , heyer & snell 2000 ; reynolds , sterling & haffner 2001 ) .
in particular , oey et al .
( 2005 ) offer evidence that ic 1795 was actually triggered by w4 and that ic 1795 has recently triggered the formation of three ultracompact regions around w3 .
given that this process is still underway , a question arises : is g134.4 + 3.85 simply the product of the winds from ocl 352 , or has it been shaped by earlier activity ?
the axis of g134.4 + 3.85 passes through the position of ocl 352 , and there can be little doubt that this star cluster is now driving the evolution of the bubble .
however , if winds from earlier massive stars swept out the environment , g134.4 + 3.85 could have extended to its present height relatively quickly .
oey et al .
argue further that the very low scale height ( @xmath425 pc ) deduced by bjm99 is the result of earlier clearing of the environment by a long - gone generation of stars .
however , the exact alignment of ocl 352 with the axis of g134.4 + 3.85 suggests that the superbubble is not simply filling an old cavity , which would most probably have no such alignment .
ntd96 calculate that the combined winds of the 9 o - type stars in the cluster could blow out the region up to a height of 110 pc in 5.7 myr which roughly agrees with cluster age estimates of 3.7 to 4.3 myr . in light of evidence that the structure extends up to @xmath4230 pc ,
dts97 calculate that 6.4 to 9.6 myr would be required to allow sufficient expansion .
they cite cluster ages of 1.3 to 2.5 myr , smaller than those obtained by ntd96 , and thus claim that ocl 352 can not be solely responsible for the formation of the superbubble . in passing we
note that both the calculations of ntd96 and dts97 ignored magnetic fields .
ferrire et al .
( 1991 ) have shown that a bubble expands more slowly in a magnetized medium : an even longer time would be needed to reach the present size of g134.4 + 3.85 .
reynolds et al .
( 2001 ) observe an loop extending 1300 pc above the midplane of the galaxy and centered on coordinates ( 139 , + 15 ) , which they believe is being ionized by ocl 352 .
they also suggest that mutiple star formation epochs have been involved in the formation of this loop , and suggest a timescale of 10 to 20 myr for the process .
if ocl 352 is maintaining the ionization of material 1300 pc above the plane , then the region must be completely clear of neutral material .
this is confirmed by our observations , at least to a height of about 300 pc .
we support the suggestion that ocl 352 is the dominant source of ionizing radiation for g134.4 + 3.85 but that it is not solely responsible for the creation and evolution of the structure .
further , we postulate that there is little matter remaining interior to the upper portion of g134.4 + 3.85 and that any gas that may have existed at an earlier stage of the life of the superbubble has escaped through the apparent breaks in the structure of g134.4 + 3.85 .
we have presented evidence that there is significant magnetic field in the walls of g134.4 + 3.85 ( [ sec : best ] ) .
ferrire et al .
( 1991 ) , tomisaka ( 1998 ) , and de avillez & breitschwerdt ( 2005 ) all consider bubbles in the magnetized ism ; all conclude that bubbles elongate in a direction parallel to the field .
tomisaka ( 1998 ) concludes that the field will inhibit expansion perpendicular to the field , and can also contain the expansion of the bubble .
the expansion can still accelerate vertically , even in the presence of a field aligned with the galactic plane , but only if it is driven by enough energy to reach several scale heights .
what then do we make of the _ vertical _ elongation of g134.4 + 3.85 ?
either it indicates that the scale height is extremely low , as suggested by bjm99 , or it implies a field component orthogonal to the galactic plane ( as suggested by komljenovic et al .
1999 ) . since we do not believe that the scale height is as low as the 25 pc derived by bjm99
, we must consider the possibility of a vertical component to the field .
such a component may be the relic of an earlier bubble and we are compelled by the very shape of g134.4 + 3.85 to consider that it may have formed in a pre - processed medium , not a pristine one .
we can envisage a situation where a new bubble has filled an old one , and the elongation may have become stronger with each successive wave of star formation .
this fits well into the picture of triggered star formation for this complex ( oey et al . 2005 ) where the new generation of stars is triggered within the walls of the superbubble created by the preceding generation .
there is also useful insight to be gained from analytical approximations available in the literature . both ferrire et al .
( 1991 ) and de avillez and breitschwerdt ( 2005 ) derive analytic equations applicable to a magnetohydrodynamic medium .
de avillez & breitschwerdt ( 2005 ) calculate the age and the radius ( in the equatorial plane ) of the bubble when its expansion stalls .
their expression for the age at stall is @xmath82 where @xmath83 is the mechanical luminosity of the stellar wind in units of @xmath84 ergs@xmath75 , @xmath85 is the density in @xmath86 , and @xmath87 is the ambient magnetic field in units of @xmath31 g .
the derivation of this expression is based on the assumption of a uniform ism and a regular magnetic field .
the bubble stalls when the force from thermal pressure is balanced by magnetic force as the bubble distorts field lines .
inspection of equation 2 shows that the stall time is very strongly dependent on the magnetic field but is less influenced by density or driving energy .
if the magnetic field goes up by a factor of 4 the time to stall drops by 32 ! nevertheless , the time to stall is long .
if we insert @xmath88 , @xmath89 , and @xmath90 ( note that this is the ambient field , not the compressed field , and that we use the regular component of the field , ignoring the random component ) , then we obtain @xmath91 myr , far longer than the age of ocl 352 .
if the field is as high as @xmath92 , this time drops to about 9 myr , a more reasonable value .
the radius of the bubble at stall is @xmath93 for the situation considered above the stall radius is @xmath41000 pc ( for @xmath90 ) and @xmath4250 pc ( for @xmath94 ) .
g134.4 + 3.85 is nowhere near this size , and has not stalled .
however , bubbles do in time stall .
contraction follows stall , and occurs mostly in the equatorial region , so that old bubbles become quite elongated in the direction of the field ( de avillez & breitschwerdt 2005 ) .
however , we have just demonstrated that g134.4 + 3.85 has not stalled , so the elongation can not be the product of contraction , but must be the product of an ambient field that already has a strong vertical component .
we therefore propose a model in which g134.4 + 3.85 is driven more by pre - existing magnetic field structure than by pre - existing density structure ; this is consistent with the ideas expressed by de avillez & bretischwerdt ( 2005 ) .
ferrire et al .
( 1991 ) describe a substantial thickening of the wall of a bubble formed in a magnetized medium . in their calculation for @xmath95 , @xmath96 , and @xmath97 , plausible values for g134.4
+ 3.85 , the equatorial shell thickness is about 20% of the radius . from our observations
the value of @xmath98 is between 12 and 24% , quite compatible with their model .
pursuing this idea further , we suggest another possible explanation for the observed gaps in the shell . in the analytical models ,
the shell thickness depends on location on the shell .
the above estimate refers to the magnetic equator ; at the magnetic poles the shell becomes very thin .
we speculate that this effect might be related to the apparent break in the top of the shell .
it is conceivable that it is not a break at all , but simply a thin spot .
we note that it is exactly at such a thin spot at the magnetic pole that one would expect such a shell to break . as well as these magnetic field arguments
, we find morphological evidence that supports a scenario with multiple star forming epochs . near the cluster
there is relatively intense @xmath422 k emission at 1420 mhz . around the cluster , but not symmetrically centered on it ,
is the w4 loop , a 1 ( @xmath440 pc ) loop visible in radio continuum , infrared , and emission with shell - like structure .
this loop does not have a constant limb brightness and is interpreted as having `` broken out '' in places ( terebey et al .
2003 ) . above @xmath99 ,
the 1420 mhz intensity drops from 8.8 k to @xmath1006 k. between @xmath101 and @xmath102 , the intensity decreases only slightly , from @xmath46 k to @xmath45.2k , after which one finds an arc above which the intensity drops quickly to @xmath44.75 k in the low emission region ( see figure [ fig : ivsb_ivsl ] . further above this
we find the upper limb of g134.4 + 3.85 .
we postulate that this pattern is the result of a series of star forming epochs in the region .
detailed theoretical modeling is beyond the scope of this paper , but is needed to test whether this postulate is physically reasonable .
dts97 state that the o - type stars in ocl 352 have a total output of @xmath103 photons / s .
terebey et al .
( 2003 ) conclude that 40% @xmath28 16 % ( @xmath104 photons / s ) of this ionizing radiation escapes to heights above 60 pc .
they suggest that these photons are available to ionize the diffuse gas and the upper galactic halo .
our observations suggest that some of these photons are ionizing the gas in the upper portion of the shell .
dts97 predict that the surface brightness of the upper portion of the shell should be @xmath420 r , which is well in excess of the 0.7 to 6.7 r that is observed .
they propose that two small molecular clouds located a little above the w4 loop are `` shadowing '' the upper portion of the shell , accounting for the low observed values of the emission measure .
in contrast , terebey et al .
( 2003 ) find that fewer photons than dts97 predict are absorbed by the molecular clouds .
a schematic diagram summarizing this picture is provided in figure [ fig : ion_schema ] .
we adopt the terebey et al .
picture that 40% of the ionizing photons emitted by ocl 352 are not absorbed by the w4 loop . in light of the and continuum data presented here ,
which are reminiscent of the ionization modeling by bjm99 , we propose that some of these photons are ionizing g134.4 + 3.85 while others escape through the breaks in g134.4 + 3.85 s walls , and yet others simply pass through the thin , fully - ionized upper portion of the shell .
the average value of the observed surface brightness for the four points , h through k ( see figure [ fig : contours ] ) is 4 r. if we compare this to the predicted 20 r ( dts97 ) we find that @xmath420% ( @xmath105 photons / s ) of the photons escaping from the lower loop ionize g134.4 + 3.85 and the remaining @xmath480% ( @xmath106 photons / s ) escape . in summary ,
the cluster produces of total flux of @xmath107 photons / s , of which 40% ( @xmath104 photons / s ) escape the w4 loop according to terebey et al .
( 2003 ) . of the photons escaping the w4 loop , @xmath420% ( @xmath108 photons / s ) are needed to produce the observed ionization in the upper shell of g134.4 + 3.85 ; this is 8% of the total flux .
the amount of this ionizing flux is less than the uncertainty in the estimate of the number of photons escaping from the loop .
this helps emphasize that the majority of these loop escaping photons may leave g134.4 + 3.85 through the breaks and through the thin , fully - ionized upper shell .
reynolds et al .
( 2001 ) estimate that the total hydrogen ionization rate for the large loop observed in the wham data is @xmath109 photons / s , or @xmath41.3% of ocl 352 s total output .
however , given that the loop subtends @xmath40.4 sr as seen from the cluster , approximately 40% of the cluster s luminosity ( @xmath110 photons / s ) must escape if such escape is isotropic .
this is reasonably consistent with our estimate , and we conclude that @xmath111 photons / s ) escape into the halo .
compelling evidence has been presented for the existence of a fragmented superbubble that is in the process of evolving into a chimney .
based on low latitude data , the structure , named here g134.4 + 3.85 , was originally postulated to be a fully - formed chimney .
later , data suggested that it was closed near @xmath112 .
the radio continuum data presented in this paper show a structure similar to that seen in , though two breaks , one at the top and one in the low longitude wall , are apparent in the combined data .
polarized intensity images , showing depolarization extending from the w4 region up the walls of the superbubble , confirm that the observed continuum emission and emission are at the same distance as the w4 region .
in contrast , the data make it clear that there is no corresponding neutral shell .
we speculate that one of the breaks could be associated with the thinning of the ionized shell at a magnetic pole .
however the 2 breaks in the ionized skin of the bubble , and the lack of a closed neutral hi shell , are similar to features in computational simulations ( in the literature ) of superbubbles which are evolving into chimneys . in these simulations
fragmentation can be caused when magnetohydrodynamic instabilities mix the magnetized shell with the ambient medium .
the fraction of ionizing photons emitted by ocl 352 that are not absorbed by the w4 loop is more than sufficient to account for the continuum emission from g134.4 + 3.85 .
the excess can escape both through the two breaks observed in the shell as well as through the thin , fully ionized upper portion of the shell .
g134.4 + 3.85 is @xmath4165 pc wide and extends @xmath4240 pc above the mid - plane of the galaxy , reaching beyond 135 pc which is the one-@xmath19 height of the global gaussian distribution .
while this is not into the galactic halo , the photons from ocl 352 can contribute to the ionization of the reynolds layer of the galaxy and are likely the main source of ionization of the large loop seen in the wham data .
this work was supported by the national science and engineering research council of canada .
the dominion radio astrophysical observatory synthesis telescope is a national facility operated by the national research council of canada .
the canadian galactic plane survey is a canadian project with international partners , and is supported by a grant from the natural sciences and engineering research council of canada .
data from the canadian galactic plane survey are publicly available through the facilities of the canadian astronomy data centre ( http://cadc.hia.nrc.ca ) .
the simbad database is operated at the centre de donnes astronomiques de strasbourg ( cds ) , strasbourg , france .
abt , h.a . ,
levy , s.g . ,
gandet , t.l .
1972 , , 77 , 138 basu , s. , johnstone , d. , martin , p.g .
1999 , , 516 , 843 beck , r. 2001 , space sci .
, 99 , 243 brown , j - a . c. 2002 , ph .
d. thesis , university of calgary , calgary , canada brand , j. , blitz , l. 1993 , , 275 , 67 carpenter , j.m . ,
heyer , m. , snell , r.l .
2000 , , 130 , 381 de avillez , m.a . , berry , d.l .
2001 , , 328 , 708 de avillez , m. a. , breitschwerdt , d. 2005 , , 436 , 585 dennison , b. , topasna , g.a . , simonetti , j.h .
1997 , , 474 , l31 ferrara , a. & tolstoy , e. 2000 , , 313 , 291 ferrire , k.m . ,
mac low , m .- m . , zweibel , e.g. 1991 , , 375 , 239 fehrenbach , c. , duflot , m. , genty , v. , amieux , g. 1996 , bicds , 48 , 11 gray , a.d . , landecker , t.l .
, dewdney , p.e . ,
willis , a.g . , normandeau , m. 1999 , , 514 , 221 haslam , c.g.t . , salter , c.j . , stoffel , h. , wilson , w.e .
1982 , , 47 , 1 kolmjenovic , p.t . ,
basu , s. , johnstone , d. 1999 , in asp conf .
168 _ new perspectives on the interstellar medium _ , ed .
taylor , t.l .
landecker , g. joncas ( san francisco : asp ) , 299 lada , c.j . ,
elmegreen , b.g . , cong , h - i , thaddeus , p. 1978 , , 226 , l39 landecker , t.l . , dewdney , p.e . ,
burgess , t.a .
, gray , a.d . ,
higgs , l.a .
, hoffmann , a.p . , hovey , g.j . ,
karpa , d.r . ,
lacey , j.d . , prowse , n. , purton , c.r . ,
roger , r.s . , willis , a.g . ,
wyslouzil , w. , routledge , d. , vaneldik , j.f .
2000 , , 145 , 509 lockman , f.j . , hobbs , l.m . , shull , j.m .
1986 , , 301 , 380 maciejewski , w. , murphy , e.m . ,
lockman , f.j . , savage , b.d .
1996 , , 469 , 238 mcclure - griffiths , n.m . , dickey , j.m .
, gaensler , b.m . , green , a.j .
2003 , , 594 , 833 maclow , m .-
m . , mccray , r. , norman , m.l .
1989 , , 337 , 141 norman , c.a . , ikeuchi , s. 1989 , , 345 , 372 normandeau , m. , taylor , a.r .
, dewdney , p.e .
1996 , , 380 , 687 normandeau , m. , taylor , a.r . , dewdney , p.e .
1997 , , 108 , 279 oey , m.s . ,
watson , a.m. , kern , k. , walth , g.l .
2005 , , 129 , 393 reich , p. , reich , w. 1986 , 63 , 205 reich , p. , reich , w. , frst , e. 1997 , , 126 , 413 reich , w. 1982 , , 48 , 219 reich , w. , reich , p. , frst , e. 1990 , , 83 , 539 reynolds , r.j . , sterling , n.c . , haffner , l.m .
2001 , , 558 , l101 taylor , a.r . ,
gibson , s.j . ,
peracaula , m. , martin , p.g . ,
landecker , t.l . ,
brunt , c.m . , dewdney , p.e . ,
dougherty , s.m . ,
gray , a.d .
, higgs , l.a .
, kerton , c.r . ,
knee , l.b.g .
, kothes , r. , purton , c.r . , uyaniker , b. , wallace , b.j . ,
willis , a.g . , and durand , d. 2003 , , 125 , 3145 taylor , j.h . , & cordes , j.m .
1993 , , 411 , 674 tenorio - tagle , g. , bodenheimer , p. 1988
, ara&a , 26 , 145 terebey , s. , fich , m. , taylor , a.r . ,
cao , y. , hancock , t. 2003 , , 590 , 906 thronson , h.a .
, campbell , m.f . , hoffmann , w.f .
1980 , , 239 , 533 tomisaka , k. 1998 , , 298 , 797 west , j. l. 2003 , `` w4 revisited : a chimney candidate in the milky way galaxy explored using radio continuum and polarization observations '' , university of manitoba , masters thesis .
willis , a. g. 1999 , , 136 , 603 | new observations at high latitudes above the region w4 show that the structure formerly identified as a chimney candidate , an opening to the galactic halo , is instead a superbubble in the process of fragmenting and possibly evolving into a chimney .
data at high galactic latitudes ( @xmath0 ) above the w3/w4 star forming region at 1420 and 408 mhz stokes i ( total power ) and 1420 mhz stokes q and u ( linear polarization ) reveal an egg - shaped structure with morphological correlations between our data and the data of dennison , topasna , & simonetti .
polarized intensity images show depolarization extending from w4 up the walls of the superbubble , providing strong evidence that the radio continuum is generated by thermal emission coincident with the emission regions .
we conclude that the parts of the region hitherto known as w4 and the newly revealed thermal emission are all ionized by the open cluster ocl 352 . assuming a distance of 2.35 kpc , the ovoid structure is 164 pc wide and extends 246 pc above the mid - plane of the galaxy .
the shell s emission decreases in total - intensity and polarized intensity in various locations , appearing to have a break at its top and another on one side . using a geometric analysis of the depolarization in the shell s walls
, we estimate that a magnetic field line - of - sight component of 3 to 5 @xmath1 exists in the shell .
we explore the connection between w4 and the galactic halo , considering whether sufficient radiation can escape from the fragmenting superbubble to ionize the kpc - scale loop discovered by reynolds , sterling & haffner . |
one of the interesting phenomenon in the vortex system of high temperature superconductors ( hts ) is the second - peak ( sp ) effect , which represents an anomalous increase of critical current with the increase of magnetic field . for the strong layered superconductor bi-2212 ,
it has already been pointed out that the sp may originate from the phase transition between a quasi - ordered bragg glass ( bg ) and a disordered vortex phase@xcite .
the existence of bg has been supported by neutron scatting experiments@xcite , but it is a long standing question about what is the nature of the disordered vortex phase in high field region .
some literatures@xcite reported that a 3d vortex glass ( vg ) exists above the second peak field and the flux dynamics is elastic with a diverging barrier in the small current limit .
the vortex glass is assumed to melt into vortex liquid at a temperature @xmath8@xcite , around which a scaling behavior of e - j curves is given . on the other hand
it was predicted from numerical simulation that a plastic deformation will occur via the proliferation of vortex defects in high field region@xcite .
similarly vortex dislocations are supposed to exist above the sp field@xcite , and kierfeld _ et al . _
concluded that the disordered vortex phase and vortex liquid will be thermodynamically indistinguishable after the critical endpoint of the first - order melting ( fot ) line@xcite .
experimentally miu _
et al._@xcite claimed the existence of plastic vortex creep above the second magnetization peak based on the data measured at a single temperature t = 25 k. it remains , however , unclear whether this plastic phase appears in the high field region for all temperatures . in this paper
we will show clear evidence that the flux motion above the sp field is dominated by plastic motion with a finite barrier in the small current limit .
several bi-2212 single crystals with typical size of @xmath9 and thickness of @xmath10 were used in the experiment .
all of them are optimal doped or slightly overdoped with @xmath11 and @xmath12 determined in resistive measurement .
the samples were measured by a vibrating sample magnetometer ( vsm 8 t , oxford 3001 ) and a superconducting quantum interfere device ( squid , quantum design , mpms5.5 ) . in all measurements the magnetic field h was applied parallel to the c - axis and the magnetization data were collected in the field descending process in order to eliminate the surface barrier . each sample shows similar results , therefore we will not distinguish them for convenience .
fig . [ fig1 ] shows the relationship between the width of the magnetization - hysteresis - loops ( mhl ) @xmath13 and @xmath14 in double - logarithmic coordinates .
an interesting feature is that each curve has a dip , as indicated by the dotted line in fig . 1 . at t = 4 , 5 , 6 , 8 , 10 , 12 , 14,16 , 18 , 20 , 22 , 24 , 26 k from top to bottom .
the dotted line indicates the dip of each curve .
the sp is very clear as pointed by the arrow at 22k . ]
the dip field decreases with the increase of temperature and finally drops to the valley field associated with the sp at 22 k and keeps stable at higher temperatures .
recently we have pointed out that the sp field reappears and is independent of temperature well below 20 k@xcite , thus the dips at low temperatures must have no direct relation with the sp in respect that the dip fields ( @xmath15 oe when @xmath16 k ) are much higher than the sp field ( about 280 oe for this sample ) . and
q at t = 10 k. four regions corresponding to different vortex dynamical properties are marked by i , ii , iii and iv .
the inset in ( b ) gives the @xmath17 curves at 17 k measured with different field sweep rates : 100 , 50 , 20 , 10 , 5 , 1 , 0.3 oe / s .
it is clear that the dip has a dynamical feature and merges to the valley position of sp at higher temperatures . ] .
[ fig2 ] in order to find out the underlying physics corresponding to the dip on @xmath18 , we have measured the dynamical relaxation rate defined as @xmath19 by using different field sweep rates@xcite .
it has been proved that @xmath20 determined in the dynamical relaxation process is equivalent to the conventional relaxation rate defined as @xmath21 .
the results are given in fig .
[ fig2](a ) . just like the dip on each @xmath18 curve
, each @xmath22 curve shows a peak . for having a deeper insight to these features , we plot @xmath20 and @xmath13 vs. @xmath14 at 10 k together in fig . [
fig2](b ) .
it is found that each @xmath23 curve can be divided into four parts marked as i , ii , iii and iv . in region
i , both @xmath13 and @xmath20 are flat , i.e. , independent of the magnetic field . in region ii , @xmath20 increases monotonously with the increase of magnetic field , and @xmath13 begins to drop .
while in region iii , @xmath20 decreases with the increase of magnetic field , and the dropping rate of @xmath13 becomes slower and thus results in the dip in mhls .
therefore , the dip on mhl is directly related to the peak on the @xmath23 curve . the inset of fig .
[ fig2](b ) gives the @xmath24 curves of another sample at 17 k and different field sweep rates .
it is clear that the dip field moves to lower value when the field sweep rate decreases ( corresponding to smaller current ) and finally merges to the valley field of sp .
hence both of the dip on @xmath18 and the peak on @xmath22 have dynamical features and should disappear above the sp field in the small current limit in low temperature region . after decreasing in region iii , @xmath20 upturns and increases monotonously with the field up to the irreversibility field @xmath25 ( @xmath26@xcite)in region iv . due to the dynamical feature of the dip and peak mentioned above
, it can be expected that the region iv will dominate the whole @xmath27 phase diagram above the sp field in long time or small current limit . after checking about 10 pieces of bi-2212 single crystals
, it is assured that the dip and peak feature appears in all samples without one exception .
in fact , cohen _ et al . _ have observed the same dip and peak phenomena in bi-2212 .
they related the peak in @xmath20 to the penetration field@xcite . since the peak can still be found if we consider only the descending branch of mhls , which strongly indicates that there is no relation between the peak and the penetration field . for bi-2212 single crystals
, it was found that the geometrical barrier plays an important role on the magnetization in the field ascending process@xcite .
the total irreversible magnetic moment of the sample is contributed by both the surface shielding current from the geometrical barrier and the bulk screening current in the field ascending process when the external field is not high . in this case local magnetic measurement using either the tiny hall probe@xcite or magneto - optical technique@xcite were suggested . while when the external field is much higher than the second peak field , the contribution from the bulk current is much larger than that from the geometrical barrier and a quasi - bean critical state model
is assumed to be valid@xcite .
this is essentially true in the field descending process when h is decreased from a high field to a field which is still higher than the second peak field .
no superposition of the magnetic moment due to the bulk current and the surface shielding current are expected here . as mentioned before our magnetization relaxation measurement was conducted in the field descending process with the external field above the second peak field .
therefore it is justified for us to derive the bulk screening current density @xmath28 from the magnetization .
it has already been pointed out that the @xmath29 curve can be obtained by assuming @xmath30 and @xmath31@xcite .
[ fig3](a ) shows the @xmath29 curves at 1 t based on this method .
all the curves above 12 k are obvious concave , suggesting a finite linear resistance .
worthy of noting is that the irreversibility temperature @xmath32 at 1 t is about 30 k. this is a strong evidence that the vortex motion in region iv is not elastic . in order to check whether the vortex glass exists below 12 k ,
a naive idea is to treat 12 k as the vortex glass melting temperature @xmath8 and scale the @xmath29 curves using 3d or quasi-2d glass theory@xcite .
the scaled curve looks not good and the critical exponents are unreasonably large , e.g. , @xmath33 , which is commonly believed to be less than 2@xcite .
hence we fit it to the 2d vg theory@xcite , which predicts @xmath34 . according to 2d vg theory
the correlation length @xmath35 diverges at @xmath36 k , @xmath37 , where @xmath38 and @xmath39 are the 2d vg exponent and the spacing of the vortex lattice respectively ; @xmath40 is the core energy of a vortex segment of length d. the linear resistivity @xmath41 due to thermal activation is given by @xmath42 , where @xmath43 is a characteristic temperature of the same order of magnitude as @xmath40 , and @xmath44 .
therefore we can scale the @xmath29 data as follows@xcite , @xmath45 where g is an universal scaling function .
the scaled result with @xmath46 , @xmath47 , @xmath48 is given in fig.[fig3](b ) .
it is interesting to note that the values found here are very close to those found for very thin @xmath49 films and @xmath50 thin films at a high magnetic field.@xcite above 12 k , all data collapse onto one branch , while the deviation begins at about 12 k. this deviation accords with the behavior of @xmath4 ( shown below ) which turns from negative to positive around 12 k. in fact , we can find the same result if scaling the data based on plastic creep theory@xcite .
this is not strange if we recall that the plastic creep theory assumes also a correlation length @xmath51 diverging at 0 k , @xmath52 , and a linear resistivity when @xmath28 approaches zero , @xmath53 , where @xmath54 is the plastic pinning barrier@xcite .
therefore , we conclude that the flux motion in region iv is plastic .
this indicates that the vortex glass phase which is characterized by a diverging activation barrier in the small current limit may not exist in region iv .
actually , miu _ et al .
_ have also found the plastic vortex creep above sp , but their claim was made for only a single temperature ( t = 25 k ) .
curves determined from the relaxation data at h = 1 t and t = 18 , 16 , 14 , 12 k from left to right .
the @xmath32 at 1 t is about 30 k. the linear lines are drawn to help identifying the curvature of these curves .
( b ) scaling the @xmath29 data at h = 5 koe and t from 5 to 20 k step 0.5 k according to 2d vg or plastic theory with @xmath46 , @xmath47 , @xmath55 .
the scaling fails below 12 k , as indicated by the arrow .
the inset of ( b ) gives the u(j ) dependence determined by the maley s method . ] and the exponent @xmath4 at ( a ) @xmath56 = 12 k and @xmath14 = 10 to 10000 oe , ( b ) @xmath14 = 500 oe and @xmath56 = 5 to 25 k for another sample@xcite .
the linear approach in low temperature region as shown by the arrow in ( b ) suggests that there is weak effect of quantum tunnelling creep . ]
an important parameter characterizing the vortex dynamics is the exponent @xmath4 , which can be derived from the conventional magnetic relaxation measurement .
et al._@xcite assumed a general pinning barrier @xmath57 $ ] from which thompson _ et al .
_ gave an interpolation formula to obtain @xmath4@xcite , @xmath58 where @xmath59 and @xmath60 can be treated as constants approximately .
to obtain @xmath4 values we have measured @xmath61 vs. time by squid and the calculated results at @xmath62k and various fields are given in fig .
[ fig4](a ) , in which the conventional magnetic relaxation rate @xmath21 is also given . to avoid the geometrical barrier in the field penetration process ,
each measurement was done as following : the field is raised to 5.5 t and then lowered to the expected field .
it is clear that @xmath4 is negative in region iv .
a negative @xmath4 indicates a finite @xmath63 when @xmath28 approaches to zero , thus conflicts with the vortex glass theory .
the @xmath64 and @xmath65 curves have very similar feature as @xmath66 and @xmath67 , as shown in fig . [ fig4](b ) for another sample with @xmath68 around 380 oe@xcite .
the negative @xmath4 values found here corroborate the picture that the vortex motion in region iv is plastic .
next we try to clarify the origin of a positive @xmath4 value and the drop of @xmath20 with increasing field in region iii .
et al . _ have measured the ac susceptibility of @xmath69 and found the elastic ( el ) creep in some part of the disordered phase@xcite , which is explained as @xmath70 , where @xmath71 is the dimension of the creeping vortex bundle and @xmath72 is the mean distance between two dislocations .
therefore , we attribute the vortex motion in region iii to the 2d el too . it was predicted that @xmath73 above the sp field@xcite , @xmath74 where @xmath75 is the critical current separating the 2d el creep and the variable - range - hopping regions ( vrh ) .
assuming @xmath76 , taking @xmath77 , we have @xmath78 when @xmath79 .
it is thus tempting to relate the positive @xmath4 value as an evidence of 2d el creep .
these regions are marked on the u(j ) curves respecting to different current values in the insert of fig.3(b ) . by further lowering t and/or h ( increasing j ) the optimal hopping distance @xmath80 will be smaller than @xmath81 eventually , and the vrh process will take place , leading to the fast increase of @xmath82 with h and t in region ii .
it is interesting to mention that in @xmath49 crystals with columnar defects thompson _
et al._@xcite found the similar peak on s(t ) curve and dip on @xmath83 curve , which was also interpreted as the crossover from elastic creep to vrh .
it is yet , however , to be understood why the @xmath4 values determined in region i and ii are negative .
there are two possibilities for the negative @xmath4 values in regions i and ii : ( 1)the total magnetic moment in the time window we used varies too little for the program to fit and get a valid and physical meaningful @xmath4 value ( for region ii ) ; ( 2)a superposition of the bulk current and the geometrical surface current may occur in region i. although all these hypotheses about the dynamical processes in regions i , ii and iii need further clarification , it seems quite certain that the region iv has a feature of plastic creep ( pl ) and it will become dominant above the sp field throughout whole temperature region in the small current limit ( shown by the inset of fig.2(b ) ) .
[ fig5 ] gives the phase diagram of bi-2212 .
el and vrh region will gradually reduce when @xmath28 decreases , and the whole region ( shown by the shaded area ) above sp field is dominated by plastic creep ultimately with @xmath84 .
is determined from the deviating point on the m(t ) curves measured in the zfc - fc processes .
the @xmath68 is taken from our earlier data@xcite and shown as the square symbol .
the boundaries between pl , el and vrh are determined from the peak and valley positions of @xmath20 in fig .
2(a ) . in small current limit ,
the el and vrh region will disappear and the whole region above sp field becomes plastic . ]
in conclusion , we have found a general dip on @xmath85 and @xmath86 curves and peak on @xmath2 and @xmath3 curves .
they have dynamical features and are explained as the crossovers between different vortex phases .
plastic flux motion is found to be dominant far below the irreversibility line and takes over the whole region above the sp field in the small current limit .
this work is supported by the national science foundation of china ( nsfc 19825111 , 10274097 ) , the ministry of science and technology of china ( project : nkbrsf - g1999064602 ) , the knowledge innovation project of chinese academy of sciences .
we are grateful for fruitful discussions with dr .
sean ling in brown university , rhode island , and dr .
thiery giamarchi in university of geneva .
00 t. giamarchi , p. le doussal , phys .
b * 55 * ( 1997 ) 6577 ; phys .
rev . lett .
* 72 * ( 1994 ) 1530 .
r. cubitt , _ et al .
_ , nature * 365 * ( 1993 ) 407 ; t. klein , _ et al .
_ , nature * 413 * ( 2001 ) 404 .
h. safar , _ et al .
68 ( 1992 ) 2672 .
m. p. a. fisher , phys .
* 62 * ( 1989 ) 1415 .
d. s. fisher , m. p. a. fisher , d. a. huse , phys .
b * 43 * ( 1991 ) 130 ; d. a. huse , d. s. fisher , m. p. a. fisher , nature * 358 * ( 1992 ) 553 . for a review see g. blatter _ et al .
* 66 * ( 1994 ) 1125 .
r. h. koch _
et al . _ ,
phys . rev .
* 63 * ( 1989 ) 1551 .
s. ryu , a. kapitulnik , s. doniach , phys .
* 77 * ( 1996 ) 2300 .
j. kierfeld , physica c * 300 * ( 1998 ) 171 .
j. kierfeld , v. vinokur , phys .
b * 61 * ( 2000 ) r14928
. s. l. li , h. h. wen , phys .
b 65 ( 2002 ) 214515 .
m. jirsa,_et al . _ , physica * c207 * ( 1993 ) 85 .
h. h. wen , _
et al . _ ,
physica * c241 * ( 1995 ) 353 .
h. h. wen , _
et al . _ ,
* 79 * ( 1997 ) 1559 .
l. f. cohen , _ et al .
_ , supercond .
* 10 * ( 1997 ) 195 .
e. zeldov , _ et al .
lett . * 73*(1994 ) 1428 .
n. morozov , _ et al .
_ , physica * c 291 * ( 1997 ) 113 .
e. zeldov , _ et al .
_ , europhys . lett . * 30 * ( 1995 ) 367 .
b. khaykovich , phys .
* 76 * ( 1996 ) 2555 . c. j. van der beek , phys .
* 84 * ( 2000 ) 4196 .
l. miu , _ et al .
_ , physica c * 334*(2000)1 .
a. m. malozemoff , physica * c 185 - 189 * ( 1991 ) 264 .
j. r. thompson , _ et al .
b * 49 * ( 1994 ) 13287 .
g. ries , _
et al . _ ,
technol . * 5 * ( 1992 ) 81 .
h. yamasaki , et al . ,
phys . rev . *
b 50 * ( 1994 ) 12959 .
m. p. a. fisher , t. a. tokuyasu _
et al . _ ,
* 66 * ( 1991 ) 2931 .
c. dekker , _ et al .
* 69 * ( 1992 ) 2717 ; h. h. wen , _ et al . _ , phys . rev . lett .
* 80 * ( 1998 ) 3859 .
c. reichhardt , a. van otterlo , g. t. zimnyi , phys . rev .
lett * 84 * ( 2000 ) 1994 .
v. m. vinokur , _ et al .
* 65 * ( 1990 ) 259 .
i. joumard , t. klein , j. marcus , phys .
* 87 * ( 2001 ) 167002 .
m. p. maley , _ et al .
b * 42 * ( 1990 ) 2639 .
j. r. thompson , _ et al .
* 78 * ( 1997 ) 3181 . | in magnetic measurements on bi@xmath0sr@xmath0cacu@xmath0o@xmath1 ( bi-2212 ) single crystals , a general peak with a dynamical feature on both @xmath2 and @xmath3 curves was found with s the magnetic relaxation rate . at higher fields ,
the characteristic exponent @xmath4 becomes negative , together with the positive curvature of @xmath5 vs. @xmath6 and the scaling based on the 2d vortex glass theory or plastic creep theory , we conclude that the vortex motion above the second peak is plastic when @xmath7 and there is no vortex glass phase at finite temperatures in bi-2212 .
the peak of s is then explained as the crossover between different meta - stable vortex states .
vortex glass , vortex phase diagram , bragg glass 74.60.ge , 74.60.ec , 74.72.hs |
in 1978 gribov showed that the fixing of the divergence of the gauge field in non - abelian gauge theory does not uniquely fix its gauge .
the restriction of the gauge field such that its faddeev - popov determinant is positive ( gribov region ) is necessary yet insufficient for excluding the ambiguity .
the restriction to the fundamantal modular region , i.e. minimizing the norm of the gauge field , cancels the infrared singularity of the perturbation theory and makes the gluon to have the complex mass .
@xmath9 @xmath10 where @xmath11 is the dimension of the adjoint representation of the su(n ) group .
in order to measure the propagator on a lattice , it is necessary to avoid the gribov copy .
we confirmed that it is possible by the smeared gauge fixing@xcite , when @xmath8 in the case of @xmath12 .
on the lattice we define an @xmath13 lattice action for @xmath14 in the pseudo - fermion scheme .
we measure @xmath15 where @xmath16 is the fourier component of @xmath17 , and @xmath18 is summed over transverse polarization with respect to * k * .
the gauge field on links is defined as @xmath19 the gauge transformation is @xmath20 where @xmath21 and @xmath22 is a traceless antihermitian the connected part of the propagator is defined by the normalized fourier transform @xmath23 as @xmath24 where @xmath25 specifies the expectation value of the gauge field in the landau gauge . for this evaluation
it is necessary to fix the global gauge transformation .
we fix the global gauge such that the average over t of the zero - mode of a sample : @xmath26 where @xmath27 is the su(3 ) gell - mann matrix , is diagonalized and the smallest eigenvalue appears at the top and the largest eigenvalue at the bottom .
we observed that the 2nd term of @xmath28 is almost independent of t and that the value is several % of the 1st term of @xmath29 .
the value is larger for the unquenched case as compared to the quenched case .
the fourier transform of @xmath30 in the unquenched @xmath12 and in the quenched @xmath31 lattice simulation @xcite suggest the suppression of the infrared part of the gluon propagator . .
ordinates are the euclidean 4-momentum @xmath32 in the unit of @xmath33 .
@xmath34(left ) and @xmath35(right ) . @xmath12
lattice.,title="fig : " ] .
ordinates are the euclidean 4-momentum @xmath32 in the unit of @xmath33 .
@xmath34(left ) and @xmath35(right ) . @xmath12 lattice.,title="fig : " ] in the stingl s factorizing - denominator rational approximants ( fdra ) method@xcite , the transverse gluon propagator is expressed in the 2nd order as @xmath36}={\rho\over p^2+u_+\lambda^2}+{\tau\over p^2+v_+\lambda^2}+c.c.\ ] ] we compared the lattice fourier transform of the above expression with @xmath37 and observed that in the confinement region there appears a pair of complex zero near @xmath38 as is suggested by the eq ( 2 ) .
in the landau gauge , the transverse gluon propagator in the infrared region is finite at @xmath39 @xciteand the position of the zero of @xmath40 approaches @xmath39 as the lattice size increases , while the propagator of the faddeev - popov ghost which is measured in the quenched @xmath31 lattice shows the infrared singularity .
these results support the possibility of a realization of the confinement mechanism of the gribov - zwanziger theory .
this work was supported by high enery accelerator research organization , as kek supercomputer project ( project no.98 - 34 ) .
gribov , nucl . phys . *
b139 * , 1 ( 1978 ) ; d. zwanziger , nucl . phys . *
b364 * , 127 ( 1991 ) .
h. nakajima and s. furui , nucl phys .
* b * ( proc . suppl . ) * 63 * , 976 ( 1998 ) , hep - lat/9710028 .
j.e . hetrick and ph .
de forcrand , nucl .
phys . * b*(proc . suppl.)*63 * , 838 ( 1998 ) .
j. fromm and m. stingl , preprint , mnster university(1997 ) ; m. stingl , z. phys .
* a353 * , 423 ( 1996 ) , hep - th/9502157 h. nakajima and s. furui , lattice98 proceedings ( 1998 ) a. cucchieri , phys . lett .
* b422 * 233 ( 1998),hep - lat/9709015 .
leinweber , j.i skullerud , a.g .
williams and c. parrinello , phys . rev
. * d58 * ( 1998 ) , hep - lat/9803015 | we measured the gluon propagator in the landau gauge fixed qcd langevin simulation@xcite and studied the infra - red behaviour of the gluon propagator .
the @xmath0 lattice siulation was done for quenched @xmath1 and @xmath2 and unquenched @xmath3 and @xmath4 , using each 100 independent samples .
the landau gauge fixing was done by an extension of the fourier acceleration method with the condition @xmath5 , and the field a is related to the link variable by @xmath6 instead of the usual u - linear definition .
we confirmed gauge fixing with smearing preconditioning@xcite works perfectly for the purpose of finding the global minimum of the squared norm of the gauge field when @xmath7 is large ( e.g. @xmath8 ) .
our simulation results suggests the possibility of a realization of the infra - red behaviour of the gribov - zwanziger theory . |
the australian square kilometre array pathfinder ( askap ; @xcite,@xcite ) , represents a significant advance in radio telescope design .
this facility will combine high resolution imaging ( through the use of a 36-element aperture synthesis array with a maximum baseline of @xmath0 km ) with a wide field of view ( achieved with innovative focal plane array technology ) at frequencies between 700 mhz and 1.8 ghz .
installation of the first askap antenna at the murchison radio observatory site , western australia , occured in early 2010 , and the 6-antenna beta test array will operate from september 2011-march 2013 .
it is anticipated that full science operations will be underway by 2014 .
processing and data transport requirements for askap are described in @xcite , and @xcite provides an overview of the data infrastructure requirements .
wallaby @xcite is one of ten askap survey science projects currently in the design study phase .
wallaby aims to significantly enhance our understanding of the extragalactic neutral hydrogen ( hi ) universe .
the survey will cover 75% of the sky , detecting @xmath1 million galaxies to redshift @xmath2 ( lookback time @xmath3 gyr ) .
key science outcomes are studies of galaxy formation and the missing satellite problem in the local group , evolution and star formation in galaxies , mergers and interactions , determination of the hi mass function and its variation with galaxy density , and the nature of the cosmic web . unlike previous hi surveys
, it will not be feasible to keep all of the raw data ( i.e. fourier visibilities ) from askap observations for subsequent reprocessing . instead
, pipeline - preprocessed spectral data cubes will be provided for analysis .
each wallaby spectral cube is anticipated to comprise 6144 @xmath4 6144 spatial pixels and 16,384 spectral channels ( i.e. @xmath5 gigavoxels or volume elements in total ) , requiring 2.5 terabytes ( tb ) of storage .
a total of 1200 cubes will be required to cover the sky south of declination @xmath6 .
likely additional outputs are integrated ( moment ) maps , continuum images , sub - cubes ( individual objects or scaled versions of larger datasets ) , and full parameterisation of all galaxies , resulting in several petabytes of data products .
such data volumes pose considerable challenges for the existing work practices of astronomers . indeed , visualisation and analysis ( hereafter
, `` v+a '' ) of wallaby data products will require both evolutionary and revolutionary changes to existing software and hardware , with a likely move away from desktop - only solutions , and a greater reliance on remote services .
a brief overview of the wallaby workflow from data collection to catalogue is as follows : 1 .
observe field .
2 . generate spectral data cube from visibilities .
visualise cube as quality control prior to deletion of raw data .
transfer preprocessed data cube to archive .
perform source finding on data cube .
fit models to candidates and perform related quantitative analysis .
. add parameterised candidates to catalogue .
apart from personnel , the main resource for completion of all of these stages is access to appropriate computing infrastructure ( hardware and software ) . as a framework within which to assess the practicalities of achieving each step in the wallaby workflow , we begin ( section ii ) by considering desktop and high performance computing ( hpc ) cluster resources available and used by astronomers today , and project these forward to configurations available by 2014 . in section iii , we present five challenges that v+a of wallaby data products will face in the likely computing environment .
we consider tasks that can be done essentially the same way they are now , and those requiring an investment in new technology or the development of new software , in order to deal with data sets orders of magnitude larger than previous extragalactic hi survey projects .
we make our concluding remarks in section iv . throughout , we make comparisons with the existing hi parkes all sky survey ( hipass ; @xcite ) , conducted with the parkes multibeam receiver @xcite . the southern catalogue , hicat ( @xmath7 ; @xcite ) , comprised 4315 galaxies , and the northern extension , nhicat ( @xmath8 ; @xcite ) , a further 1002 sources .
russell jurek ( australia telescope national facility ; atnf ) has combined the 388 individual southern sky data cubes into a single all - sky cube with @xmath9 gigavoxels , and a file size of 12 gb .
the configurations of ( typical ) desktop and hpc resources available to astronomers are fundamental to the capacity of existing or new software to enable each stage of the wallaby workflow . in table
i , we present hardware parameters for today s mid - range desktop computer . using typical growth rates in the computing industry ( e.g. moore s law ; kryder s law @xcite ) , we extrapolate to 2014 ( i.e. `` tomorrow '' ) .
quoted processing speeds are theoretical ( i.e. peak ) , single precision values ; these assume 100@xmath10 efficient algorithms using all available processing cores / streams .
table ii presents similar per - node comparisons for cluster - based hpc configurations .
other hpc configurations are possible , but we restrict our discussion to facilities similar to the swinburne `` green machine '' supercomputer @xcite , with which we are most familiar .
most specifications and capabilities of desktop and hpc compute platforms will simply evolve and grow as they have done over the past decades .
two significant revolutions in compute capability of processors , however , are currently underway ( e.g. @xcite ) : 1 .
central procesing units ( cpus ) are gaining increased compute capacity in the form of multiple cores , rather than increased clock speeds .
2 . graphics processing units ( gpus ) are boosting compute capacity by around 1050 times by functioning as streaming co - processors , at very low cost .
this `` concurrency '' revolution , based on the availability of high levels of parallelism on a single chip , requires major software work @xcite and a re - examination of algorithms so that scientists can benefit fully from this new processing paradigm ( see @xcite and @xcite for astronomy - related solutions ) .
while cpus are optimised for sequential programs thanks to sophisticated control logic and large memory caches ( to reduce instruction and data access latencies ) , gpus maximise the chip area for computation .
the advent of programming libraries such as cuda and opencl has enabled the use of gpus for general purpose computation , with the gpu acting as a computational co - processor .
typical ( single precision ) theoretical peaks are already over 1 teraflop / s for cards like the nvidia tesla c2050 .
some of the challenges we identify in this paper can only be solved with gpus .
.desktop computing for wallaby [ cols="<,<,<",options="header " , ]
perhaps the biggest challenge to planning strategies for visualisation and analysis is that no askap data exists yet .
we do not know what the exact imaging properties of askap will be .
although simulated data cubes are now being generated , until the full askap system undergoes commissioning , we will not fully understand all of the calibration , noise , interference , etc . issues that will arise with the relatively new technology of focal plane arrays .
testing source finders often includes injecting fake sources , with a given signal level , and then seeing how often they are recovered . with real wallaby data cubes unavailable until 2014 , progress in testing source finders will necessarily be limited . while we can do our best to plan source finders based on existing datasets , and early science data from the beta configuration ( september 2011-march 2013 )
, we may find that our techniques do not work adequately for the full dataset . by considering the various v+a tasks now , and identifying approaches based on new hardware and software that were not available or feasible for earlier surveys ,
we can hope to minimise the impact of the `` unknown unknowns '' of askap .
graphics processing units offer an intriguing solution to a number of the current desktop - bound problems .
table iv summarises our thoughts regarding the visualisation and analysis tasks that will require either an evolution of existing software and hardware , or a revolution in how they are approached . by planning today
, we aim to maximise the scientific return from wallaby tomorrow .
the authors would like to thank brbel koribalski ( atnf ) and russell jurek ( atnf ) for discussions relating to the wallaby project , and john taylor ( csiro ) and tim cornwell ( atnf ) for enabling access to the csiro gpu cluster .
t. j. cornwell and g. van diepen , `` scaling mount exaflop : from the pathfinders to the square kilometre array '' , 2010 [ online ] available : http://www.atnf.csiro.au/people/tim.cornwell/publications/mountexaflop .
/ people / tim.cornwell / publications / mountexaflop .
pdf ] p. quinn , `` data infrastructure needs for radio astronomy , '' 2010 [ online ] .
available : http://astronomyaustralia.org.au/pq-dataworkshop-whitepaper.pdf c. walter , `` kryder s law , '' scientific american , vol .
293 , july 2005 [ online ] .
available : http://www.scientificamerican.com/article.cfm?id=kryders-law swinburne supercomputer [ online ] .
available : http://astronomy.swin.edu.au/supercomputing h. sutter , `` the free lunch is over : a fundamental turn toward concurrency in software , '' dr .
dobb s journal , vol . 30 , issue 3 , march 2005 [ online ] .
available : http://www.gotw.ca/publications/concurrency-ddj.htm karma .
[ online ] .
available : http://www.atnf.csiro.au/computing/software/karma aips .
[ online ] .
available : http://www.aips.nrao.edu casa .
[ online ] .
available : http://casa.nrao.edu fits .
[ online ] .
available : http://fits.gsfc.nasa.gov netcdf .
[ online ] .
available : http://www.unidata.ucar.edu/software/netcdf hdf5 .
[ online ] .
available : http://www.hdfgroup.org/hdf5 optiportal .
[ online ] .
available : http://wiki.optiputer.net/optiportal a. h. hassan , c. j. fluke , and d. g. barnes , `` interactive visualization of the largest radioastronomy cubes , '' newa , vol .
16 , pp . 100109 , february 2011 .
[ online ] .
available : http://www.atnf.csiro.au/people/matthew.whiting/duchamp b. s. koribalski , `` the local volume hi survey ( lvhis ) , '' in galaxies in the local volume , astrophysics and space science proceedings , b.s.koribalski and h.jergen , eds . berlin : springer , 2008 , pp . 4144 .
f. walter , e. brinks , w. j. g. de blok , f. bigiel , r. c. kennicutt , m. d. thornley , and a. leroy , `` things : the hi nearby galaxy survey , '' aj , vol .
136 , pp . 25632647 , december 2008 .
sloan digital sky survey .
[ online ] .
available : http://www.sdss.org/background/data.html wigglez .
[ online ] .
available : https://wigglez.swin.edu.au millenium simulation .
[ online ] .
available : http://www.g-vo.org/millennium | visualisation and analysis of terabyte - scale datacubes , as will be produced with the australian square kilometre array pathfinder ( askap ) , will pose challenges for existing astronomy software and the work practices of astronomers . focusing on the proposed outcomes of wallaby ( widefield askap l - band legacy all - sky blind survey ) , and using lessons learnt from hipass ( hi parkes all sky survey ) , we identify issues that astronomers will face with wallaby data cubes .
we comment on potential research directions and possible solutions to these challenges .
computer aided analysis ; distributed computing ; radio astronomy ; visualisation . |
gravitational waves , the ripples in the fabric of space - time , were predicted by einstein s general theory of relativity .
although astronomical observations have inferred the existence of gravitational waves , they have yet to be detected directly .
the laser interferometer gravitational - wave observatory ( ligo ) is one of the large - scale gravitational - wave detectors currently being built worldwide .
the pre - stabilized laser ( psl ) subsystem is the light source for the ligo detector as shown in figure [ psl - position ] .
the output of the psl is modematched into the suspended modecleaner before being coupled into the ligo interferometer . the term _ pre - stabilized _ is used because the laser undergoes two stages of stabilization prior to being injected into the interferometer .
the 10-w laser used is configured as a master - oscillator - power - amplifier ( mopa ) , with a 700mw single - frequency , single - mode non - planar ring oscillator used as the master oscillator .
the control strategy uses the actuators of the master oscillator in order to stabilize the frequency .
power stabilization is achieved by control of the power amplifier output .
the psl topology is shown in figure [ psl - parts ] .
light from the laser is modematched into a high - throughput , ring fabry - perot cavity called the pre - modecleaner ( pmc ) .
the psl has a design requirement that the output be close to the shot - noise limit for 600mw of detected light at the interferometer modulation frequency of 25mhz .
as this is beyond the bandwidth of any electronics servo , it is done by passive filtering by the pmc . by appropriate choice of mirror reflectivity ,
the pmc acts as a tracking bandpass filter with a pole at the cavity half - bandwidth .
one of the pmc mirrors is epoxied to a piezoelectric transducer ( pzt ) to vary the length of the cavity .
the servo electronics constantly adjusts the pzt voltage in order to keep the incident light resonant with the cavity .
astrophysical models suggest that in order to plausibly detect candidate gravitational - wave sources , the ligo detector must achieve a displacement sensitivity of better than 10@xmath0@xmath1 at 100hz .
this corresponds to a frequency noise of 10@xmath2@xmath3 at 100 hz .
the frequency stabilization servo utilizes three frequency actuators inside the 10-w laser . a thermo - electric cooler ( tec ) bonded to the laser gain medium actuates on the laser frequency by thermally changing the optical path length .
dc1hz adjustments to the laser frequency are made with the tec .
this actuator , modeled as three poles at 0.1hz , has a coefficient of 4ghz / v and is used for large scale adjustments to the laser frequency .
also bonded to the laser gain medium is a pzt , which covers dc10khz . a voltage applied to
the pzt stresses the laser medium and induces refractive index changes to change the laser frequency .
the pzt has a flat response to @xmath4100khz and is known to have a number of mechanical resonances beyond 100khz .
fast frequency fluctuations beyond 10khz are handled by the third frequency actuator , a pockels cell located between the master oscillator and power amplifier .
a small fraction of the output of the pmc is sampled and frequency shifted through an 80mhz acousto - optic modulator ( aom ) .
the output of the aom is focussed into a phase modulator that imparts sidebands at 21.5mhz .
the output of the phase modulator is then modematched into a high - finesse , linear fabry - perot cavity which is used as a frequency reference against which the laser frequency is stabilized .
the frequency stabilization scheme employs the well - known pound - drever - hall technique in which the light incident on the reference cavity is phase - modulated @xcite .
both the carrier and sideband light reflected from the reference cavity is focused onto a tuned photodetector .
the output of the tuned photodetector is bandpass filtered and synchronously demodulated to derive the error signal . in order to ensure closed - loop stability ,
the open - loop gain of the pzt actuator must be well below that of the pockels cell at the pzt mechanical resonance frequency . to ensure this ,
the pzt actuator path is aggressively rolled off after the designed 10khz crossover . in the absence of the pockels cell ,
the pzt path is naturally unstable at @xmath415khz . with a dynamic range some 30 times greater than that of the pockels cell ,
a self - sustaining oscillation may arise if saturation occurs in the pockels cell path .
limiting the dynamic range of the pzt actuator prevents this instability .
photons in the laser light induce a source of noise in the interferometer known as radiation pressure noise .
this noise arises from the momentum imparted to the mirrors as statistically different numbers of photons reflect off the mirrors in the interferometer . to minimize the movement of the interferometer mirrors due to radiation pressure , the intensity fluctuations of the laser
must be stabilized to the level of @xmath410@xmath51/@xmath6 . currently in the prototype design phase
, the intensity servo utilizes a current shunt for fast regulation of the power amplifier pump diode current .
placed in parallel with the power amplifier pump diodes , the current shunt was designed to carry @xmath7250ma .
the intensity stabilization servo adopts a dual - loop topology as illustrated in figure [ psl - iss ] .
inputs from photodetectors located after the pmc and modecleaner are used in either a single or dual sensor configuration . in the single sensor configuration
, the outer - loop photodetector provides the signal to the servo electronics . in the case where the modecleaner is not locked
, the single - sensor signal comes from the inner loop photodetector . in the dual sensor case ,
both the inner and outer feedback paths provide signals to the servo electronics . in the dual loop configuration , noise suppression
is established in two phases . closing the inner loop yields a high - bandwidth , well - behaved inner loop with partial noise suppression . the outer loop is then closed around the inner loop to provide the balance of the noise suppression .
the user control and interface is via the experimental physics and industrial control system ( epics ) . through epics
the operator can remotely monitor the performance of the psl and adjust the various servo loop gains and settings .
the operator interface is a series of graphical screens , that indicate the current status of the psl .
processing the data and events is the input / output controller ( ioc ) , a baja4700e mips - based processor running the vxworks kernel .
the ioc performs the real - world input / output tasks and local control tasks , and provides status information through the channel access network protocol .
the control software for the psl is event - driven and is written in state notation language . although not fully debugged , automated operation from cold start through to full operation has been demonstrated .
one software routine constantly adjusts the tec on the 10-w laser to keep the laser frequency well within the dynamic range of the pzt .
one consequence of this is that lock re - acquisition is instantaneous once the cause of the loss of lock is removed . at present a dozen signals
are acquired and logged through the ligo data acquisition system .
fast signals are acquired at the rate of 16khz whilst slower signals are acquired at 256hz .
all signals are recorded and logged .
we thank the entire ligo team for assistance and support .
this work is supported by the national science foundation under cooperative agreement phy9210038
. 4 alex abramovici _
et.al._ , `` ligo : the laser interferometer gravitational - wave observatory '' , science , * 256 * , 325 , april ( 1992 ) .
a. lazzarini and r. weiss , `` ligo science requirements document ( srd ) '' , internal ligo document e950018 - 02-e .
p. king , r. savage and s. seel , `` ( infrared ) pre - stabilized ( psl ) design requirements '' , internal ligo document t970080 - 09-d .
r.w.p . drever _
et.al._ , `` laser phase and frequency stabilization using an optical resonator '' , appl . phys . *
31 * , 97 , ( 1983 ) . | to meet the strain sensitivity requirements @xcite , @xcite of the laser interferometer gravitational wave observatory ( ligo ) , the laser frequency and amplitude noise must initially be reduced by a factor of 1000 in the pre - stabilized portion of the interferometer @xcite .
a control system was implemented to provide laser noise suppression , data acquisition interfaces , diagnostics , and operator control inputs .
this paper describes the vme - based analog and digital controls used in the ligo pre - stabilized laser ( psl ) . |
the detection of a sparse set of facial landmarks in still images has been a widely - studied problem within the computer vision community .
interestingly , many face analysis methods either systematically rely on video sequences ( e.g. , facial expression recognition @xcite ) or can benefit from them ( e.g. , face recognition @xcite ) .
it is thus surprising that facial landmark tracking has received much less attention in comparison .
our focus in this paper is on one of the most important problems in model - specific tracking , namely that of updating the tracker using previously tracked frames , also known as incremental ( face ) tracking .
the standard approach to face tracking is to use a facial landmark detection algorithm initialised on the landmarks detected at the previous frame .
this exploits the fact that the face shape varies smoothly in videos of sufficiently high framerates : if the previous landmarks were detected with acceptable accuracy , then the initial shape will be close enough for the algorithm to converge to a `` good '' local optimum for the current frame too . hence ,
tracking algorithms are more likely to produce highly accurate fitting results than detection algorithms that are initialised by the face detector bounding box . however , in this setting the tracker still employs a generic deformable model of the face built offline using a generic set of annotated facial images , which does not include the subject being tracked .
it is well known that person - specific models are far more constrained and easier to fit than generic ones @xcite .
hence one important problem in tracking is how to improve the generic model used to track the first few frames into an increasingly person - specific one as more frames are tracked .
learned offline is updated with each new frame . ]
this problem can be addressed with incremental learning , which allows for the smart adaptation of pre - trained generic appearance models .
incremental learning is a common resource for generic tracking , being used in some of the state - of - the - art trackers @xcite , and incremental learning for face tracking is by no means a new concept , please see ross et al .
@xcite for early work on the topic .
more recently , incremental learning within cascaded regression , the state - of - the - art approach for facial landmark localisation , was proposed by xiong & de la torre @xcite and independently by asthana et al .
@xcite . however , in both @xcite and @xcite the model update is far from being sufficiently efficient to allow real - time tracking , with @xcite mentioning that the model update requires 4.7 seconds per frame .
note that the actual tracking procedure ( without the incremental update ) is faster than 25 frames per second , clearly illustrating that the incremental update is the bottleneck impeding real - time tracking .
if the model update can not be carried out in real time , then incremental learning might not be the best option for face tracking - once the real - time constraint is broken in practice one would be better off creating person - specific models in a post - processing step @xcite ( e.g. , re - train the models once the whole video is tracked and then track again ) .
that is to say , without the need and capacity for real - time processing , incremental learning is sub - optimal and of little use . our main contribution in this paper is to propose the first incremental learning framework for cascaded regression which allows real - time updating of the tracking model . to do this ,
we build upon the concept of continuous regression @xcite as opposed to standard sampling - based regression used in almost all prior work , including @xcite and @xcite .
we note that while we tackle the facial landmark tracking problem , cascaded regression has also been applied to a wider range of problems such as pose estimation @xcite , model - free tracking @xcite or object localisation @xcite , thus making our methodology of wider interest .
we will release code for training and testing our algorithm for research purposes .
our main contributions are as follows : * we propose a complete * new formulation for continuous regression * , of which the original continuous regression formulation @xcite is a special case .
crucially , our method is now formulated by means of a * full covariance matrix capturing real statistics * of how faces vary between consecutive frames rather than on the shape model eigenvalues .
this makes our method particularly suitable for the task of tracking , something the original formulation can not deal with .
* we incorporate continuous regression in the cascaded regression framework ( coined cascaded continuous regression , or * ccr * ) and demonstrate its performance is equivalent to sampling - based cascaded regression .
* we derive the * incremental learning for continuous regression * , and show that it * is an order of magnitude faster * than its standard incremental sdm counterpart . *
we evaluate the incremental cascaded continuous regression ( * iccr * ) on the 300vw data set @xcite and show the importance of incremental learning in achieving state - of - the - art performance , especially for the case of very challenging tracking sequences .
facial landmark tracking methods have often been adaptations of facial landmark detection methods .
for example , active appearance models ( aam ) @xcite , constrained local models ( clm ) @xcite or the supervised descent method ( sdm ) @xcite were all presented as detection algorithms .
it is thus natural to group facial landmark tracking algorithms in the same way as the detection algorithms , i.e. splitting them into discriminative and generative methods @xcite . on the generative side ,
aams have often been used for tracking .
since the model fitting relies on gradient descent , it suffices to start the fitting from the last solution .
tracking is particularly useful to aams since they are considered to have frequent local minima and a small basin of attraction , making it important that the initial shape is close to the ground truth .
aams have further been regarded as very reliable for person specific tracking , but not for generic tracking ( i.e. , tracking faces unseen during training ) @xcite .
recently @xcite showed however that an improved optimisation procedure and the use of in - the - wild images for training can lead to well - behaving person independent aam . eliminating the piecewise - affine representation and adopting a part - based model led to the gauss - newton deformable part model ( gn - dpm ) @xcite , which is the aam state of the art . historically , discriminative methods relied on the training of local classifier - based models of appearance , with the local responses being then constrained by a shape model @xcite .
these algorithms can be grouped into what is called the constrained local models ( clm ) framework @xcite . however , the appearance of discriminative regression - based models quickly transformed the state - of - the - art for face alignment .
discriminative regressors were initially used within the clm framework substituting classifiers , showing improved performance @xcite . however , the most important contributions came with the adoption of cascaded regression @xcite and direct estimation of the full face shape rather than first obtaining local estimates @xcite .
successive works have further shown the impressive efficiency @xcite and reliable performance @xcite of face alignment algorithms using cascaded regression .
however , how to best exploit discriminative cascaded regression for tracking and , in particular , how to best integrate incremental learning , is still an open problem .
in this section we revise the preliminary concepts over which we build our method . in particular , we describe the methods most closely related to ours , to wit the incremental supervised descent method @xcite and the continuous regressor @xcite , and motivate our work by highlighting their limitations . a face image
is represented by @xmath0 , and a face shape is a @xmath1 matrix describing the location of the @xmath2 landmarks considered .
a shape is parametrised through a point distribution model ( pdm ) @xcite . in a pdm
, a shape @xmath3 is parametrised in terms of @xmath4 \in \real^{m}$ ] , where @xmath5 represents the rigid parameters and @xmath6 represents the flexible shape parameters , so that @xmath7 , where @xmath8 is a procrustes transformation parametrised by @xmath9 .
@xmath10 and @xmath11 are learned during training and represent the linear subspace of flexible shape variations .
we will sometimes use an abuse of notation by referring treating shape @xmath3 also as function @xmath12 .
we also define @xmath13 as the feature vector representing shape @xmath12 .
an asterisk represents the ground truth , e.g. , @xmath14 is the ground truth shape for image @xmath15 . given a test image @xmath0 , and a current shape prediction @xmath16 , the goal of linear regression for face alignment is to find a mapping matrix @xmath17 able to infer @xmath18 , the increment taking directly to the ground truth , from @xmath19 . by using @xmath20 training images , and @xmath21 perturbations per image ,
the mapping matrix @xmath22 is typically learned by minimising the following expression w.r.t .
@xmath22 : @xmath23 where the bias term is implicitly included by appending a 1 to the feature vector . in order to produce @xmath21 perturbed shapes @xmath24 per image , it suffices to draw the perturbations from an adequate distribution , ideally capturing the statistics of the perturbations encountered at test time .
for example , during detection , the distribution should capture the statistics of the errors made by using the face detection bounding box to provide a shape estimation .
the minimisation in eq .
[ eq : standard_linearreg_problem ] has a closed - form solution .
given @xmath20 images and @xmath21 perturbed shapes per training image , let @xmath25 and @xmath26 represent the matrices containing in its columns the input feature vectors and the target output @xmath27 respectively .
then , the optimal regressor @xmath22 can be computed as : @xmath28 given a test shape @xmath12 , the predicted shape is computed as @xmath29 .
continuous regression ( cr ) @xcite is an alternative solution to the problem of linear regression for face alignment .
the main idea of continuous regression is to treat @xmath18 as a continuous variable and to use _ all samples _ within some finite limits , instead of sampling a handful of perturbations per image .
that is to say , the problem is formulated in terms of finite integrals as : @xmath30 where @xmath31 is the eigenvalue associated to the @xmath32-th flexible parameter of the pdm , @xmath33 represent the number of flexible parameters , and @xmath34 is a parameter determining the number of standard deviations considered in the integral .
unfortunately , this formulation does not have a closed - form solution .
however , it is possible to solve it approximately in a very efficient manner by using a first order taylor expansion of the loss function . following the derivations in @xcite , we denote @xmath35 as the jacobian of the image features with respect to the shape parameters evaluated at the ground truth @xmath36 , which can be calculated simply as @xmath37 . a solution to eq .
[ eq : old_continuous ] can then be written as : @xmath38 where @xmath39 is a diagonal matrix whose @xmath32-th entries are defined as @xmath40 .
cr formulated in this manner has the following practical limitations : 1 .
it does not account for correlations within the perturbations .
this corresponds to using a fixed ( not data - driven ) diagonal covariance to model the space of shape perturbations , which is a harmful oversimplification .
because of 1 , it is not possible to incorporate cr within the popular cascaded regression framework in an effective manner .
derivatives are computed over image pixels , so more robust features , e.g. , hog or sift , are not used .
4 . the cr can only account for the flexible parameters , as the integral limits are defined in terms of the eigenvalues of the pdm s pca space .
in section [ ssec : continuous_regression ] we will solve all of these shortcomings , showing that it is possible to formulate the cascaded continuous regression and that , in fact , its performance is equivalent to the sdm .
the main limitation of using a single linear regressor to predict the ground truth shape is that the training needs to account for too much intra - class variation .
that is , it is hard for a single regressor to be simultaneously accurate and robust . to solve this , @xcite successfully adapted the cascaded regression of framework of dollr et al .
@xcite to face alignment .
however , the most widely - used form of face alignment is the sdm @xcite , which is a cascaded linear regression algorithm . at * test time * , the sdm takes an input @xmath41 , and then for a fixed number of iterations computes @xmath42 and @xmath43 .
the key idea is to use a different regressor @xmath44 for each iteration .
the input to the * training * algorithm is a set of images @xmath45 and corresponding perturbed shapes @xmath46 . the training set @xmath32 is defined as @xmath47 , with @xmath48 , and @xmath49 , with @xmath50
. then regressor @xmath32 is computed using eq .
[ eq : linear_reg ] on training set @xmath32 , and a new training set @xmath51 is created using the shape parameters @xmath52 .
incremental versions of the sdm have been proposed by both xiong & de la torre @xcite and asthana et al . @xcite .
the latter proposed the _
parallel sdm _ , a modification of the original sdm which facilitates the incremental update of the regressors .
more specifically , they proposed to alter the sdm training procedure by modelling @xmath53 as a normal distribution @xmath54 , allowing training shape parameters to be sampled for the next level of the cascade as : @xmath55 once the parallel sdm is defined , its incremental extension is immediately found . without loss of generality , we assume that the regressors are updated in an on - line manner , i.e.
, the information added is extracted from the fitting of the last frame .
we thus define @xmath56 , arrange the matrices @xmath57 and @xmath58 accordingly , and define the shorthand @xmath59 , leading to the following update rules @xcite : @xmath60 where @xmath61 is the @xmath21-dimensional identity matrix .
the cost for these incremental updates is dominated by the multiplication @xmath62 , where both matrices have dimensionality @xmath63 , which has a computational complexity of @xmath64 .
since @xmath65 is high - dimensional ( @xmath66 ) , the cost of updating the models becomes prohibitive for real - time performance .
once real time is abandoned , offline techniques that do not analyse every frame in a sequential manner can be used for fitting , e.g. , @xcite .
we provide a full analysis of the computational complexity in section [ sec : computational_complexity ] .
in this section we describe the proposed incremental cascaded continuous regression , which to the best of our knowledge is the first cascaded regression tracker with * real - time incremental learning * capabilities .
to do so , we first extend the continuous regression framework into a fully fledged cascaded regression algorithm capable of performance on par with the sdm ( see sections [ ssec : continuous_regression ] and [ ssec : cascaded_continuous_regression ] ) .
then , we derive the incremental learning update rules within our cascaded continuous regression formulation ( see section [ ssec : incremental_learning_update_rules ] )
. we will show in section [ sec : computational_complexity ] that our newly - derived formulas have complexity of one order of magnitude less than previous incremental update formulations .
we first modify the original formulation of continuous regression .
in particular , we add a `` data term '' , which is tasked with encoding the probability of a certain perturbed shape , allowing for the modelling of correlations in the shape dimensions . plainly speaking
, the previous formulation assumed an i.i.d .
uniform sampling distribution .
we instead propose using a data - driven full covariance distribution , resulting in regressors that model the test - time scenario much better .
in particular , we can see the loss function to be optimised as : @xmath67 ] it is interesting to note that this equation appears in @xcite , where the sdm equations are interpreted as a mcmc sampling - based approximation of this equation .
contrariwise , the continuous regression proposes to use a different approximation based on a first - order taylor approximation of the _ ideal loss function _ defined in eq .
[ eq : train_problem_def ] .
however , the continuous regression proposed in @xcite extends the functional data analysis @xcite framework to the imaging domain , without considering any possible data correlation .
instead , the `` data term '' in eq .
[ eq : train_problem_def ] ( which defines how the data is sampled in the mcmc approach ) , will serve to correlate the different dimensions in the continuous regression .
that is to say , the data term " does not play the role of how samples are taken , but rather helps to find an analytical solution in which dimensions can be correlated .
these differences are illustrated in figure [ fig : differences ] .
the first - order approximation of the feature vector is given by : @xmath68 where @xmath35 is the jacobian of the feature representation of image @xmath45 at @xmath36 .
while @xcite used a pixel - based representation , the jacobian under an arbitrary representation can be computed empirically as : @xmath69 ) - f ( \i , [ \s_x - \delta x , \s_y ] ) } { 2 \delta x}\ ] ] where @xmath70 are the @xmath71 coordinates of shape @xmath3 , and @xmath72 indicates that @xmath73 is added to each element of @xmath70 ( in practice , @xmath73 is the smallest possible , 1 pixel ) .
@xmath74 can be computed similarly . then @xmath75 \frac{\partial\s } { \partial \p_j^*}$ ] .
[ eq : train_problem_def ] has a closed form solution as : @xmath76 where @xmath77 and @xmath78 are the mean and covariance of the data term , @xmath79 .
finally , we can see that eq .
[ eq : closed_form_cont ] can be expressed in a more compact form .
let us first define the following shorthand notation : @xmath80 $ ] , @xmath81 , @xmath82 $ ] and @xmath83 $ ] . then : @xmath84 where @xmath85 . through this arrangement , the parallels with the sampling - based regression formula are clear ( see eq .
[ eq : linear_reg ] ) .
it is interesting that , while the standard linear regression formulation needs to sample perturbed shapes from a distribution , the continuous regression training formulation only needs to extract the features and the jacobians on the ground - truth locations .
this means that once these features are obtained , re - training a new model under a different distribution takes seconds , as it only requires the computation of eq .
[ eq : compact_form_cont ] .
now that we have introduced a new formulation with the continuous regression capable of incorporating a data term , it is straightforward to extend the cr into the cascade regression formulation : we take the distribution in equation [ eq : parsdm_sampling ] as the _ data term _ in eq .
[ eq : train_problem_def ] .
one might argue that due to the first - order taylor approximation required to solve equation [ eq : train_problem_def ] , ccr might not work as well as the sdm .
one of the main experimental contributions of this paper is to show that in reality this is not the case : in fact ccr and sdm have equivalent performance ( see section [ sec : experimental_results ] ) .
this is important because , contrary to previous works on cascaded regression , incremental learning within ccr allows for real time performance .
once frame @xmath15 is tracked , the incremental learning step updates the existing training set @xmath86 with @xmath87 , where @xmath88 denotes the predicted shape parameters for frame @xmath15 .
note that in this case @xmath89 consists of only one example compared to @xmath21 examples in the incremental sdm case .
the update process consists of computing matrix @xmath90 , which stores the feature vector and its jacobian at @xmath88 and then , using the shorthand notation @xmath91 , updating continuous regressor as : @xmath92 in order to avoid the expensive re - computation of @xmath93 , it suffices to update its value using the woodbury identity @xcite : @xmath94 note that @xmath95 , where @xmath96 accounts for the number of shape parameters .
we can see that computing eq .
[ eq : closed_form_icont ] requires computing first @xmath97 , which is @xmath98 .
this is a central result of this paper , and reflects a property previously unknown .
we will examine in section [ sec : computational_complexity ] its practical implications in terms of real - time capabilities .
in this section we first detail the computational complexity of the proposed iccr , and show that it is real - time capable . then
, we compare its cost with that of incremental sdm , showing that our update rules are an order of magnitude faster .
* iccr update complexity : * let us note the computational cost of the feature extraction as @xmath99 .
the update only requires the computation of the feature vector at the ground truth , and in two adjacent locations to compute the jacobian , thus resulting in @xmath100 complexity .
interestingly , this is independent from the number of cascade levels .
then , the update equation ( eq . [ eq : closed_form_icont ] ) , has a complexity dominated by the operation @xmath101 , which has a cost of @xmath102 .
it is interesting to note that @xmath103 is a matrix of size @xmath104 and thus its inversion is extremely efficient .
the detailed cost of the incremental update is : @xmath105 * incremental sdm update complexity : * incremental learning for sdm requires sampling at each level of the cascade .
the cost per cascade level is @xmath106 , where @xmath21 denotes the number of samples .
thus , for @xmath107 cascade levels the total cost of sampling is @xmath108 .
the cost of the incremental update equations ( eqs .
( [ eq : incremental_discrete_eq1]-[eq : incremental_discrete_eq4 ] ) ) , is in this case dominated by the multiplication @xmath109 , which is @xmath64 .
the detailed computational cost is : @xmath110 * detailed comparison and timing : * one advantage of iccr comes from the much lower number of feature computations , being as low as 3 vs. the @xmath111 computations required for incremental sdm .
however , the main difference is the @xmath112 complexity of the regressor update equation for the incremental sdm compared to @xmath113 for the iccr . in our case , @xmath114 , while @xmath115 . the feature dimensionality results from performing pca over the feature space , which is a standard procedure for sdm .
note that if we avoided the use of pca , the complexity comparison would be even more in our favour .
a detailed summary of the operations required by both algorithms , together with their computational complexity and the execution time on our computer are given in algorithm [ algo_iccr ] .
note that @xmath116 is the cost of projecting the output vector into the pca space .
note as well that for incremental sdm , the `` sampling and feature extraction '' step is repeated @xmath107 times .
_ iccr update ( total : 72 ms . ) : _
+ ' '' '' _ isdm @xcite update ( total : 705 ms . ) : _ +
this section describes the experimental results .
first , we empirically demonstrate the performance of ccr is equivalent to sdm . in order to do so ,
we assess both methods under the same settings , avoiding artefacts to appear , such as face detection accuracy .
we follow the vot challenge protocol @xcite . then , we develop a fully automated system , and we evaluate both the ccr and iccr in the same settings as the 300vw , and show that our fully automated system achieves state of the art results , illustrating the benefit of incremental learning to achieve it .
* training data : * we use data from different datasets of static images to construct our training set . specifically , we use helen @xcite , lfpw @xcite , afw @xcite , ibug @xcite , and a subset of multipie @xcite .
the training set comprises @xmath1177000 images .
we have used the facial landmark annotations provided by the 300 faces in the wild challenge @xcite , as they offer consistency across datasets .
the _ statistics _ are computed across the training sequences , by computing the differences of ground - truth shape parameters between consecutive frames .
given the easiness of the training set with respect to the test set , we also included differences of several frames ahead . this way , higher displacements are also captured .
* features : * we use the sift @xcite implementation provided by xiong & de la torre @xcite .
we apply pca on the output , retaining 2000 dimensions .
we apply the same pca to all of the methods , computed during our sdm training .
* test data : * all the methods are evaluated on the test partition of the 300 videos in the wild challenge ( 300vw ) @xcite .
the 300vw is the only publicly - available large - scale dataset for facial landmark tracking .
its test partition has been divided into categories 1 , 2 and 3 , intended to represent increasingly unconstrained scenarios .
in particular , category 3 contains videos captured in totally unconstrained scenarios .
the ground truth has been created in a semi - supervised manner using two different methods @xcite .
* error measure : * to compute the error for a specific frame , we use the error measure defined in the 300vw challenge @xcite .
the error is computed by dividing the average point - to - point euclidean error by the inter - ocular distance , understood as the distance between the two outer eye corners . in order to demonstrate the performance capability of our ccr method against sdm
, we followed the protocol established by the visual object tracking ( vot ) challenge organisers for evaluating the submitted tracking methods @xcite . specifically ,
if the tracker error exceeds a certain threshold ( 0.1 in our case , which is a common definition of alignment failure ) , we proceed by re - initialising the tracker . in this case
, the starting point will be the ground truth of the previous frame .
this protocol is adopted to avoid the pernicious influence on our comparison of some early large failure from which the tracker is not able to recover , which would mean that successive frames would yield a very large error .
results are shown in fig .
[ fig : ccrvssdm ] ( * left * ) .
we show that the ccr and the sdm provide similar performance , thus ensuring that the ccr is a good starting point for developing an incremental learning algorithm .
it is possible to see from the results shown in fig .
[ fig : ccrvssdm ] that the ccr compares better and even sometimes surpasses the sdm on the lower levels of the error , while the sdm systematically provides a gain for larger errors with respect to the ccr .
this is likely due to the use of first - order taylor approximation , which means that larger displacements are less accurately approximated .
instead , the use of _ infinite _ shape perturbations rather than a handful of sampled perturbations compensates this problem for smaller errors , and even sometimes provides some performance improvement .
we now show the benefit of incremental learning with respect to generic models .
the incremental learning needs to filter frames to decide whether a fitting is suitable or harmful to update the models .
that is , in practice , it is beneficial to filter out badly - tracked frames by avoiding performing incremental updates in these cases .
we follow @xcite and use a linear svm trained to decide whether a particular fitting is `` correct '' , understood as being under a threshold error . despite its simplicity
, this tactic provides a solid performance increase .
results on the test set are shown in fig .
[ fig : ccrvssdm ] ( * right * ) . , title="fig : " ] , title="fig : " ] we developed a fully automated system to compare against state of the art methods .
our fully automated system is initialised with a standard sdm @xcite , and an svm is used to detect whether the tracker gets lost .
we assessed both our ccr and iccr in the most challenging category of the 300vw , consisting of 14 videos recorded in unconstrained settings . for a fair comparison
, we have reproduced the challenge settings ( a brief description of the challenge and submitted methods can be found in @xcite ) .
we compare our method against the top two participants @xcite .
results are shown in fig .
[ fig : soacomp ] .
the influence of the incremental learning to achieve state of the art results is clear .
importantly , as shown in the paper , our iccr allows for real - time implementation .
that is to say , our iccr reports state of the art results whilst working in near real - time , something that could not be achieved by previous works on cascaded regression .
code for our fully automated system is available for download at www.cs.nott.ac.uk/~psxes1 . ]
in this article we have proposed a novel facial landmark tracking algorithm that is capable of performing on - line updates of the models through incremental learning . compared to previous incremental learning methodologies
, it can produce much faster incremental updates without compromising on accuracy .
this was achieved by firstly extending the continuous regression framework @xcite , and then incorporating it into the cascaded regression framework to lead to the ccr method , which we showed provides equivalent performance to the sdm .
we then derived the incremental learning update formulas for the ccr , resulting in the iccr algorithm .
we further show the computational complexity of the incremental sdm , demonstrating that iccr is an order of magnitude simpler computationally .
this removes the bottleneck impeding real - time incremental cascaded regression methods , and thus results in the state of the art for real - time face tracking .
the work of snchez - lozano , martinez and valstar was supported by the european union horizon 2020 research and innovation programme under grant agreement no 645378 , aria - valuspa .
the work of snchez - lozano was also supported by the vice - chancellor s scholarship for research excellence provided by the university of nottingham .
the work of tzimiropoulos was supported in part by the epsrc project ep / m02153x/1 facial deformable models of animals .
we are also grateful for the given access to the university of nottingham high performance computing facility , and we would like to thank jie shen and grigoris chrysos for their insightful help in our tracking evaluation .
as shown in the paper , linear regression aims at minimising the average expected training error , respect to @xmath22 .
the average expected error is formulated as follows : @xmath118 where @xmath119 represents the ground truth shape parameters , and @xmath18 the parameters displacement . due to the intractability of the integral ,
this is typically solved by a mcmc ( sampling - based ) approximation , in which samples are taken from the distribution @xmath79 .
the continuous regression framework avoids the need to sample by performing the first order taylor approximation of the function @xmath120 , defined as : @xmath121 where @xmath122 , evaluated at @xmath123 , is the jacobian of the feature representation of image @xmath45 , respect to shape parameters @xmath124 , at @xmath36 . combining this approximation with the integral in eq .
[ eq : train_problem_def_appx ] leads to : @xmath125 d \delta \p \hspace{-5pt } \label{eq : long_deriv_ccr_appx}\end{aligned}\ ] ] where recall @xmath126 is the shorthand of @xmath127 .
if we group independent , linear , and quadratic terms , respect to @xmath18,we can express eq .
[ eq : long_deriv_ccr_appx ] as : @xmath128 d \delta \p , \label{eq : long_deriv_ccr2_appx}\end{aligned}\ ] ] where @xmath129 and @xmath130 .
let us assume that @xmath79 is parametrised by its mean @xmath77 and covariance @xmath78 .
then , it follows that : @xmath131 which means that the expected error , for the @xmath15-th training example , has a closed - form solution as follows : @xmath132 now , @xmath22 is obtained after minimising eq .
[ eq : closed_form_appx ] , whose derivatives are obtained as follows :
@xmath133 this leads to the solution presented in the paper : @xmath134
we can see that our new formulation generalises that of @xcite . more specifically , if we solve eq .
[ eq : train_problem_def ] for the non - rigid parameters only , we can define @xmath79 to be a uniform distribution defined within the limits @xmath135 , with @xmath31 the eigenvalue assciated to the @xmath32-th basis of the pdm , and @xmath34 the number of standard deviations considered for that eigenvalue . in such case , eq .
[ eq : train_problem_def ] would be defined , for @xmath136 non - rigid parameters , as : @xmath137 which is the problem definition that appeared in @xcite .
moreover , we can see that such a uniform distribution would be parametrised by a zero - mean vector and a diagonal covariance matrix whose entries are @xmath40 .
in such case , eq . [ eq : closed_form_cont ] would be reduced to the solution presented in @xcite .
that is to say , the continuous regression presented in @xcite assumed a uniform distribution , without connection to tracking statistics , and no correlation between target dimensions were possible .
instead , our formulation accepts a data term " , which correlates the target dimensions , and allows for its solution for rigid parameters as well .
data term " is crucial to the performance of the ccr . | this paper introduces a novel real - time algorithm for facial landmark tracking .
compared to detection , tracking has both additional challenges and opportunities .
arguably the most important aspect in this domain is updating a tracker s models as tracking progresses , also known as incremental ( face ) tracking .
while this should result in more accurate localisation , how to do this online and in real time without causing a tracker to drift is still an important open research question .
we address this question in the cascaded regression framework , the state - of - the - art approach for facial landmark localisation . because incremental learning for cascaded regression is costly
, we propose a much more efficient yet equally accurate alternative using continuous regression .
more specifically , we first propose cascaded continuous regression ( ccr ) and show its accuracy is equivalent to the supervised descent method .
we then derive the incremental learning updates for ccr ( iccr ) and show that it is an order of magnitude faster than standard incremental learning for cascaded regression , bringing the time required for the update from seconds down to a fraction of a second , thus enabling real - time tracking . finally , we evaluate iccr and show the importance of incremental learning in achieving state - of - the - art performance .
code for our iccr is available from http://www.cs.nott.ac.uk/~psxes1 16subnumber1763 |
since the pioneering works of fuller , fowler , & newman ( 1980 , 1982a , 1982b , 1985 ) , calculations of stellar weak - interaction rates have entered an era of precision science .
more recently an authoritative work of langanke & martinez - pinedo ( 2000 ) on this subject appeared .
since the presupernova stellar evolution and stellar nucleosynthesis critically depend on the details of the stellar weak - interaction rates ( e.g. , wanajo et al .
2002 ) , it is extremely important to calculate accurately the screening corrections to the electron capture rates in dense stars by the relativistically degenerate electron liquid .
this problem has been already addressed by many authors ( couch & loumos 1974 ; takahashi , el eid , & hillebrandt 1978 ; gutierrez et al .
1996 ; luo & peng 1996 ; bravo & garcia - senz 1999 ) .
the plasma effects on the chemical potential of the nucleus and hence on the threshold energy for the electron capture , in particular , have been discussed by couch & loumos ( 1974 ) , gutierrez et al .
( 1996 ) , as well as by bravo & garcia - senz ( 1999 ) .
in this paper we will address ourselves to the calculation of the effective potential energy felt by the relativistically degenerate electron .
we will use the linear response theory in order to calculate the screening potential caused by the relativistically degenerate electron liquid .
the present paper is organized as follows . in
2 we will calculate the effective potential energy felt by the electron using the longitudinal dielectric function of the relativistically degenerate electron liquid derived by jancovici ( 1962 ) .
we will thereby calculate the screening potential which will be used for the calculation of the screening corrections to the electron capture rates . in
3 we will summarize the numerical results in the form of analytic fitting formulae which will facilitate the application of the present results .
we will give concluding remarks in
the coulomb corrections to the beta decay rates have been discussed in many standard textbooks on beta decay ( e.g. , morita 1973 ) .
some authors ( takahashi , el eid , & hillebrandt 1978 ; fuller , fowler , & newman 1980 ) have discussed possible importance of the screening effects on the electron capture rates at extremely high densities .
therefore , it is extremely important to calculate accurately the screening corrections to the electron capture rates by the relativistically degenerate electron liquid .
we consider the density - temperature regime in which the electrons are strongly degenerate .
this condition is expressed as ( itoh et al .
1983 ) @xmath0 \ , [ { \rm k } ] \ , , \end{aligned}\ ] ] where @xmath1 is the electron fermi temperature , @xmath2 is the mass density in units of @xmath3 g @xmath4 , and @xmath5 and @xmath6 are the atomic number and mass number of the nucleus considered , respectively . we assume complete pressure ionization .
therefore , no bound electrons will exist around the nucleus .
this assumption is valid when the following condition is satisfied ( itoh et al .
1996 ) : @xmath7 \ , , \end{aligned}\ ] ] where @xmath8 is the mass density .
the potential energy of the electron due to the pure coulomb potential @xmath9 is changed to @xmath10 because of the screening by the relativistically degenerate electron liquid . in this paper
we use the static longitudinal dielectric function due to the relativistically degenerate electron liquid calculated by jancovici ( 1962 ) based on the relativistic random - phase approximation .
jancovici s static longitudinal dielectric function is written in the form ( itoh et al .
1984 ) @xmath11 , \\
q & = & \frac{k}{2 k_{f } } \ , , \\ x & = & \frac { \hbar k_{f}}{m_{e}c } \ , = \ ,
\frac{1}{137.036 } \left ( \frac{9 \pi}{4 } \right)^{1/3 } \ , r_{s}^{-1 } \ , , \\
r_{s } & = & \frac { \ , a_{e } \ , } { \displaystyle { \frac { \hbar^{2 } } { m_{e}e^{2 } } } } = 1.388 \times 10^{-2 } \left ( \frac{a}{z \rho_{6 } } \right)^{1/3 } \ , , \\ & & \frac{4}{3 } \pi a_{e}^{3 } \ , n_{e } \ , = \ , 1 \ , , \end{aligned}\ ] ] where @xmath12 is the electron number density .
the electron fermi wavenumber is expressed as @xmath13 \ , .\end{aligned}\ ] ] the ion - sphere radius @xmath14 is written as ( itoh & kohyama 1993 ) @xmath15 \ , , \end{aligned}\ ] ] where @xmath16 is the ion number density .
we also have the relationship @xmath17 the potential energy of the electron @xmath10 which takes into account the screening by the relativistially degenerate electron liquid is written as @xmath18}{q \ , \epsilon(q , 0 ) } \ , d q \ , .\end{aligned}\ ] ] therefore the increment of the potential energy due to the screening by the relativistically degenerate electron liquid is @xmath19}{q \ , \epsilon(q , 0 ) } \ , d q \right\ } \nonumber \\ & = & 7.525 \times 10^{-3 } \ , z \left ( \frac{z}{a } \rho_{6 } \right)^{1/3 } [ { \rm mev } ] \frac{1}{2k_{f}r } \left\ { 1 - \frac{2}{\pi } \ , \int_{0}^{\infty } \frac { { \rm sin } [ ( 2 k_{f}r)q]}{q \ , \epsilon(q , 0 ) } \ , d q \right\ } .\end{aligned}\ ] ] equation ( 2.12 ) is based on the linear response theory ( e.g. , pines & nozires 1966 ) .
dharma - wardama & perrot ( 1982 , 1990 ) have carried out the density - functional study of hydrogen plasmas as well as the density - functional study of c , si , and ge metallic liquids and have found that the detailed results of the density - functional calculations of these systems are close to the results obtained by the linear response theory .
they have also confirmed that the density - functional theory as well as the linear response theory reproduce satisfactorily the experimental results on the ge metallic liquid , thereby proving the excellence of these theories for this system .
this fact gives great support to our use of the linear response theory in the present problem .
compared with the case of the pure coulomb potential , the effective electron energy at the position of the nucleus is increased by @xmath20 . hence the usual factor in the electron - capture rates @xmath21 ( morita 1973 ; fuller , fowler , & newman 1980 ; langanke & martinez - pinedo 2000 ) should be replaced by @xmath22^{2 } - m_{e}^{2 } \right\}^{1/2 } \left [ e - v_{s}(0 ) \right ] \ , f \left[z , e - v_{s}(0 ) \right ] \ , .\end{aligned}\ ] ] at the same time the shift in the threshold energy for the electron capture should be taken into account ( couch & loumos 1974 ; gutierrez et al . 1996 ; bravo & garcia - senz 1999 ) . in figure 1
we show the function @xmath23}{q \ , \epsilon(q , 0 ) } \ , d q\end{aligned}\ ] ] as a function of @xmath24 for various values of @xmath25 . in figure
2 we show the function @xmath26}{q \ , \epsilon(q , 0 ) } \ , d q \right\}\end{aligned}\ ] ] as a function of @xmath24 for various values of @xmath25 .
in figure 3 we enlarge figure 2 for small values of @xmath24 . in figure 4 we show @xmath27 as a function of @xmath25 .
equation ( 2.13 ) can be rewritten as @xmath28 \ , .\end{aligned}\ ] ] the nuclear radius @xmath29 can be expressed as ( morita 1973 ) @xmath30 \ , .\end{aligned}\ ] ] from equations ( 2.8 ) and ( 2.18 ) we have @xmath31 from figure 3 we see that the function @xmath32 is almost constant for small values of @xmath33 which corresponds to @xmath34 given by equation ( 2.19 ) , when the mass density is not extremely high .
this fact justifies our use of @xmath20 for the screening potential at the nuclear radius .
when the mass density becomes extremely high , then one should use the value of the screening potential @xmath35 by adopting the value of the function @xmath36 .
the electron fermi energy is given by @xmath37 \
, [ { \rm mev } ] \ , .\end{aligned}\ ] ] when the electrons are extremely relativistic , we have @xmath38 therefore , we find that the screening potential at the origin @xmath20 is typically a few percent of the electron fermi energy .
thus we conclude that the screening corrections to the electron capture rates by the relativistically degenerate electron liquid are not as great as anticipated by takahashi , el eid , & hillebrandt ( 1978 ) and also by fuller , fowler , & newman ( 1980 ) . in passing ,
it is interesting to compare our present detailed calculation with the fermi - thomas model which is generally more crude than the random - phase approximation that has been adopted in the present paper ( pines and nozires 1966 ) . in the fermi - thomas model the screening potential at the origin
is expressed as ( pines and nozires 1966 ) @xmath39_{ft } & = & z e^{2 } \
, k_{ft } \
, , \end{aligned}\ ] ] where @xmath40 is the fermi - thomas screening wave number .
when the electrons are extremely relativistic , the fermi - thomas screening wave number is expressed as ( flowers & itoh 1976 ) @xmath41 by comparing with figures 2 , 3 , 4 , we find that the prediction of the fermi - thomas model is quite good ( the accuracy is about 7% ) for the present problem when the electrons are extremely relativistic .
in this section we will present accurate analytic fitting formulae for @xmath32 in order to facilitate the application of the numerical results obtained in the present paper .
we have carried out the numerical calculations of @xmath32 for @xmath42 , @xmath43 .
we express the analytic fitting formula by @xmath44 the coefficients @xmath45 are presented in table 1 .
the accuracy of the fitting is generally better than 0.1% .
the value of @xmath46 for @xmath47 is of course obtained by the fitting formulae ( 3.1 ) , ( 3.2 ) , ( 3.3 ) . for the sake of simplicity
we will give a separate fitting formula @xmath48 the coefficients @xmath49 are presented in table 2 .
the accuracy of the fitting is generally better than 0.1% .
we have studied the screening corrections to the electron capture rates by the relativistically degenerate electron liquid . in particular , we have calculated the screening potential caused by the relativistically degenerate electron by using jancovici s ( 1962 ) static longitudinal dielectric function .
we have found that the screening potential is typically a few percent of the electron fermi energy .
hence we conclude that the screening corrections to the electron capture rates at high densities are not as great as anticipated by takahashi , el eid , & hillebrandt ( 1978 ) and also by fuller , fowler , & newman ( 1980 ) .
we have presented accurate analytic fitting formulae which will be useful when one wishes to apply the present results to the calculations of the screening corrections to the electron capture rates at high densities .
we wish to thank k. nomoto for drawing our attention to the present important problem .
we are grateful to y. takada for clarifying the validity of the linear response theory in the present problem .
we also thank y. oyanagi for allowing us to use the least - squares fitting program sals . this work is financially supported in part by grants - in - aid of the japanese ministry of education , culture , sports , science , and technology under contracts 13640245 , 13740129 .
bravo , e. , & garcia - senz , d. 1999 , mnras , 307 , 984 couch , r. g. , & loumos , g. l. 1974 , apj , 194 , 385 dharma - wardana , m. w. c. , & perrot , f. 1982 , phys .
a26 , 2096 dharma - wardana , m. w. c. , & perrot , f. 1990 , phys .
letters , 65 , 76 flowers , e. , & itoh , n. 1976 , apj , 206 , 218 fuller , g. m. , fowler , w. a. , & newman , m. j. 1980 , apjs , 42 , 447 fuller , g. m. , fowler , w. a. , & newman , m. j. 1982a , apj , 252 , 715 fuller , g. m. , fowler , w. a. , & newman , m. j. 1982b , apjs , 48 , 279 fuller , g. m. , fowler , w. a. , & newman , m. j. 1985 , apj , 293 , 1 gutierrez , j. , garcia - berro , e. , iben , i. , isern , j. , labay , j. , & canal , r. 1996 , apj , 459 , 701 itoh , n. , hayashi , h. , nishikawa , a. , & kohyama , y. 1996 , apjs , 102 , 411 itoh , n. , & kohyama , y. 1993 , apj , 404 , 268 itoh , n. , kohyama , y. , matsumoto , m. , & seki , m. 1984 , apj , 285 , 758 ( erratum 404 , 418 [ 1993 ] ) itoh , n. , mitake , s. , iyetomi , h. , & ichimaru , s. 1983 , apj , 273 , 774 jancovici , b. 1962 , nuovo cimento , 25 , 428 langanke , k. , & martinez - pinedo 2000 , nucl .
phys . , a673 , 481 luo , z. q. , & peng , q. h. 1996 , science in china ( series a ) , 39 , 776 morita , m. 1973 , _ beta decay and muon capture _ ( reading , massachusetts : w. a. benjamin ) pines , d. , & nozires , p. 1966 , _ the theory of quantum liquids _
( new york : w. a. benjamin ) takahashi , k. , el eid , m. f. , & hillebrandt , w. 1978 , a & a , 67 , 185 wanajo , s. , itoh , n. , ishimaru , y. , nozawa , s. , & beers , t. c. 2002 , submitted to apj | we calculate the screening corrections to the electron capture rates in dense stars by the relativistically degenerate electron liquid . in order to calculate the screening corrections we adopt the linear response theory which is widely used in the field of solid state physics and liquid metal physics .
in particular , we use the longitudinal dielectric function for the relativistically degenerate electron liquid derived by jancovici .
we calculate the screening potential at the position of the nucleus . by using this screening potential
one can calculate the screening corrections to the electron capture rates .
we will present accurate analytic fitting formulae which summarize our numerical results .
these fitting formulae will facilitate the application of the present results .
the screening corrections to the electron capture rates are typically a few percent . |
numerical simulations of galaxy formation and interactions often include rules for star formation . in many cases ,
the local rate of star formation , @xmath0 , is related to the local gas density , @xmath1 , by a power law : @xmath2 where @xmath3 is a constant . this prescription has been justified both empirically and theoretically . from an empirical perspective , ( [ sfr - power - law ] )
resembles the `` schmidt law '' @xcite , which relates star formation per unit surface area , @xmath4 to gas surface density , @xmath5 ; a recent determination of the schmidt law @xcite is @xmath6 @xcite adopted a rule equivalent to ( [ sfr - power - law ] ) with @xmath7 as an approximation to a schmidt law with index @xmath8 , while @xcite took @xmath9 , and presented numerical tests showing this gave a reasonable match to a schmidt law with index @xmath10 .
a more theoretical approach , adopted by @xcite and @xcite , sets @xmath11 , where @xmath12 , the time - scale for star formation , is basically proportional to the local collapse time of the gas , @xmath13 .
this yields ( [ sfr - power - law ] ) with @xmath9 and @xmath14 , where @xmath15 is a dimensionless constant . despite the apparent convergence of observation and theory on the index @xmath9 , it s unlikely that ( [ sfr - power - law ] ) really captures the process of star formation .
for one thing , only about @xmath16 per cent of the gas actually forms stars per collapse time @xmath17 ; in other words , consistency with the observations implies that @xmath18 .
current theories of star formation do nt offer any straightforward way to calculate this quantity .
gravitational collapse is evidently not the limiting factor which sets the rate of star formation ; `` feedback '' processes are important in determining the fraction of available interstellar material which ultimately becomes stars . while several groups have now devised simulations including star formation regulated by various forms of feedback ( e.g. @xcite ; @xcite ) , this approach still has some way to go .
moreover , models based on ( [ sfr - power - law ] ) do nt reproduce the large - scale star formation seen in many interacting galaxies .
@xcite found that most of the star formation was confined to the central regions of their model galaxies , while @xcite noted that ( [ sfr - power - law ] ) underestimated the rate of star formation in regions where interstellar gas exhibits large velocity dispersions and gradients , or where the gas appears to be undergoing strong shocks .
@xcite , @xcite , and @xcite have modeled the central bursts of star formation seen in ultraluminous infrared galaxies ( * ? ? ? * and references therein ) .
but nuclear starbursts , while a necessary stage in the transformation of merger remnants into elliptical galaxies ( e.g. * ? ? ?
* ) , may not be sufficient to accomplish this transformation .
for example , merger - induced starbursts create massive young star clusters ( e.g. * ? ? ? * ) which may subsequently evolve into globular clusters , but these clusters will be confined to the nuclei of remnants _ unless _ the starbursts are spatially extended . in view of these considerations , it s worth examining alternatives to ( [ sfr - power - law ] ) .
one long standing idea is that collisions of molecular clouds trigger of star formation ( e.g. @xcite ) .
this has been implemented in `` sticky particle '' schemes which model the interstellar medium as a collection of discrete clouds undergoing inelastic collisions ( e.g. @xcite ; @xcite ) .
however , molecular clouds have relatively long mean free paths , and direct collisions between clouds at velocities of @xmath19 or more may result in disruption rather than star formation .
@xcite proposed a model of shock - induced star formation in interacting galaxies ; specifically , they suggested that fast collisions between extended h clouds create a high - pressure medium which compresses pre - existing molecular clouds and thereby induces bursts of star formation .
high - pressure regions , and especially large - scale shocks in colliding galaxies , may favor the formation of bound star clusters @xcite .
recent observations at optical , infrared , and radio wavelengths continue to reveal large - scale star formation in interacting galaxies , including ngc4038/9 , arp299 , and ngc4676 ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
several of these studies explicitly invoke shock - induced star formation in discussing the observational data . in this paper
i focus on ngc4676 , a strongly interacting pair of disk galaxies with long tidal tails ( @xcite , hereafter tt72 ; @xcite ) , as a test - case for models of shock - induced and density - dependent star formation .
2 describes the star formation algorithms .
3 presents simulations of ngc4676 , and contrasts the results of shock - induced and density - dependent star formation rules .
conclusions are given in
within the framework of `` smoothed particle hydrodynamics '' or sph ( @xcite ; @xcite ; @xcite and references therein ) , the interstellar material can be modeled as an isothermal gas with a representative temperature @xmath20 .
this approach is less complex than others which include radiative cooling ( e.g. * ? ? ?
* ) , heating by massive stars @xcite , or other feedback effects @xcite .
however , previous studies indicate that simple isothermal models capture much of the relevant dynamical behavior of interstellar gas in interacting galaxies , including large - scale shocks , nuclear inflows , and the formation of extended gas disks in merger remnants ( @xcite ; @xcite ; @xcite ) . with the gas temperature fixed ,
there remain two interesting local quantities which are easily evaluated in an sph simulation : the gas density , @xmath1 , and the rate of mechanical heating due to shocks and @xmath21 work , @xmath22 .
( an isothermal model assumes that radiative processes locally balance mechanical heating or cooling . ) in galactic simulations , the gas temperature is much lower than the virial temperature , so most shocks are highly supersonic and shock heating exceeds @xmath21 work by about two orders of magnitude .
thus shocks can be identified as regions with @xmath22 significantly larger than zero ; this criterion serves to locate shock fronts @xcite . making use of both @xmath1 and @xmath22 , i consider a general star - formation prescription of the form @xmath23 where the max function , which returns the larger of its two arguments , is used to handle situations in which @xmath24 .
this prescription is implemented in a probabilistic fashion ; the chance of gas particle @xmath25 undergoing star formation in a time interval @xmath26 is @xmath27 where @xmath28 and @xmath29 , the gas density and heating rate for particle @xmath25 , are defined in appendix a. at each time - step , @xmath30 is computed for each gas particle and compared with a random number @xmath31 freshly drawn from a uniform distribution between @xmath32 and @xmath16 . if @xmath33 then the gas particle is instantly and completely converted to a star particle , which subsequently evolves collisionlessly . for later analysis , each star particle formed during a simulation is tagged with the value of the time @xmath34 at its moment of birth .
while ( [ sfr - two - power - law ] ) can encompass a wide range of rules , this paper focuses on two limiting cases .
first , setting @xmath35 and @xmath36 yields _ density - dependent _ star formation , in which @xmath37 with _ no _ dependence on @xmath22 , and regions of higher density produce more stars .
second , setting @xmath38 and @xmath39 yields _ shock - induced _ star formation , in which @xmath40 ; star formation is strictly limited to regions with @xmath41 , which typically correspond to shocks .
more general rules with both @xmath35 and @xmath39 are beyond the scope of this paper , but it may be possible to anticipate their behavior using experience gained with the simple cases described here .
simulations of shock - induced star formation are subject to resolution effects .
the width of a shock region in sph is @xmath42 times the local smoothing scale @xmath43 @xcite ( in these experiments , @xmath43 is set to enclose @xmath44 gas particles within a radius @xmath45 of each gas particle @xmath25 ; see appendix b ) .
consider a situation in which a gas particle encounters a shock of some fixed strength .
the time that the particle spends in the shock scales as @xmath43 , while the total amount of energy it dissipates is independent of @xmath43 , so @xmath46 .
thus the net probability that the gas particle is transformed into a star particle while passing through the shock is proportional to @xmath47 .
the specific choice @xmath48 should produce results which are approximately independent of the spatial resolution .
for other choices for @xmath49 the star formation rate will depend on resolution ; to obtain roughly equivalent results in simulations with different spatial resolutions , it will be necessary to adjust the parameter @xmath3 .
resolution effects also arise in simulations with density - dependent star formation rules .
such simulations do a good job of reproducing the observed relationship ( [ schmidt - kennicutt - law ] ) between gas density and star formation rate ( @xcite ; @xcite ) .
but current sph simulations generally do nt resolve the vertical structure of gas in galactic disks , so this success may be due , in part , to compensating errors at different distances from the disk midplane .
more critically , although shocks with mach numbers of @xmath50 to @xmath51 will boost the density of an isothermal gas by factors of @xmath52 to @xmath53 @xcite , these high - density regions are not resolved in existing sph calculations , and the localized bursts of star formation which should result have not been seen in the numerical models . in principle , simulations using density - dependent rules can exhibit shock - induced star formation , but the computing power required to achieve the necessary resolution may not yet be available .
on account of their rounded bodies and long tails , the two galaxies making up ngc4676 were dubbed the `` playing mice '' by @xcite .
the far left panel of fig .
[ fig01 ] shows a deep optical image , overlain by contours of h @xcite , while the adjacent panel shows a shallower image .
the individual galaxies are ngc4676a , to the north , and ngc4676b , to the southeast .
the length and straightness of the bright northern tail imply that we are viewing the aftermath of a nearly - direct passage roughly edge - on to the orbit and disk planes , while the curve of the fainter southern tail is consistent with a more face - on view of an inclined disk .
the fairly equal lengths of the two tails suggests that the galaxies involved had roughly equal masses .
along with long - slit data indicating that both `` hulks '' rotate with north receding ( @xcite ; @xcite ) , these considerations led tt72 to a simple model for this system .
their model required that the northern hulk recede faster than the southern one , a prediction soon confirmed with h@xmath54 spectroscopy by @xcite .
most kinematic studies , whether based on h@xmath54 spectroscopy @xcite , h interferometry @xcite , or co interferometry @xcite , yield results consistent with @xcite s . on the other hand , h@xmath54 spectroscopy by @xcite indicated a remarkably large difference of @xmath55 between the nucleus of ngc4676a and its associated tail , and a large velocity dispersion within the tail itself
; it s tempting to interpret this as measurement error , but @xcite reported somewhat similar results , so the situation is not entirely clear . nonetheless , h interferometry appears the best choice for tracing the large - scale dynamics of the system , inasmuch as it provides a full map of the velocity field , and neutral hydrogen is unlikely to be disturbed by star formation . the model of ngc4676 used in this paper is based on work with j. hibbard ( hibbard & barnes , in preparation ; see also @xcite ) .
in view of some concerns noted below , this model is somewhat preliminary , but it seems to describe the overall evolution of the system reasonably well
. the initial conditions for the individual galaxies each have three components : a bulge with a shallow cusp @xcite , an exponential disk with constant scale height ( @xcite ; @xcite ) , and a dark halo with a constant - density core , based on a `` gamma model '' ( @xcite ; @xcite ) .
density profiles for these components are @xmath56 where @xmath57 is the spherical radius , @xmath58 is the cylindrical radius in the disk plane , and @xmath59 is the distance from the disk plane .
numerical results are presented using a system of units with @xmath60 . in these units ,
the components have masses @xmath61 , @xmath62 , @xmath63 , and length scales @xmath64 , @xmath65 , @xmath66 , and @xmath67 . with these parameters ,
the disk has a rotation period of @xmath68 time units at @xmath69 .
scale factors relating simulation units to real physical quantities will be determined as part of the model - matching process . beginning with tt72 s model
, we ran experiments systematically varying the relative orbit and orientation of the two identical galaxies .
the following parameters produced the reasonably good fit to ngc4676 shown on the right and far right of fig .
[ fig01 ] .
first , the initial orbit was parabolic , with a pericentric separation @xmath70 .
second , the disks of ngc4676a and ngc4676b were initially inclined to the orbital plane by @xmath71 and @xmath72 .
third , the initial angle between each disk s line of nodes and the direction of separation at pericenter was @xmath73 and @xmath74 .
fourth , the system was evolved @xmath16 time unit past pericenter .
we used an interactive display program to rotate the model to best match the observed morphology and kinematics ngc4676 ; further details of the matching process will appear in a subsequent paper . comparing the model with the observations , the main discrepancy is the position angle of ngc4676b s bar ; optical images show the bar runs roughly nne ssw , while the bar in the model is roughly n s or even nnw sse .
the pattern speed of a bar depends on the details of the rotation curve , so this discrepancy might be resolved by adjusting the mass model used for ngc4676b .
it s also possible that ngc4676b s bar predates its encounter with ngc4676a , in which case the bar s current position angle would depend on its phase at pericenter @xcite .
however , the hypothesis that ngc4676b s bar was induced by the tidal interaction is more parsimonious and has not yet been tested by experiments using a range of mass models .
the time yielding the best match between the model and ngc4676 proved somewhat ambiguous ; by adjusting the viewing angle and scaling factors , acceptable matches to the tail morphology and kinematics were obtained for @xmath75 to @xmath76 time units after pericenter . at @xmath77 , a good match was obtained by equating one length unit to @xmath78 , and one velocity unit to @xmath79 .
this scaling assumes that ngc4676 s distance is @xmath80 , consistent with @xmath81 and ngc4676 s systemic velocity of @xmath82 @xcite .
one time unit then works out to @xmath83 , and pericenter occurred @xmath84 ago .
finally , one mass unit is @xmath85 . given this scaling ,
the initial parameters of the model galaxies can be computed .
for example , the initial disks have reasonable scale lengths @xmath86 and masses @xmath87 ; on the other hand , they have maximum circular velocities of @xmath88 , which seem rather high and might be repaired by increasing the halo scale radius @xmath89 .
since tt72 s work , ngc4676 has been modeled by several other groups .
@xcite essentially produced a self - consistent clone of tt72 s model , even adopting a short - period elliptical orbit like the one tt72 s test - particle methodology dictated ; they favored an earlier viewing time which matched the position angle of ngc4676b s bar but produced tails significantly shorter than those observed .
@xcite used @xcite s ( @xcite ) velocity data to construct a self - consistent model with a parabolic initial orbit . in most respects ,
their model is very similar to the one described here , although we favor a closer passage to increase the heft of the tidal tails .
@xcite produced a test - particle model , restricted to ngc4676a , which matched the general trend of their velocity data but not the large velocity dispersion they reported within the northern tail .
their model requires relatively massive dark halos to account for these high tail velocities . , the rate of star formation , while the bottom panel shows the mass of gas remaining ; both quantities are given in simulation units ( @xmath60 ) .
pericenter occurred at @xmath90 ; the horizontal bar labeled `` now '' indicates the range of times matching the morphology and kinematics of ngc4676 .
solid line : density - dependent star formation with @xmath9 , @xmath36 , @xmath91 .
broken lines : shock - induced star formation ; dashed is @xmath92 , @xmath93 , @xmath94 , dotted is @xmath38 , @xmath48 , @xmath95 .
note that the shock - induced models produce strong bursts of star formation at pericenter ( @xmath90 ) , while the density - dependent model exhibits little or no immediate response.,width=317 ] fig .
[ fig02 ] compares global star formation rates for three different simulations of ngc4676 .
to illustrate the range of possibilities , i contrast one simulation with density - dependent star formation ( solid line : @xmath9 , @xmath36 , @xmath96 ) and two simulations with shock - induced star formation ( dashed line : @xmath38 , @xmath93 , @xmath94 ; dotted line : @xmath38 , @xmath48 , @xmath95 ) .
all of these simulations were based on the initial orbit and disk orientations described in 3.1 ; the gas was initially distributed like the disk stars , and the gas mass in each disk was @xmath97 .
further numerical details are given in appendix b. in simulations with density - dependent star formation , setting the constant @xmath96 yielded a baseline star formation rate per disk of @xmath98 ( see appendix c ) ; at this rate , an undisturbed disk would consume its gas on a time - scale of @xmath99 time units ( @xmath100 ) . in the simulation with shock - induced star formation and @xmath93 ,
a similar baseline rate of @xmath101 was obtained by setting @xmath94 ; this gave reasonable behavior throughout the encounter .
on the other hand , it was necessary to adopt a much lower baseline rate of @xmath102 in the simulation of shock - induced star formation with @xmath48 to avoid using up most of the gas in the earliest stages of the encounter ; setting @xmath95 insured that this case ultimately consumed about as much gas as the other two .
[ fig02 ] shows , the density - dependent simulation predict that the star formation rate remains fairly constant until some time after the first passage .
eventually , as the gas encountered shocks and gravitationally interacted with the tidally induced bars in the stellar disks , it began accumulating in the centers of the galaxies , driving a fairly gradual increase in star formation . the two shock - induced simulations , on the other hand , produced sharp bursts of star formation as the disks interpenetrated ( @xmath103 ) , followed by brief lulls as the disks separated .
soon thereafter , star formation rates rose again as gas flows intersected in the perturbed disks ; this produced the broad peaks , centered on @xmath104 , accounting for most of the stars formed before @xmath105 .
as one might expect , the simulation with @xmath48 exhibited the most extreme form of this behavior , with a higher peak at @xmath90 and a deeper minimum immediately thereafter .
these various star formation histories have interesting observational implications . according to the density - dependent simulation ,
star formation in ngc4676 has just begun responding to the encounter , and the peak star formation rate will occur several hundred myr from now .
in contrast , the shock - induced models predict that we are observing the system a short time after the broad peaks of star formation at @xmath106 .
this difference in star formation history is fairly insensitive to the choice of @xmath3 , which influences the overall amplitude of star formation but does little to change the timing of starbursts .
[ fig03 ] contrasts the spatial distribution of star formation in density - dependent and shock - induced simulations .
the top row of images set the stage by showing distributions of old disk stars and gas , taken from a simulation without star formation .
the middle row shows distributions of old disk stars and star particles formed according to the density - dependent rule ( @xmath9 , @xmath36 , @xmath96 ) ; colors indicate ages of star particles , with red being youngest and blue oldest .
the bottom row presents results for a simulation with shock - induced star formation ( @xmath38 , @xmath93 , @xmath94 ) using the same color coding .
results for the other shock - induced simulation are not shown since they are very similar .
comparison of the middle and bottom rows of fig .
[ fig03 ] shows that the star formation rule clearly influences the distribution of newly - formed stars . in the density - dependent model ,
star formation was always strongly concentrated toward the central regions of the disks . some star formation did occur outside the centers , but it was relatively weak and short - lived .
in contrast , the shock - induced models produced large , spatially - extended bursts of star formation as the disks collided at @xmath90 , and relatively high levels of activity persisted in the bridge and tail regions until @xmath107 . by the last time ( @xmath108 ) shown in fig .
[ fig03 ] , products of star formation are much more widely distributed in the shock - induced simulations than in the density - dependent simulation .
rather interesting in this regard are the gas - poor features which extend to roughly five oclock from the upper galaxy and to eleven oclock from the lower galaxy .
these regions were swept free of gas during the interpenetrating encounter of the two disks . in the simulations with shock - induced star formation ,
some of this gas was converted to stars _ before _ losing much momentum ; as a result , star formed in the pericentric bursts are scattered throughout these gas - poor regions . by contrast , these regions are bereft of both gas _ and _ star - formation products in the density - dependent simulation .
[ fig04 ] presents views which approximate the observed morphology and kinematics of ngc4676 .
these views are plotted at time @xmath77 , but fairly similar results are found between @xmath109 and @xmath110 .
the simulations shown here are the same ones shown in the last figure , with the density - dependent case in the middle and the shock - induced one on the right . in both cases , most of the ongoing star formation
is now confined to central regions of the galaxies .
however , products of star formation are found throughout the tails and other tidal features , especially in the simulation with shock - induced star formation .
these results invite comparison with observations of past star formation in ngc4676 .
@xcite found that the spectrum of ngc4676 is dominated by a stars , and attributed this to `` rapid , widespread star formation that effectively ceased at least @xmath111 years ago '' ; this diagnosis is supported by more recent observations ( @xcite ; @xcite ; @xcite ; @xcite ) .
the shock - induced simulations produced just such bursts at first passage ( @xmath103 ) , and the products of these bursts are widely distributed throughout the tidal tails at later times .
in contrast , the density - dependent simulation had far less star formation outside the galactic centers , and the first passage did not induce a prompt burst of extended star formation .
ngc4676 has regions of ongoing star formation which are evident in h@xmath54 images ( @xcite , fig .
the central regions of the two galaxies are quite prominent in h@xmath54 ( @xcite ) ; central activity is seen in all the simulations .
outside of the centers of the galaxies , star - forming regions are observed in the straight tail extending to the north of ngc4676a .
none of the simulations presented here reproduce this phenomenon , although the shock - induced models come closer inasmuch as they show enhanced star formation in their tails until @xmath107 ; the density - dependent model seems less promising in this regard since star formation in its tails declines after first passage .
images also show ongoing star formation at the base of the curved tail to northeast of ngc4676b .
this region has a relatively high surface brightness and is conspicuously blue in the acs images ( e.g. @xcite ) .
it is nicely reproduced in the shock - induced simulations ; gas from the curved tail falls back into its original disk , and the resulting shock sustains a small , coherent patch of extra - nuclear star formation . on the whole ,
the simulations with shock - induced rules qualitatively match most of the recent large - scale star formation in ngc4676 .
eventually , this comparison can be made more precise by modeling the spectrophotometric evolution of star particles , but such refinements are beyond the scope of this paper . the balance between central and extended star formation can be quantified by measuring the mass in stars formed since the start of the interaction .
specifically , the production of stars was integrated from @xmath112 , which is when fig .
[ fig02 ] shows the first increase in interaction - induced star formation , until @xmath77 ; particles then found at radii @xmath113 from the centers of both galaxies were considered to be part of the extended tidal features .
table [ starmass ] list the results for all three experiments . at the epoch
matching the present configuration of ngc4676 , the shock - induced models have formed about twice as many stars as the density - dependent model ; of course , this ratio depends on the values chosen for @xmath3 . in the shock - induced simulations , @xmath114 percent of the newly - formed stars
are found within the tidal features , while the corresponding figure for the density - dependent simulation is only @xmath115 percent .
these percentages should be insensitive to reasonable variations of @xmath3 , and preliminary experiments with other parameter values support this expectation .
.mass of stars formed since start of interaction , expressed as a percentage of initial gas mass .
[ cols= " > , > , > , > , > " , ] tables [ burstradii ] and [ burstmass ] summarize properties of the starbursts produced in these simulations . here
a burst is defined as all stars formed in a certain time interval ; these intervals bracket various peaks of @xmath116 in fig .
[ fig05 ] . for the shock - induced simulations ,
burst `` b@xmath117 '' corresponds to @xmath118 , `` b@xmath119 '' to @xmath120 , `` b@xmath121 '' to @xmath122 , and `` b@xmath123 '' to @xmath124 . for the density - dependent simulation , `` b@xmath119 ''
corresponds to @xmath125 , and `` b@xmath123 '' to @xmath126 . in effect , bursts b@xmath117 and b@xmath121 occur when the galaxies interpenetrate at first and second passages , respectively , b@xmath119 is the broad peak of star formation in the individual galaxies after their first passages , and b@xmath123 is the final merger - induced starburst .
the last column in each table , labeled `` all '' , gives figures for all stars formed during these simulations ( @xmath127 ) .
the radial distribution of burst populations in the final ( @xmath128 ) merger remnants is characterized in table [ burstradii ] . here , the center of each remnant @xmath129 was defined by locating the minimum of the potential well ; next , the median of @xmath130 for all particles @xmath25 in each burst population was computed ; common logarithms of these median radii are listed . on the whole ,
the burst populations in the shock - induced simulations are @xmath42 times more extended than those in the density - dependent simulation ; this ratio applies not only to the stars produced by bursts b@xmath119 and b@xmath123 but also to the entire population of stars formed during the simulations .
the extended distribution of star formation products in the shock - induced models may provide a better match to galaxies like ngc7252 , where @xcite found spectra dominated by a stars both within the nucleus and at a projected distance of @xmath131 .
in addition , the final b@xmath123 populations in the shock - induced simulations , while still compact , may be extended enough to moderate the central `` spikes '' of young stars that @xcite found in their simulated merger remnants .
table [ burstmass ] lists masses of stars formed .
these masses are given as percentages of the total initial gas mass , @xmath132 .
the two shock - induced simulations , by design , formed comparable amounts of stars , while the density - dependent simulation ultimately formed somewhat more .
burst b@xmath119 , triggered within the individual galaxies after their first passage , yielded @xmath133 percent of all stars formed in the density - dependent model , but only @xmath134 percent in the shock - induced models . the final burst b@xmath123 accounts for between @xmath135 percent and @xmath136 percent of the total star formation .
earlier merger simulations including shock - induced star formation have largely been based on `` discrete cloud '' models of the ism , making a direct comparison with the present results difficult . @xcite
presented cloud - collision rates for several mergers ; assuming that each cloud collision resulted in star formation , he predicted that merging galaxies undergo `` repetitive starbursts '' as the relative orbit of their nuclei decays .
the present calculations also produce bursts of star formation associated with pericentric passages , but orbit decay in these fully self - consistent calculations is much more abrupt than in @xcite s experiments , possibly because the self - gravity of the gas is included .
thus , while there is a good deal of temporal structure in the star formation rate , the nearly - periodic starbursts @xcite reported , which seem to reflect the gradual decay of elongated orbits in a nearly - harmonic core , are not seen here .
@xcite modeled the evolution of a spectrum of cloud masses during a merger , and reported a dramatic increase in the rate of disruptive cloud - cloud collisions within the inner @xmath137 of their model galaxies . assuming that star formation is associated with disruptive and glancing cloud - cloud collisions
, they obtained peak luminosities and luminosity - to - gas ratios typical of luminous ir galaxies .
density - dependent and shock - induced models of star formation yield qualitatively and quantitatively different results .
these differences arise despite the fact that the gas density @xmath1 and the dissipation rate @xmath22 are not completely independent variables ; indeed , they can be explained in terms of the global relationship between @xmath1 and @xmath22 .
dissipation is a necessary precursor for any significant increase in central gas density , since gas can only accumulate in galactic centers as a result of an irreversible process .
the rate of increase in gas density closely tracks the net energy radiated away in shocks ( @xcite , fig .
6 ) ; globally , @xmath22 and @xmath138 are strongly correlated .
thus , models of shock - induced star formation can respond promptly to external disturbances , while little activity occurs in density - dependent models until @xmath1 has had time to build up .
the larger spatial extent of star formation in shock - induced models is , in part , a corollary of the earlier onset of activity in such models , since the gas is more widely distributed at earlier times .
density - dependent rules may be used to implement unified models of star formation in normal and interacting galaxies .
the disks of normal galaxies _ and _ the central regions of starburst galaxies fit onto the same power law @xcite .
thus , by setting @xmath3 to match the baseline rate of star formation in unperturbed disks , @xcite were able to produce starbursts comparable to those inferred in ultraluminous infrared galaxies @xcite ; this represents a real success for the model defined by ( [ sfr - power - law ] ) with @xmath9 .
but there is abundant evidence for large - scale star formation in interacting galaxies .
this includes ongoing star formation in the `` overlap regions '' of systems like ngc4038/9 and arp299 , as well as the h@xmath54 emission from the tails of ngc4676 .
it also includes the a - star spectra seen in the tails of ngc4676 and throughout the body of ngc7252 .
density - dependent star formation ca nt easily explain these observations ; interactions funnel gas from the disks to the central regions of galaxies , thereby promoting rapid star formation in galactic nuclei but reducing the supply of gas needed for star formation elsewhere .
moreover , violent relaxation ceases long before binding energies are effectively randomized ( e.g. @xcite ; @xcite ) , so merging is ineffective at transporting the products of star formation outward from nuclei to the bodies of merger remnants .
thus , density - dependent rules offer at best an incomplete description of star formation in interacting galaxies . in simulations with shock - induced rules ,
the star formation rate depends on dynamical circumstances .
shocks in unperturbed disks are associated with transient spiral patterns , and matching the baseline level of star formation depends on reproducing the `` right '' level of spiral structure .
this , in itself , is a tough problem , and it s not clear that shock - induced models will soon yield a unified description of star formation in normal _ and _ interacting galaxies . until this becomes possible ,
unperturbed disks ca nt be used to set @xmath3 , so the amplitude of a shock - induced starburst ca nt be predicted _ a priori_. nonetheless , shock - induced star formation is an important element of galactic collisions , and its implementation in numerical simulations is a useful step toward increased realism .
these simulations have contrasted two limits of ( [ sfr - two - power - law ] ) : density - dependent star formation , with @xmath139 and @xmath36 , and shock - induced star formation , with @xmath38 and @xmath140 .
these cases were chosen as instructive examples , but ( [ sfr - two - power - law ] ) is general enough to accommodate additional possibilities .
first , setting @xmath35 and @xmath49 just slightly larger than @xmath32 would yield a modified law in which star formation is proportional to @xmath141 but occurs _ only _ in regions with @xmath41 ; this resembles the rule adopted by @xcite , who basically took @xmath142 but restricted star formation to regions with convergent flows .
notice that cases with @xmath143 are _ not _ continuous with the case @xmath36 ; in the latter , star formation is strictly independent of @xmath22 .
second , setting @xmath35 and @xmath39 would yield hybrid rules in which star formation depends on both @xmath1 and @xmath22 .
the consequences of such rules can sometimes be inferred from the limiting cases considered here .
for example , in the simulations of ngc4676 with shock - induced star formation , the gas involved in burst b@xmath121 has densities @xmath144 times higher than the gas involved in burst b@xmath117 ; setting @xmath35 would boost b@xmath121 relative to b@xmath117 by a factor of @xmath145 , assuming that neither burst was limited by the supply of gas .
perhaps most exciting are the new avenues for research created by an alternative description of star formation in galaxy interactions : * in contrast to density - dependent star formation , which seems fairly insensitive to most details of galactic encounters @xcite , different encounter geometries may yield a variety of shock - induced star formation histories .
a survey of different encounters , along the lines of other surveys without star formation ( e.g. @xcite ) , might establish if widespread star formation at first passage is generic or limited to a subset of close encounters , and also determine if shock - induced star formation can account for ultraluminous infrared galaxies .
* shock - induced star formation makes definite predictions about the _ timing _ of starbursts triggered by galactic encounters .
as already mentioned , models including the photometric evolution of starburst populations could sharpen the comparison between simulations and observations of systems like ngc4676 . in addition , these models might predict the ages of embedded young star clusters ; such predictions could be checked by multi - object spectroscopy of the clusters . * in the context of merger simulations , shock - induced star formation also yields predictions for the spatial distribution and kinematics of starburst populations .
it would be quite interesting to compare these predictions with observations of the distribution and kinematics of metal - rich globular clusters in elliptical galaxies , which may have been formed in merger - induced starbursts ( e.g. @xcite ) .
* encounters between gas - rich galaxies at redshifts of a few seem a natural setting for shock - induced star formation . while many high-@xmath59 galaxies have peculiar morphologies ( e.g. @xcite ) , the bridges and tails characteristic of low-@xmath59 encounters ( tt72 ) are not very evident . instead , the optical morphology of these objects may be dominated by rest - frame uv from widespread starbursts ; if so , the photometric modeling approach described above could help interpret existing and forthcoming observations of interacting galaxies at intermediate and high redshifts .
i thank john hibbard for permission to discuss our modeling of ngc4676 in advance of publication , and for helpful discussions .
i also thank chris mihos , lars hernquist , and especially francois schweizer for comments on this paper , and the referee for a prompt and helpful review .
i use an adaptive sph code in which each gas particle @xmath25 is assigned smoothing length @xmath43 which depends on local conditions .
the gas density at particle @xmath25 is estimated using the symmetric form suggested by @xcite , @xmath146 where @xmath147 is the particle mass and @xmath148 . here ,
@xmath149 $ ] , and @xmath150 is the spline interpolation kernel @xcite , which vanishes identically for @xmath151 . as discussed in 2 , the code uses an isothermal equation of state , so the internal energy @xmath152 of each gas particle is constant .
one may nonetheless compute the mechanical heating rate @xmath153 associated with the pressure and ( artificial ) viscous forces on each particle @xmath25 , and assume that this heating is _ exactly _ balanced by radiative processes .
this requires only a modest number of computations over and above those required to compute the forces themselves , and allows a global check of energy conservation .
the mechanical heating rate is estimated using @xmath154 where @xmath155 is the pressure at particle @xmath25 for an isothermal gas with sound speed @xmath156 . here
, @xmath157 approximates a combination of bulk and von neuman - richtmyer artificial viscosity @xcite , @xmath158 where @xmath54 and @xmath159 are parameters of order unity , @xmath160 @xmath161 , and @xmath162 is a parameter of order @xmath163 .
the disk galaxies used in the star formation simulations had a total of @xmath164 particles each , with @xmath165 for the gas , @xmath166 for the stellar disk , @xmath167 for the bulge , and @xmath168 for the halo .
the gas was initially distributed like the stellar disk , and accounted for @xmath169 per cent of the total disk mass . in simulation units ,
the gas had a fixed sound speed @xmath170 , mimicking a warm ism with @xmath20 .
these calculations were run with a sph code featuring adaptive smoothing and individual particle timesteps .
the smoothing radius @xmath43 of each gas particle was set so that a sphere of radius @xmath45 contains exactly @xmath44 gas particles .
a modified courant condition with @xmath171 was used to determine individual particle timesteps , which in some cases were as small as @xmath172 time units .
the artificial viscosity parameters appearing in ( [ sph - visc ] ) were @xmath173 and @xmath174 .
gravitational forces were calculated using a modified tree code with plummer smoothing @xcite and a smoothing length of @xmath175 .
collisionless particles were integrated with a fixed time - step of @xmath176 time units . including the energy radiated ( or absorbed ) by the gas in order to maintain a constant temperature , energy was conserved to about @xmath177 percent through the final merger of the galaxy models .
depending on the star formation parameters , each simulation took between @xmath178 and @xmath179 on a @xmath180 processor ; the run without star formation was the slowest since its gas attained the highest densities . in practice ,
the limited resolution of computer simulations introduces effects which complicate the interpretation of numerical experiments .
smoothing over a sphere of radius @xmath45 inevitably suppresses structures on smaller scales .
the gas disks used in this paper provide good examples ; for the simulation parameters given above , only half of the disk particles have smoothing radii @xmath43 less than the vertical scale height @xmath181 ! clearly , the vertical structure of these disks is not resolved and the same is true for most sph simulations of interacting galaxies published to date . in this circumstance , sph underestimates @xmath1 close to the midplane , and overestimates @xmath1 far from the midplane .
to better characterize the behavior of the galaxy models used in the simulations of ngc4676 , i ran some test calculations of isolated disks .
these can be used to establish the baseline rate of star formation in non - interacting disks , and to examine the radial distribution of star formation for various choices of @xmath182 and @xmath49 .
[ fig06 ] shows global star formation rates for isolated disk models like the ones used in the ngc4676 simulations . as in figs .
[ fig02 ] and [ fig05 ] , the solid lines show results for the density - dependent rule ( @xmath9 , @xmath36 , @xmath96 ) while the broken lines show results for shock - induced rules ( dashed line : @xmath183 , @xmath93 , @xmath94 ; dotted line : @xmath38 , @xmath48 , @xmath184 ) . since the last of these yields a very low baseline rate , an additional calculation was run with @xmath3 increased by a factor of @xmath185 ( dot - dashed line : @xmath38 , @xmath48 , @xmath186 ) . in every case ,
the isolated disk models proved stable for the duration of the calculation .
apart from the brief initial bursts in the shock - induced models due to imperfect initialization of gas velocities , the only trends seen are _ slight _ declines in global star formation due to gradual depletion of the gas .
; in addition , the dot - dashed line is @xmath38 , @xmath48 , @xmath186 .
note that here the top panel shows @xmath187 .
the simulations using shock - induced star formation produce brief initial bursts , but settle down promptly ; no other features are seen in these plots.,width=317 ] given the rather stable behavior of the isolated disk models , it s reasonable to measure radial star formation profiles by counting stars formed in concentric annuli .
[ fig07 ] plots @xmath4 as a function of cylindrical radius .
much as expected , the density - dependent rule yields a good approximation to the standard schmidt law with index @xmath188 .
less predictably , the shock - induced rules also approximate schmidt - type laws , albeit with somewhat shallower slopes .
this outcome must depend on the dynamical structure of the disks ; for a hypothetical and rather extreme example , a bar - unstable disk would presumably yield different star formation profiles before and after the bar formed .
a shock - induced model in which the average shock strength is independent of @xmath58 should yield a schmidt - type law with an index of unity , regardless of the chosen value of @xmath49 . here
, however , the shock - induced models with @xmath189 appear to yield slightly steeper @xmath4 profiles than does the model with @xmath93 , hinting that the shocks are somewhat stronger near the center of this disk . | star formation plays an important role in the fate of interacting galaxies . to date , most galactic simulations including star formation have used a density - dependent star formation rule designed to approximate a schmidt law . here , i present a new star formation rule which is governed by the local rate of energy dissipation in shocks . the new and old rules are compared using self - consistent simulations of ngc4676 ; shock - induced star formation provides a better match to the observations of this system .
[ firstpage ] galaxies : mergers galaxies : kinematics and dynamics galaxies : structure |
quantum field theory in curved space - time @xcite describes quantum fields propagating in a classical gravitational field background .
important processes described by the theory are vacuum polarization and particle creation in cosmological models and black - hole evaporation @xcite .
a black hole emits thermal radiation due to quantum effects with an effective temperature inversely proportional to its mass , @xmath0 .
the basic question that naturally arises is the following : is there a way to measure such a radiation in a suitable setup ?
a partial answer to this question was provided by unruh , who introduced the idea of studying analogous condensed matter systems which reproduce kinematical features of black - hole physics .
unruh showed @xcite that the propagation of sound waves in an irrotational and inviscid supersonic fluid is equivalent to the propagation of scalar waves in a black - hole space - time . since this seminal paper
, the possibility of simulating aspects of general relativity and quantum fields in curved space - time through analog models has been widely discussed in the literature @xcite . in particular , the possibility of constructing an acoustic analog of a black hole and measuring sound waves with thermal spectrum can provide an experimental verification of the existence of hawking radiation in a condensed matter setting . in this respect , there exist interesting recent proposals to generate an acoustic metric with sonic horizon in atomic bose - einstein condensates and other superfluids @xcite .
there is , however , a serious difficulty in the semiclassical picture underlying the derivation of the thermal spectrum of back - hole radiation . tracing the hawking radiation back in time ,
one has to undo an exponentially strong gravitational red - shift in the vicinity of the horizon .
this so - called trans - planckian problem can spoil the derivation of the hawking effect .
the gravitational back reaction also raises questions on the applicability of the semiclassical theory of gravity . within this perspective
, models were formulated @xcite with the aim of studying the effects of fluctuations of the black - hole horizon on the hawking radiation spectrum . in the absence of knowledge of the precise nature of the metric fluctuations near the horizon ,
the assumptions made here are that they can be treated classically and their effects on the propagation of quantum fields can be described via random differential equations .
we emphasize that this is quite different from the stochastic gravity program @xcite , where the einstein - langevin equation enables one to find the dynamics of metric fluctuations generated by the fluctuations on the stress tensor of quantum fields .
once established the analogy between black holes and fluids , it is a plausible approach to treat random media as models for fluctuations of the effective geometry of a sonic black hole . in such a scenario , recently an analog model for quantum gravity effects was proposed @xcite .
the model builds on the work of ford and collaborators @xcite and hu and shiokawa @xcite .
two general features of waves propagating in random fluids underlie the model .
first , acoustic perturbations in a fluid define discontinuity surfaces that provide a causal structure with sound cones .
second , propagation of acoustic excitations in random media are generally described by wave equations with random speed of sound @xcite . in ref .
@xcite , the quantum field theory of a scalar field associated with acoustic waves was analyzed in a situation where the speed of propagation of the acoustic wave , and hence the sound cone , fluctuates .
a stochastic ensemble of fluctuating geometries was assumed in that work . in the present paper
we take a pragmatic point of view towards seeking experimental consequences of quantum gravity effects in a specific scenario .
specifically , we analyze the question whether a quantum device can detect such effects by considering how fluctuations of a black - hole event horizon affect the transition rate of a two - level system which interacts with a massless scalar field . since we are not assuming the rotating wave approximation , the two - level system measures the vacuum noise in its world line .
the simplest assumption one can make for modeling the event horizon fluctuations is to assume a wave equation with random coefficients .
the differential equation governing the random wave propagation can not be solved in closed form , but it can be treated in perturbation theory @xcite , using as a small expansion parameter the intensity of the noise correlation function . using such a perturbative expansion ,
the wave multi - scattering processes can be interpreted in terms of feynman diagrams .
the positive - frequency wightman function is calculated at the one loop level of the noise averaging process . with this result in hand
, it is possible to calculate the distortion in transition probabilities of a two - level system induced by the fluctuating horizon .
although the horizon fluctuations do not invalidate the semiclassical derivation of the hawking effect , we show that predictions provided by the radiative processes in the semiclassical theory scenario may differ from those of this quantum gravity effect scenario , where the quantum fluctuations of the metric are treated using a stochastic ensemble of geometries . in this context , we would like to call the attention of the reader to the fact that hu and roura @xcite discussed a few years ago the possibility of studying the positive - frequency two - point wightman function in the presence of metric fluctuations and the response function of a detector coupled to the field .
also , a deviation from the thermal spectrum was found by takahashi and soda @xcite using a different model for a fluctuating black - hole horizon .
the organization of the paper is as follows . in section
ii we discuss quantum field theory in the presence of a schwarzschild event horizon .
we use the fact that , close to the horizon the schwarzschild metric takes the form of the rindler line element .
note that the rindler s line element is static , and consequently there is a straightforward way to define positive and negative frequency modes in order to impose the canonical quantization in rindler s spacetime .
with these considerations , we describe an apparatus device which is sensitive to fluctuations of the event horizon . in section iii
we discuss quantum fields in fluctuating disordered medium . in section
iv we present the distortion caused by the fluctuating horizon in the transition probabilities .
also , in this section , using topological arguments we show that fluctuations in the horizon do not change the temperature associated with the acceleration , but only the spectrum of the thermal radiation .
finally , section v contains our conclusions . to simplify presentation we assume units such that @xmath1 .
our aim is to discuss a particular model for fluctuations of the black - hole event horizon .
we are interested to know how such fluctuations can affect the thermal radiation due to the presence of the event horizon .
therefore , let us consider the line element of a four - dimensional schwarzschild space - time which describes a non - rotating uncharged black hole of mass @xmath2 : @xmath3 where @xmath4 is the metric of a unit @xmath5-sphere .
close to the horizon , @xmath6 .
therefore eq .
( [ det1 ] ) can be written as @xmath7 where @xmath8 . in these
coordinates the horizon is at @xmath9 .
the quantity @xmath10 describes the line element of a @xmath5-sphere of radius @xmath11 .
the other two terms can be identified with the line element of the two - dimensional rindler edge by setting @xmath12 and @xmath13 , for @xmath14 and @xmath15 . then : @xmath16 the null asymptotes @xmath17 , @xmath18 act as event horizons .
note also that lines of constant @xmath19 are hyperbolae , hence they represent the world lines of uniformly accelerated observers .
one sees that , close to event horizon the schwarzschild metric takes the form of the rindler line element .
therefore , in order to capture the essential physical features of such a situation , we consider an uniformly accelerated pointlike two - level system in a minkowski space - time with light - cone random fluctuations .
although out of the scope of the present publication to answer the important question what is a detector and what is the phenomenon of detecting particles " , it nevertheless requires some discussion .
it is a known fact that there is a conceptual problem in quantum field theory in the construction of the hilbert space of particles when particles are observer - dependent .
the situation can be clarified to some extent by considering the response of an accelerated detector . in the context of quantum optics , glauber @xcite , loudon @xcite , and nussenzveig
@xcite proposed ideal photocounter detectors .
afterwards , unruh @xcite and dewitt @xcite proposed scalar - particle detector models . crudely speaking ,
the so - called unruh - dewitt detector " is a two - level system with nonzero matrix elements of a monopole operator .
the detector has the feature that its response to an interaction with a scalar field in the minkowski vacuum depends on its state of motion .
when in inertial motion , the detector has a vanishing asymptotic probability to wind up in an excited state , while if it moves with a constant proper acceleration , it has a finite asymptotic probability to undergo transition to an excited state .
moreover , the accelerated detector , with proper acceleration @xmath20 , interacting with the scalar field in the minkowski vacuum is equivalent to the situation of the detector in inertial motion but in contact with a bath of thermal radiation at the temperature @xmath21 . following these results
, many papers appeared in the literature studying such a detector in many different situations .
for the reader interested in more details , we recommend ref .
@xcite ( and references therein ) , in which the authors studied the unruh - dewitt detector in a situation where the probability of its excitation is evaluated over a finite time interval . after this brief digression ,
let us describe our idealized model .
in this paper we consider an unruh - dewitt detector ; a two - level system coupled via a monopole interaction with a massless scalar field .
we will be working in four - dimensional minkowski space - time , whose line element is given by : @xmath22 in order to find the distortion caused by the fluctuating event horizon in the decay and excitation rates of a quantum system , let us discuss the response function of a two - level system .
let @xmath23 and @xmath24 be energy eigenstates of the system , with eigenenergies @xmath25 and @xmath26 , and @xmath27 the gap between the two states .
next , we suppose that the system is weakly coupled to a real scalar field with interaction lagrangian @xcite : @xmath28 where @xmath29 is the world line of the two - level system parametrized by the proper time @xmath30 , and @xmath31 is the monopole - moment operator of the two - level system . the quantity @xmath32 is a small coupling constant between the detector and the scalar field .
it is clear that this is an oversimplified model , with an atom represented by a two - level system that interacts with a real massless scalar field .
however , this model contains all the properties needed to understand the basic features of radiative processes of atoms near a fluctuating event horizon .
we can define an initial state at @xmath33 given by @xmath34 and a final state @xmath35 , at time @xmath30 . here , where @xmath36 and @xmath37 are the vacuum and final states of the field .
in first order approximation perturbation theory in the monopole coupling constant @xmath32 , eq .
( [ nova1 ] ) , the transition probability is given by : @xmath38 where @xmath39 is the selectivity of the two - level system , and @xmath40 is the response function : @xmath41 using the definition of the positive - frequency wightman function : g^+(x , x ) = 0|(x)(x)|0 , and introducing the variables @xmath42 and @xmath43 , the response function can be rewritten as : @xmath44 for a free massless scalar field , one has @xmath45 , and letting @xmath46 , the double integration would reduce to a fourier transform of the wightman function , times an infinite time integral .
similarly , we shall show later on in the paper that random fluctuations of the light cone introduce an unbounded function @xmath47 that depends on the functional form of the noise correlation function . however , the transition probability per unit proper time should be finite .
such circumstances often arise in quantum field theory and may be dealt with by adiabatically switching off the coupling as @xmath18 .
to circumvent such a problematic situation , we assume the field - detector interaction occurring during a finite time interval and , because of this , we choose to evaluate the response function over a finite proper time interval .
one should recall that the detector defined above responds to the vacuum fluctuations because we do not assume the rotating wave approximation .
the two - level system is measuring the vacuum noise in its world line .
consequently @xmath40 defines the spectrum of the vacuum noise .
determination of @xmath40 requires the positive frequency wightman function @xmath48 . in the next section
we determine @xmath48 and discuss the consequences of a disordered medium on the response function @xmath40 .
in order to access the modification caused by the fluctuating event horizon on transition probabilities , we implement a perturbation calculation similar to the one used in the context of problems of fluctuating disordered media and discussed in refs .
let us consider the random massless scalar klein - gordon equation in a four dimensional space - time , given by : @xmath49 \varphi(t,{\bf r})=0 , \label{nami20}\ ] ] where @xmath50 is the three dimensional laplacian .
note that @xmath51 works like a local refractive index , in that it perturbs locally the wave speed due to the replacement : ( 1+(*r * ) ) .
[ random - speed ] for the random function @xmath52 we will take a zero - mean gaussian random function : @xmath53 with a white - noise correlation function given by : @xmath54 where @xmath55 gives the intensity of random fluctuations .
the symbol @xmath56 denotes an average over all possible realization of this random variable . on the other hand , in principle
it is possible to extend the method to colored and/or non - gaussian noise functions .
note that we assume a time - independent random function for inertial observers .
however , in the accelerated detector world line @xmath57 becomes a rindler time - dependent random function .
we will discuss this point later on in the paper .
following refs .
@xcite , the random klein - gordon equation of eq .
( [ nami20 ] ) can be solved using a perturbation expansion in the noise function . in this way
, the positive- frequency wightman function can be written as : @xmath58 where @xmath59 is the usual positive - frequency wightman function without random fluctuations , and @xmath60 with @xmath61 being the random differential operator : l_1 ( x ) = l_1(t,*r * ) = - ( * r * ) .
[ l1-coord ] in eq .
( [ genericterm ] ) , it is to be understood that @xmath62 and that there is no integration in @xmath63 .
details on the derivations of the above expressions can be found in ref .
@xcite . due to the gaussian nature of the noise averaging
, higher order correlation functions of the form @xmath64 can be easily expressed as the sum of products of two - point correlation functions corresponding to all possible partitions of @xmath65 .
an interesting feature of wave propagation in random media is anderson localization @xcite . in this context
, we note that truncation of the series in eq .
( [ genericterm ] ) at a finite order @xmath66 will miss the singular aspect of the localization problem , which is of a non - perturbative nature see refs .
@xcite for discussions on this and related subjects . for our purposes in the present paper it is sufficient to use only the first terms of the series . after performing the averages over the noise function , the connected two - point positive - frequency wightman function associated with the massless scalar field can be written in the form of a dyson equation : @xmath67 with @xmath68 being the self - energy .
the one - loop contribution to @xmath68 is obtained from noise averaging the second order contribution in the random function @xmath57 .
specifically , one can write : @xmath69 and going over to fourier space : @xmath70 allows us to write : @xmath71 with @xmath72 being : g^+_0(,*k*)=i/(^2-*k*^2 ) , [ g0 ] and @xmath73 where the quantity @xmath74 is given by : @xmath75 plane for the positive - frequency wightman function . ]
using these results in eq .
( [ ft - g1 ] ) , one obtains the correction due to the random fluctuations to the wightman function in terms of the integral : @xmath76 the integral over @xmath77 can be done considering contour of integration in fig .
( [ contornok0 ] ) .
the integrand in eq .
( [ g11 ] ) has a second - order pole in @xmath78 .
the final result , after performing the integration @xmath79 , is : @xmath80 ^ 5
} \biggl [ ( \delta t)^5 \nn \\[0.3true cm ] & & + \ , 10 \ , ( \delta t)^3 \ , ( \delta { \bf x})^2 - 3 \ , \delta t ( \delta { \bf x})^4 \biggr ] , \label{g1}\end{aligned}\ ] ] where @xmath81 and @xmath82 .
now we are in position to take this result over different trajectories of the unruh - dewitt detector .
this will be done in the next section .
having obtained the two - point positive - frequency wightman function associated with the massless scalar field , it can be used to find the response function of the two - level system : @xmath83 where @xmath84 can be obtained from eq .
( [ gg ] ) and ( [ g1 ] ) fig .
[ int ] illustrates the coordinate system for the integration over @xmath85 and @xmath86 .
next , one shall consider the unruh - dewitt detector moving in @xmath87 plane along a hyperbolic trajectory : x = ( t^2 + ^-2)^1/2 , y = z = 0 , [ hyperb ] with @xmath20 being a constant .
as well known , this represents a detector accelerating uniformly with acceleration @xmath20 in the frame of the detector .
the detector s proper time @xmath88 is related to @xmath87 by the relations : @xmath89 and @xmath90 and @xmath86 . ]
performing such a coordinate transformation in eq .
( [ gg ] ) , one can write the response function to one - loop order as a sum of three contributions : @xmath91 the first term is the usual thermal contribution and the second is due to the switching on and off of the coupling between the two - level system and the scalar field .
we are interested in the third term since it contains the correction due to fluctuating event horizon .
the second term vanishes when large time intervals are considered . using the techniques developed in ref .
@xcite , the first term in eq .
( [ dis7 ] ) reads : @xmath92 , \label{fbeta}\end{aligned}\ ] ] where @xmath93 .
the transition rate , i.e. the probability of spontaneous and induced decay and excitation per unit time , of the two - level system given by : r_(e ) = f_(e , ) , [ rate ] can be readily computed . clearly , the result of eq .
( [ fbeta ] ) is that one has the same effect that of a bath of thermal radiation at a temperature @xmath94 see refs .
@xcite . within the perspective of the thermalization theorem ,
the result can be put in the following form : the pure state which is the vacuum from the point of view of an inertial observer is a canonical ensemble from the point of view of a uniformly accelerated observer , with a temperature proportional to the magnitude of the observer s acceleration .
the contribution @xmath95 due to the fluctuating event horizon is given by @xmath96 and can be evaluated in the same way as @xmath97 . performing the coordinate transformations given in eqs .
( [ t ] ) and ( [ x ] ) in eq .
( [ g1 ] ) , one finds : @xmath98 where we have absorbed a positive function of @xmath85 and @xmath86 into the infinitesimal parameter @xmath99 . also , the quantity @xmath100 that gives the intensity of the horizon fluctuations and is given by : @xmath101 the @xmath85 dependence comes from the fact that in the wightman function , @xmath102 is not an independent variable ; rather , it is determined by the detector s trajectory .
in other words , inertial observers in this model experience static light - cone random fluctuations see eq .
( [ nami20 ] ) .
however , for uniformly accelerated observers , such fluctuations will not be static anymore and will also depend on their proper times . on physical grounds , it is clear that for an uniformly accelerated detector the effects of the fluctuations will increase with its proper time , a result that is manifest in eq .
( [ mod ] ) . using the result of eqs .
( [ g1-eta - zeta ] ) and ( [ mod ] ) in eq .
( [ ult ] ) , one obtains the expression : @xmath103 & & \times \ , \frac { \sinh [ 5\alpha ( \tau - |\zeta|)/2 ] } { \sinh^5 ( \alpha\zeta/2 - i\epsilon\alpha)}. \label{integral}\end{aligned}\ ] ] we may express the integral above as @xmath104 & & - \ , \left[\int_{-\infty}^{-\tau}\,d\zeta f(\zeta)\ , + \,\int_{\tau}^{\infty}\,d\zeta f(\zeta ) \right ] , \label{trick}\end{aligned}\ ] ] where @xmath105}{\sinh^5(\zeta/2)}.\ ] ] the last two - terms on the right - hand side of eq .
( [ trick ] ) can be expressed as a single integral .
they have the same physical origin as the term @xmath106 .
in order to perform the integral @xmath107 we may use contour integration the contour to be used is shown in fig .
( [ contorno ] ) .
the integral over the lower part of the contour yields @xmath108 while that over the upper part yields @xmath109 .
the sum of these contributions is related to the fifth - order residue of @xmath110 at @xmath111 . finally ,
collecting these results , we have that f_1(e , ) = w(e , ) + h(e , ) , [ f1-sum ] where @xmath112 and @xmath113 & & \times \,\frac{\sinh \left[5\alpha(\tau - |\zeta|)/2 \right ] } { \sinh^5 \left[\alpha\zeta/2 \right]}. \label{r2noinercial2}\end{aligned}\ ] ] in eq .
( [ r2noinercial ] ) we consider @xmath30 as a small quantity , so that @xmath114 . . ]
eqs ( [ f1-sum])-([r2noinercial2 ] ) comprise our main result in this paper .
we have found that the correction due to horizon fluctuations gives a thermal distribution with a temperature @xmath115 , that is the same temperature as for the non - fluctuating case , _ but _ the distribution is of fermi - dirac form .
this resembles the result obtained by takagi @xcite , who studied the power spectrum of the vacuum noise measured by an accelerated detector in arbitrary dimensions and discussed the phenomenon of inversion of statistics in odd dimensions .
nevertheless , as emphasized by the literature @xcite , this is an apparent inversion of statistics . in our case
we have a fermi - dirac correction to the thermal radiation of a bosonic field , but we still have the expected bose - eintein distribution as the leading component of the radiation .
the meaning of our result is that horizon fluctuations imply in a radiation spectrum of both contributions , in that the usual bose - einstein distribution is perturbed by a fermi - dirac distribution of the same temperature .
one could expect that higher loop corrections would give additional energy - dependent terms that will be neither fermi - dirac nor bose - einstein forms .
it is easy to see from our result how the fluctuating horizon will change the transition rate of a detector undergoing an inertial trajectory by taking the zero proper acceleration limit in eq .
( [ r2noinercial ] ) : @xmath116 in the case of an inertial detector the intensity of the fluctuations will not change with time . the correction to the transition rate for the inertial detector is proportional to @xmath117 , meaning that for an inertial detector we have only spontaneous decay induced by the vacuum fluctuations .
let us study the dependence of the temperature with random fluctuations in more detail
. we will be using arguments developed by christensen and duff @xcite .
the euclidean manifold associated to inertial observers has the topology of @xmath118 , with euler - poincar characteristic @xmath119 . for the case of the uniformly accelerated observer
, we have a non - simply connected manifold , with topology @xmath120 , with euler - poincar characteristic @xmath121 . now
if we have these two distinct topological situations , we can define two different vacua , one associated with inertial observers , defined by @xmath122 and another one associated with accelerated observers , defined by @xmath123 .
next , one can show that an accelerating observer regards the @xmath122 vacuum as a thermal state at temperature @xmath21 .
the two - point schwinger function associated with the massless scalar field for both cases can be defined as : @xmath124 this two - point function obeys @xmath125 where @xmath50 is now the four dimensional laplacian . for simplicity ,
let us take the case where the two points @xmath102 and @xmath126 belong to the accelerated world - line @xmath127 , @xmath128 and @xmath129 .
the generalization for two arbitrary points can be found in ref .
we have : @xmath130 and , since @xmath131 must be periodic in @xmath30 with period @xmath132 , we have : @xmath133 in the case where @xmath121 , paths winding around the origin have topologically distinct classes with winding number @xmath66 .
in this case we have @xmath134 using the kms condition , we have that the finite temperature schwinger function satisfies @xmath135 and using the identification @xmath136 we have that @xmath137 and @xmath138 .
we conclude that the ground sate of the accelerated observer is the @xmath123 vacuum state , relative to which the @xmath122 vacuum is a thermal state . in the presence of noise , it is easy to see that the above arguments can be used in the same way . recalling eq .
( [ p32 ] ) , we get : @xmath139 in the case where @xmath121 , we also have @xmath140 using the kms condition , we have that the finite temperature schwinger function satisfies @xmath141 and using the same identification @xmath21 we have that @xmath142 and @xmath143 .
therefore , the temperature associated with the acceleration remains the same , but the fluctuating horizon does change its emitted thermal radiation .
recently , an analog model for quantum gravity effects in a condensed matter scenario was proposed in ref . @xcite . in ref .
@xcite this discussion was extended to a more general case , namely to a massive real scalar field . in the present paper
such a model was used to study a massless scalar field near a four - dimensional schwarzschild black hole with fluctuations in the event horizon . using a perturbation theory similar to the one used in problems of fluctuating disordered media
, we obtain the two - point positive - frequency wightman function associated with a real scalar field . after performing the averages over the noise function , we discuss the thermal radiation near the fluctuating event horizon .
we obtained the modification of the transition probabilities caused by the fluctuating horizon on the decay and excitation processes .
we showed that horizon fluctuations imply that the usual bose - einstein distribution is perturbed by a fermi - dirac distribution of the same temperature .
our results are obtained by assuming that the mechanical disturbances caused by the radiative processes belongs to an energy scale much smaller than the mass of the accelerated two - level system .
this means that its world line does not change due to light - cone random fluctuations .
we conclude our discussions noting that the previous treatment can be presented using the fermi golden rule and the fact that the density of states per unit volume is given by @xmath144 , where @xmath145 is the retarded green function associated with the massless scalar field .
finally , we remark that random - matrix theory can be used to find how the spectral density near the fluctuating event horizon is modified . in a similar situation , in ref .
@xcite the transition rate of a two - level atom within a chaotic cavity was presented using random matrices .
the use of random matrices to find how the spectral density near the fluctuating event horizon is modified is under investigation by the authors n. f. svaiter would like to acknowledge the hospitality of the instituto de fsica terica , universidade estadual paulista , where part of this research was carried out .
g. menezes would like to acknowledge the hospitality of the physics and astronomy department of tufts university , where part of this research was carried out .
we would like to thank l.h .
ford , e. goulart and f. f. tovar for useful discussions .
this paper was supported by capes , cnpq and fapesp ( brazilian agencies ) .
99 n. d. birrel and p. c. w. davies,_quantum fields in curved space _
( cambridge iniversity press , cambridge , 1982 ) .
a. a. grib , s. g. mamayev and v. m. mostepanenko , _ vacuum quantum effects in strong fields _ , ( friedman laboratory publishing , st .
petesburg,1994 ) .
s. w. hawking , nature ( london ) * 248 * , 30 ( 1974 ) ; comm .
math phys .
* 43 * , 199 ( 175 ) .
w. g. unruh , phys .
. lett . * 46 * , 1351 ( 1981 ) .
w. g. unruh , phys .
d * 51 * , 2827 ( 1995 ) .
m. visser , phys .
lett . * 80 * , 3436 ( 1998 ) .
t. a. jacobson and g. e. volovick , phys .
d * 58 * , 064021 ( 1998 ) .
s. liberati , s. sonego , and m. visser , class .
* 17 * , 2903 ( 2000 ) .
r. balbinot , s. fagnocchi , and a. fabri , phys .
d * 71 * , 064019 ( 2005 ) .
m. cadoni and s. mignemi , phys .
d * 72 * , 084012 ( 2005 ) .
r. schtzhold , m , uhlmann , l. petersen , h. schmitz , a. friedenauer , and t. schatz , phys .
lett . * 99 * , 201301 ( 2007 ) . , edited by m. novello , m. visser and g. volovick ( world scientific , singapure , 2002 ) . , edited by r. schutzhold and g. w. unruh ( springer , berlin , 2007 )
. l. j. garay , j. r. anglin , j. i. cirac , and p. zoller , phys .
* 85 * , 4643 ( 2000 ) ; phys .
a * 63 * , 023611 ( 2001 ) . c. barcel , s. liberati , and m. visser , class .
. grav . * 18 * , 1137 ( 2001 ) .
p. o. fedichev and u. r. fisher , phys .
lett . * 91 * , 240407 ( 2003 ) . v. a. de lorenci and r. klippert , braz .
* 34 * , 1367 ( 2004 ) .
h. nakano , y. kurita , k. ogawa , and c. moon yoo , phys .
d * 71 * , 084006 ( 2005 ) .
s. giovanazi , phys .
* 94 * , 061302 ( 2005 ) .
r. schtzhold , phys .
97 * , 190405 ( 2006 ) .
s. wster and c. m. savage , phys .
a * 76 * , 013608 ( 2007 ) . c. barrabs , v. frolov , and r. parentani , phys .
d * 59 * , 124010 ( 1999 ) ; * 62 * , 044020 ( 2000 ) . r. parentani , phys . rev .
d * 63 * , 041503 ( 2001 ) .
b. l. hu and e. verdaguer , living rev .
relativity , * 7 * , 3 ( 2004 ) .
g. krein , g. menezes , and n. f. svaiter , phys .
lett . * 105 * , 131301 ( 2010 ) .
l. h. ford , phys .
d * 51 * , 1692 ( 1995 ) .
l. h. ford and n. f. svaiter , phys .
d * 56 * , 2226 ( 1997 ) ; * 54 * , 2640 ( 1996 ) .
h. yu and l. h. ford , phys .
d * 60 * , 084023 ( 1999 ) .
r. t. thompson and l. h. ford , phys .
d * 78 * , 024014 ( 2008 ) ; class .
25 * , 154006 ( 2008 ) .
h. yu , n. f. svaiter , and l. h. ford , phys .
d * 80 * , 124019 ( 2009 ) .
b. l. hu and k. shiokawa , phys .
d * 57 * , 3474 ( 1998 ) .
a. ishimaru , _ wave propagation and scattering in random media _
( academic , new york , 1978 ) .
, edited by p. sheng ( world scientific , singapure,1990 ) .
j. pierre fouque , j. garnier , g. papanicolaou , and k. solna , _ wave propagation and time reversal in randomly layered media _
( spingerscience+bussiness media , llc,2007 ) .
e. akkermans and g. montambaux _
mesoscopic physics of electrons and photons _ ( cambridge university press , new york , 2007 ) . e. arias , e. goulart , g. krein , g. menezes and n. f. svaiter , phys .
d * 83 * , 125022 ( 2011 ) .
b. c. hu and a. roura , phys .
d * 76 * , 124018 ( 2007 ) .
t. takahashi and j. soda , class .
* 27 * , 175008 ( 2010 ) .
r. j. glauber , phys .
rev . * 130 * , 2529 ( 1963 ) .
r. loudon , _ the quantum theory of light _ ( oxford university press , new york , 1983 )
. h. m. nusenzweig , _ introduction to quantum optics _
( gordon and breach , new york , 1973 ) .
w. g. unruh , phys .
d * 14 * , 870 ( 1976 ) .
b. s. dewitt , in _ general relatiuitiy : an einstein cen tenary survey _ , edited by s.
w. hawking and w. israel ( cambridge university press , cambridge , england , 1980 ) .
b. f. svaiter and n. f. svaiter , phys .
d * 46 * , 5267 ( 1992 ) , ibid phys .
d * 47 * , 4802 ( 1993 ) .
d. w. sciama , p. candelas and d. deutsch , adv .
phys . * 30 * , 327 ( 1981 ) ; v. l. ginzburg and v. p. frolov , sov .
usp . * 30 * 1073 ( 1987 ) ; l. h. ford , n. f. svaiter and m. l. lyra , phys . rev .
a * 49 * , 1378 ( 1994 ) ; l. c. b. crispino , a. higuchi and g. e. a. matsas , rev .
phys . * 80 * , 787 ( 2008 ) .
w. anderson , phys . rev . * 109 * , 1492 ( 1958 ) .
s. john , h. sompolinsky , and m. j. stephen , phys .
b * 27 * , 5592 ( 1983 ) .
s. john and m. j. stephen , phys.rev .
b * 28 * , 6358 ( 1983 ) .
s. john , phys .
* 53 * , 2169 ( 1984 ) ; t. r. kirkpatrick , phys.rev .
b * 31 * , 5746 ( 1985 ) .
e. akkermans and r. maynard , phys.rev .
b * 32 * , 7850 ( 1985 ) . b. shapiro , phys .
* 57 * , 2168 ( 1986 ) .
m. j. stephen , phys .
b * 37 * , 1 ( 1988 ) .
v. gurarie , phys .
94 * , 245502 - 1 ( 2005 ) .
p. c. w. davies , j. phys .
a * 8 * , 365 ( 1975 ) .
s. takagi , prog .
. phys . * 72 * , 505 , ( 1984 ) ; * 88 * , 1 , ( 1986 ) .
w. g. unruh , phys .
d * 34 * , 1222 ( 1986 ) .
s. m. christensen and m. j. duff , nucl .
phys . * 146 * , 11 ( 1978 ) .
n. f. svaiter and c. a. d. zarro , class .
* 25 * , 095008 ( 2008 ) .
misirpashaev , p. w. brouer and c. w. j. bennaker , phys .
lett . * 79 * , 1841 ( 1997 ) . | we consider a pointlike two - level system undergoing uniformly accelerated motion .
we evaluate the transition probability for a finite time interval of this system coupled to a massless scalar field near a fluctuating event horizon .
horizon fluctuations are modeled using a random noise which generates light - cone fluctuations .
we study the case of centered , stationary and gaussian random processes .
the transition probability of the system is obtained from the positive - frequency wightman function calculated to one loop order in the noise averaging process .
our results show that the fluctuating horizon modifies the thermal radiation but leaves unchanged the temperature associated with the acceleration . |
the need of nonperturbative methods in in quantum physics is obvious considering the many problems where strong coupling and/or binding effects render perturbation theory inadequate . in nonrelativistic many - body physics
variational methods are widely used under these circumstances while this is _ not _ the case in relativistic field theory .
however , feynman s successful treatment of the polaron problem @xcite shows that variational methods may also be used for a nonrelativistic field theory provided that the fast degrees of freedom can be integrated out and their effect properly taken into account in the trial action . in a previous paper @xcite
( henceforth referred to as ( i ) ) we have extended the polaron variational method to the simplest scalar field theory which describes heavy particles ( `` nucleons '' ) interacting by the exchange of light particles ( `` mesons '' ) . in euclidean space - time the lagrangian of the wick - cutkosky model is given by @xmath3 here @xmath4 is the bare mass of the nucleon , @xmath5 is the mass of the meson and @xmath6 is the ( dimensionfull ) coupling constant of the yukawa interaction between them . in the quenched approximation
the meson field can be integrated out and one ends up with an effective action for the nucleons only @xmath7 = \int_0^{\beta } d \tau \
> \frac{1}{2 } \dot x^2 - \frac{g^2}{2 } \int_0^{\beta } d \tau_1 \ > \int_0^{\beta } d \tau_2 \
> \int \frac{d^4 q}{(2 \pi)^4 } \
> \frac{1}{q^2 + m^2 } \ > e^ { \
> i q \cdot ( \>x(\tau_1 ) - x(\tau_2 ) \ > ) } \ > . \label{eff action}\ ] ] note that this is formulated in terms of trajectories @xmath8 of the heavy particle ( `` particle representation '' ) which are parametrized by the proper time @xmath9 and obey the boundary conditions @xmath10 and @xmath11 . to obtain the 2-point function one has to perform the path integral over all trajectories and to integrate over @xmath12 from zero to infinity with a certain weight .
it is , of course , impossible to perform this path integral exactly and , again following feynman , we have approximated it variationally by a retarded quadratic two - time action . in ( i ) we have proposed various parameterizations for the retardation function which enters this trial action and derived variational equations for the most general case when its form was left free .
the purpose of the present paper is to investigate numerically these parametrizations as well as to solve the variational equations .
this fixes the variational parameters which will be used to calculate physical observables in forthcoming applications .
one quantity which we evaluate in the present paper is the residue on the pole of the propagator .
another one is related to the well - known instability @xcite of the wick - cutkosky model : although the effective action ( [ eff action ] ) is very similar to the one in the polaron model the ground state of the system is only metastable .
this does not show up in any order of perturbation theory but , as we have demonstrated in ( i ) , the variational approach knows about it .
indeed an approximate solution of the variational equations has revealed that there are no real solutions beyond a certain critical coupling . in the present paper
we will find exact numerical values for this critical coupling and calulate the width of the unstable particle for couplings beyond it .
this paper is organized as follows : the essential points of the polaron variational approach are collected in section [ sec : polaron ] , while section [ sec : numerical ] is devoted to the numerical methods and results . in section [ sec : width ] we investigate the instability of the wick - cutkosky model in our variational method and determine analytically and numerically the width of the dressed particle .
the variational principle can also be applied away from the pole , which is explored in section [ sec : var 2point off pole ] and used to calculate the residue at the nucleon pole .
the main results of this work are summarized in the last section .
following feynman s treatment of the polaron problem we have performed in ( i ) a variational calculation of the 2-point function with the quadratic trial action @xmath13 \ > = \ > \int_0^{\beta }
d\tau \ > \frac{1}{2 } \dot x^2 \ > + \
> \int_0^{\beta } d\tau_1 \ > \int_0^{\tau_1 } d\tau_2 \ > f ( \tau_1 - \tau_2 ) \
> \left [ \ > x(\tau_1 ) \ > - \
> x(\tau_2 ) \
> \right ] ^2 \ > .
\label{general x feynman action}\ ] ] here @xmath14 is an undetermined ` retardation function ' which takes into account the time lapse occurring when mesons are emitted and absorbed on the nucleon . in actual calculations we rather have used the fourier space form @xmath15 where the @xmath16 are the fourier components of the path @xmath17 and the fourier coefficients @xmath18 are considered as variational parameters . the variational treatment is based on the decomposition of the action @xmath19 into @xmath20 and on jensen s inequality @xmath21 near @xmath22 the 2-point function should behave like @xmath23 where @xmath24 is the residue .
as was shown in ( i ) this requires the proper time @xmath25 to tend to infinity .
one then obtains the following inequality @xmath26 where @xmath27 is a finite mass into which the divergence of the self - energy has been absorbed and @xmath28 a variational parameter . for @xmath29 all discrete sums over fourier modes @xmath18 turn into integrals over the ` profile function ' @xmath30 and one finds @xmath31 \ > , \label{omega by a(e)}\ ] ] as well as @xmath32 \ > .
\label{pot}\ ] ] here we use the abbreviations @xmath33 and @xmath34 because @xmath35 behaves like @xmath36 and @xmath37 for small and large @xmath36 , respectively , we have called it a ` pseudotime ' . note that in eq .
( [ pot ] ) the particular renormalization point @xmath38 has been used to regularize the small-@xmath39 behaviour of the integrand . as we have shown in ( i ) the total result is , of course , independent of @xmath40 . the profile function @xmath41 is linked to the retardation function @xmath42 by @xmath43 in ( i ) we have studied the following parametrizations _ ` feynman ' parametrization : _ @xmath44 which leads to @xmath45 _ ` improved ' parametrization : _
@xmath46 which entails @xmath47 \ > .
\label{improved a(e)}\ ] ] in both cases @xmath48 are variational parameters whose values have to be determined by minimizing eq .
( [ var inequality for mphys ] ) . as well as the above parametrizations , it was possible not to impose any specific form for the retardation function but to determine it from varying eq .
( [ var inequality for mphys ] ) with respect to @xmath28 and @xmath41 .
this gave the following relations @xmath49
@xmath50 \nonumber \\ & & \hspace{1 cm } \cdot \ > e \
> \left ( m \mu(\sigma ) , \frac{\lambda m_{\rm phys } \sigma } { \mu(\sigma ) } , u \right ) \ > .
\label{var eq for a(e)}\end{aligned}\ ] ] together with eq .
( [ amu2(sigma ) ] ) they constitute a system of coupled variational equations which have to be solved . assuming @xmath51 and @xmath52 we have found in ( i ) an approximate solution which had the same form of the retardation function as the ` improved ' parametrization and exhibited the instability of the system beyond a critical coupling constant . in the general case
we can read off the variational retardation function from eq .
( [ var eq for a(e ) ] ) @xmath53 \ > e \
> \left ( m \mu(\sigma ) , \frac{\lambda m_{\rm phys } \sigma } { \mu(\sigma ) } , u \right ) \ > .
\label{var retardation function}\ ] ] obviously it has the same @xmath54-behaviour for small relative times as the ` improved ' parametrization ( [ improved retard func ] ) .
finally we mention that by means of the variational eq .
( [ var eq for a(e ) ] ) , one can find the following expression for the the ` kinetic term ' @xmath55 defined in eq .
( [ omega by a(e ) ] ) @xmath56 \nonumber \\ & \cdot & e \
> \left ( m \mu(\sigma ) , \frac{\lambda m_{\rm phys } \sigma } { \mu(\sigma ) } , u \right ) \ > \frac{\partial}{\partial \sigma } \left ( \frac{\sigma } { \mu^2(\sigma ) } \right ) \ > .
\label{omega var expressed by g^2}\end{aligned}\ ] ] this is demonstrated in the appendix and will be used in chapter [ sec : width ] . that the kinetic term @xmath55 can be combined with the ` potential term '
@xmath57 is a consequence of the virial theorem for a two - time action @xcite which the variational approximation fulfills .
in this section we will compare numerically the various parametrizations for the retardation function .
because we are primarily interested in an eventual application in pion - nucleon physics , we have chosen the masses and coupling constants appropriately .
of course , the model does not really give a realistic description of the pion - nucleon interaction as spin- and isospin degrees of freedom as well as chiral symmetry are missing .
in short , we use @xmath58 and the results are presented as function of the dimensionless coupling constant @xmath59 the relevant quantity for the physical situation is the strength of the yukawa potential between two nucleons due to one - pion exchange @xcite , which is approximately given by ( depending on the spin - isospin channel ) @xmath60 where @xmath61 is the pion - nucleon coupling . in the wick - cutkosky scalar model
the corresponding strength is just the dimensionless coupling constant @xmath62 that we have defined in eq .
( [ alpha ] ) .
it should also be remembered that a yukawa potential only supports a bound state @xcite if @xmath63 we have minimized ( cf .
( [ var inequality for mphys ] ) ) @xmath64 with the ` feynman ' ansatz ( [ feynman a(e ) ] ) and the ` improved ' ansatz ( [ improved a(e ) ] ) .
this minimization was performed numerically with respect to the parameters @xmath65 by using the standard cern program minuit .
the numerical integrations were done with typically @xmath66 gauss - legendre points after mapping the infinite - range integrals to finite range . for the `
improved ' retardation function we had to calculate @xmath67 and @xmath55 numerically .
tables [ table : var feyn ] and [ table : var improved ] give the results of these calculations .
we also include the value of @xmath68 although it does nt have a physical meaning : finite terms ( which , for example , arise when a different renormalization point is chosen ) can be either grouped with @xmath68 or with @xmath57 .
however , from the variational inequality ( [ minimization ] ) we see that @xmath68 is a measure of the quality of the variational approximation : the larger @xmath68 the better the approximation .
.variational calculation for the nucleon self - energy in the wick - cutkosky model using the ` feynman ' parametrization ( [ feynman a(e ) ] ) for the profile function .
the parameters @xmath69 obtained from minimizing eq .
( [ minimization ] ) are given as well as @xmath28 and the intermediate renormalized mass @xmath68 ( see eq .
( [ finite mass ] ) ) .
the last column lists @xmath70 . [ cols="^,^,^,^,^,^",options="header " , ] it is also possible to apply the variational principle away from the pole of the two - point function by varying eq .
( [ 2point expressed by f0,f1 ] ) .
this gives @xmath71 note that on mass - shell where @xmath72 vanishes the previous variational equations follow .
we will not elaborate on eq .
( [ var eq off ] ) further but only point out that the perturbative self - energy ( [ perturb self energy ] ) is _ not _ obtained from the off - shell variational equations ( [ var eq off ] ) . in the limit @xmath73 one rather finds @xmath74 + \frac{g^2}{4 \pi^2 } \ >
\int_0 ^ 1 du \
> \frac{p^2 + m_1 ^ 2}{m^2 ( 1-u ) - p^2 u^2}\ ] ] which is an expansion of eq .
( [ perturb self energy ] ) around @xmath75 ( or @xmath76 which is the same in lowest order perturbation theory ) .
the reason for this somehow unexpected result is the neglect of exponentially suppressed terms in deriving eq .
( [ 2point expressed by f0,f1 ] ) .
indeed it is easy to see that one obtains the correct perturbative self - energy only if the upper limit of the @xmath39-integral is kept at @xmath25 and not extended to infinity as we have done in deriving eq .
( [ 2point expressed by f0,f1 ] ) .
the difference is one of the many exponentially suppressed terms which we have neglected .
thus , the off - shell variational equations ( [ var eq off ] ) only hold in the vicinity of the nucleon pole and in order to investigate variationally the two - point function far away ( say near the meson production threshold @xmath77 ) one has to include consistently all terms which are exponentially suppressed in @xmath25 .
this is beyond the scope of the present work .
in the present work we have performed variational calculations for the ` wick - cutkosky polaron ' following the approach which was developed previously @xcite . we have determined different parametrizations as well as the full variational solution for the retardation function which enters the trial action . since the nucleon mass is fixed on the pole of the 2-point function the value of
the functional which we minimize is of no physical significance but only a measure of the quality of the corresponding ansatz .
this is in contrast to the familiar quantum - mechanical case where an upper limit to the ground - state energy of the system is obtained .
however , our calculation fixes the variational parameters with which we then can calculate other observables of physical interest .
one of these quantities was the residue on the pole of the propagator for which we have compared numerically the results of the variational calculations to first order perturbation theory in table table : residue . for small couplings
all results for the residue agree , since in this case the variational approach necessarily reduces to perturbation theory independent of the value of the variational parameters .
what is rather remarkable is that for larger couplings the three parametrizations of the profile function in our variational approach yield rather similar results , which are now of course different from the perturbative calculation .
as we have seen , the ` improved ' and ` variational ' actions have the same singularity behaviour , for small relative times , as the true action , so here one might expect some similarity in the results .
this is however not true for the ` feynman ' parameterization which has a rather different form , so its agreement with the other two is not preordained .
this similarity is also exhibited in tables [ table : var feyn ] and [ table : var improved ] for @xmath28 and @xmath78 , but of course not for the parameters @xmath79 and @xmath80 which enter the respective profile functions and which are ` gauge ' ( i.e. reparametrization)-dependent quantities .
also the critical coupling at which real solutions ceased to exist was nearly identical in all three parametrizations .
the similarity of the results for the different ans " atze presumably indicates that these results are not too far away from the exact ones .
we were not only able to determine the critical coupling but also to deduce qualitatively and quantitatively the width which the particle acquires beyond the critical coupling .
this was achieved by finding _
complex _ solutions of the variational equations first approximately by an analytic approach and then exactly by an iterative method which closely followed the analytic procedure .
although the present approach does not describe tunneling ( which we expect to render the system unstable even at small coupling constants but with exponentially small width @xcite ) the polaron variational method is clearly superior to any perturbative treatment in this respect .
we have concentrated mostly , although not exclusively , on the on - shell 2-point function , i.e. the nucleon propagator .
this corresponds to the limit where the proper time goes towards infinity .
it is possible , however , to go beyond the on - shell limit .
this was necessary , for example , for the calculation of the residue of the 2-point function in section [ sec : resi ] . nevertheless , the residue is a quantity which is calculated at the pole and thus only requires off - shell information from an infinitesimal region around it .
this has the effect that the variational parameters for the calculation of the residue are the same as the on - shell ones .
as one moves a finite distance away from the pole the variational parameters themselves become a function of the off - shellness @xmath81 ( see section [ sec : var eq off mass shell ] ) . in conclusion , we think that the present variational approach has yielded nonperturbative numerical results which look very reasonable and are encouraging . we therefore believe it worthwhile to try to extend it in several ways .
first , in a sequel to this work we will generalize the present approach to the case with @xmath82 external mesons and thereby study physical processes like meson production or meson scattering from a nucleon .
this can be done by employing the quadratic trial function whose parameters have been determined in the present work on the pole of the 2-point function .
such a ` zeroth order ' calculation is similar in spirit to a quantum mechanical calculation in which wave functions determined from minimizing the energy functional are used to evaluate other observables .
more demanding is the consistent ` first - order ' variational calculation of higher - order green functions as this requires the amputation of precisely the non - perturbative nucleon propagators which have been determined in the present work . that this is indeed possible
will be demonstrated in another paper in this series . of course
, finally we would like to apply these non - perturbative techniques to theories which are of a more physical nature . among these
one may mention scalar qed , the walecka model @xcite or qed .
the latter two will require introduction of grassmann variables in order to deal with spin in a path integral .
as such , this should not pose a fundamental problem
. a greater challenge , however , is to extend such an approach beyond the quenched approximation or to nonabelian theories where the light degrees of freedom can not be integrated out analytically .
* note added * + after completion of this work we became aware of the pioneering work by k. mano @xcite in which similar methods are applied to the wick - cutkosky model with zero meson mass .
mano uses the proper time formulation , the quenched approximation and the feynman parametrization for the retardation function to derive a variational function for the self - energy of a scalar nucleon ( the expression following his eq .
( 6.18 ) ) which is identical with our eq .
( [ var inequality for mphys ] ) after proper identification of quantities is made .
however , for minimizing the variational function mano sets ( in our nomenclature ) @xmath83 , expands to second order in @xmath84 and finds an instability of the ground state for @xmath85 . note that @xmath86 so that this translates into a critical coupling @xmath87 which is much smaller than the value which we obtain from the exact minimization .
in addition , in the present work we consider non - zero meson masses , employ more general retardation functions , and calculate residue and width of the dressed particle .
here we derive eq .
( [ omega var expressed by g^2 ] ) for the kinetic term @xmath55 when the variational equations are fulfilled .
we first perform an integration by parts in the definition ( [ omega by a(e ) ] ) of @xmath55 .
the slow fall - off of the variational profile function with @xmath89 @xmath90 leads to a contribution at @xmath91 @xmath92 \ > .
\label{omega with part int}\ ] ] we then write the variational equation ( [ var eq for a(e ) ] ) for @xmath41 in the form @xmath93 where @xmath94 \ > e \ > \left ( m \mu(\sigma ) , \frac{\lambda m_{\rm phys } \sigma } { \mu(\sigma ) } , u \right ) \>. \label{x(sigma)}\ ] ] the integration over @xmath89 can now be performed giving a factor @xmath95 due to eq .
( [ amu2(sigma ) ] ) . therefore we have @xmath96 = \ > - \frac{g^2}{16 \pi } \ > \int_0^{\infty } d\sigma \ > \frac{1}{\mu^2(\sigma ) } \ > x(\sigma)\ ] ] which is just one term in the expression ( [ omega with part int ] ) for @xmath55 . to get the other one we differentiate the variational equation for @xmath41 with respect to @xmath89 and observe that @xmath97 one has to be careful not to interchange the @xmath89-integration and the @xmath39-differentiation .
we therefore perform an integration by parts and obtain @xmath98\nonumber \\ & = & \frac{g^2}{16 \pi } \left [ \ > - \lim_{\sigma \to 0 } \frac{\sigma x(\sigma)}{\mu^2(\sigma ) } + \int_0^{\infty } d\sigma \ >
\frac{1}{\mu^2(\sigma ) } \ > \left ( 2 + \frac{\partial}{\partial \sigma } \sigma \right ) \ >
x(\sigma ) \right ] \ > .\end{aligned}\ ] ]
note that the boundary term at @xmath99 gives a contribution because of @xmath100 .
this contribution exactly cancels the term @xmath101 in eq .
( [ omega with part int ] ) . combining both terms for @xmath55 ( which do not exist separately due to the slow fall - off of @xmath41 ) we obtain @xmath102 & = & \frac{g^2}{16 \pi } \left [ - 1 + \int_0^{\infty }
d\sigma \ >
\frac{1}{\mu^2(\sigma ) } \ > \left ( \ > 1 - \frac{\partial}{\partial \sigma}\sigma \
> \right ) \
> x(\sigma ) \right ] \nonumber \\ & = & \frac{g^2}{16 \pi } \left [ - 1 + \int_0^{\infty } d\sigma \ >
x(\sigma ) \left ( 1 + \sigma \frac{\partial}{\partial \sigma } \ >
\right ) \frac{1}{\mu^2(\sigma ) } \ > \ > \right ] \end{aligned}\ ] ] from which eq . ( [ omega var expressed by g^2 ] ) follows . in the last line again an integration by parts has been performed but this time there is no contribution from the boundary terms . | for the scalar wick - cutkosky model in the particle representation we perform a similar variational calculation for the 2-point function as was done by feynman for the polaron problem .
we employ a quadratic nonlocal trial action with a retardation function for which several anstze are used .
the variational parameters are determined by minimizing the variational function and in the most general case the nonlinear variational equations are solved numerically .
we obtain the residue at the pole , study analytically and numerically the instability of the model at larger coupling constants and calculate the width of the dressed particle . * polaron variational methods in the particle representation of field theory : ii .
numerical results for the propagator * r. rosenfelder @xmath0 and a. w. schreiber @xmath1 @xmath0 paul scherrer institute , ch-5232 villigen psi , switzerland @xmath2 triumf , 4004 wesbrook mall , vancouver , b.c . ,
canada v6 t 2a3 pacs numbers : 11.80.fv , 11.15.tk , 11.10.st |
the existence and uniqueness of a self - interacting measure - valued diffusion that is either attracted to or repelled from its centre of mass is shown below .
it is natural to consider a super ornstein uhlenbeck ( sou ) process with attractor ( repeller ) given by the centre of mass of the process as it is the simplest diffusion of this sort .
this type of model first appeared in a recent paper of engl " ander @xcite where a @xmath0-dimensional binary brownian motion , with each parent giving birth to exactly two offspring and branching occurring at integral times , is used to construct a binary branching ornstein uhlenbeck process where each particle is attracted ( repelled ) by the center of mass ( com ) .
this is done by solving the appropriate sde along each branch of the particle system and then stitching these solutions together .
this model can be generalized such that the underlying process is a branching brownian motion ( bbm ) , @xmath1 ( i.e. , with a general offspring distribution ) .
we might then solve an sde on each branch of @xmath1 : @xmath2 for @xmath3 , where @xmath4 labels the @xmath5th particle of @xmath1 alive from time @xmath6 to @xmath7 , @xmath8 is the parent of @xmath5 and @xmath9 is the center of mass . here , @xmath10 is the population of particles alive from time @xmath6 to @xmath7 .
this constructs a branching ou system with attraction to the com when @xmath11 and repulsion when @xmath12 .
it seems reasonable then , to take a scaling limit of branching particle systems of this form and expect it to converge in distribution to a measure - valued process where the representative particles behave like an ou process attracting to ( repelling from ) the com of the process .
though viable , this approach will be avoided in lieu of a second method utilizing the historical stochastic calculus of perkins @xcite which is more convenient for both constructing the sou interacting with its com and for proving various properties .
the idea is to use a supercritical historical brownian motion to construct the interactive sou process by solving a certain stochastic equation .
this approach for constructing interacting measure - valued diffusions was pioneered in @xcite and utilized in , for example , @xcite . a supercritical historical brownian motion , @xmath13 ,
is a stochastic process taking values in the space of measures over the space of paths in @xmath14
. one can think of @xmath13 as a supercritical superprocess which has a path - valued brownian motion as the underlying process . that is , if @xmath15 is a @xmath0-dimensional brownian motion , then @xmath16 is the underlying process of @xmath13 .
more information about @xmath13 is provided in section [ sec2 ] .
it can be shown that if a path @xmath17 is chosen according to @xmath18 ( loosely speaking this is made rigorous below in definition [ dcamp ] ) , then @xmath19 is a brownian motion stopped at @xmath20 .
projecting down gives @xmath21 which is a ( supercritical ) super brownian motion .
a key advantage to projecting down and constructing measure - valued processes is that it is possible to use the historical stochastic calculus to couple different projections together ( and hence couple measure - valued diffusions ) .
one can sensibly define the ornstein uhlenbeck sde driven by @xmath22 according to @xmath18 as the solution of a stochastic equation .
let @xmath23 be borel measurable .
we say that @xmath24 is a solution to the strong equation [ se1 ] if the pair satisfies @xmath25 \\[-10pt ] \nonumber & & \mbox{(b)}\quad x_t(a ) = \int1(z_t\in a)k_t(dy)\qquad \forall a\in\b(\r ^d)\ \forall t\ge0,\end{aligned}\ ] ] where @xmath26 and @xmath27 are appropriately adapted .
we will henceforth call the projection @xmath26 the ordinary super ornstein uhlenbeck process . if @xmath11 , @xmath28 is attracted to the origin , and if @xmath12 it is repelled .
the approximate meaning of @xmath13-a.e . in ( a )
is that the statement holds for @xmath18-a.a .
@xmath22 , for all @xmath18 , @xmath29-a.s .
the exact definition is given in the next section .
the projection @xmath30 is a sou process with attraction to ( repulsion from ) the origin at rate @xmath31 .
intuitively , @xmath13 tracks the underlying branching structure and @xmath28 is a function transforming a typical brownian path into a typical ornstein uhlenbeck path .
note that in the definitions of [ se1 ] and [ se2 ] given below , part ( b ) is unnecessary to solve the equations .
it has been included to provide an easy comparison to the strong equation of chapter v.1 of @xcite .
for all the results mentioned in the remainder of this work , the standing assumption ( unless indicated otherwise ) will be that @xmath32 [ tsourep ] there is a pathwise unique solution @xmath24 to [ se1 ] .
that is , @xmath26 is unique @xmath29-a.s . and @xmath27 @xmath13-a.e .
furthermore , the map @xmath33 is continuous and @xmath26 is a @xmath34-super - critical super ornstein uhlenbeck process .
similar to the above , we establish a function of the path @xmath22 that is a path of an ou process with attraction ( repulsion ) to the com and project down .
let @xmath35 be borel measurable .
define ( @xmath36 ) as the solution of @xmath37 \\[-10pt ] \nonumber & & \textup{(b ) } \quad x'_t(a ) = \int1(y_t\in a)k_t(dy)\qquad \forall a\in\b(\r ^d)\ \forall t\ge0 , \ ] ] where the com is @xmath38 we will call the projection @xmath39 the super ornstein uhlenbeck process with attraction ( repulsion ) to its com , or the interacting super ornstein uhlenbeck process . note that by our definitions of @xmath26 and @xmath39 as solutions to @xmath40 , respectively , that ( [ efinmass ] ) is the same as saying that @xmath41 as these quantities equal that in ( [ efinmass ] ) . [ texuniq ] there is a pathwise unique solution to [ se2 ] . one could prove this theorem using a combination of the proof of theorem [ tsourep ] and a localization argument .
we find it more profitable however , to employ a correspondence with the ordinary sou process @xmath26 .
this correspondence plays a central role in the analysis of @xmath39 , and indeed reveals a very interesting structure : we have that for any @xmath42 , @xmath43 where @xmath44 is the com of @xmath26 , defined as @xmath45 the correspondence essentially says that the sou process with attraction ( repulsion ) to its com is the same as the ordinary sou process being dynamically pushed by its com . from this equation ,
a relation between @xmath46 and @xmath44 can be established : @xmath47 define @xmath48 and @xmath49 . as the goal of this work is to prove that @xmath50 has interesting limiting behavior as @xmath20 approaches infinity , ( [ ecoupl ] ) yields a method of approach :
show first that the time integral in ( [ ecoupl ] ) converges in some sense and establish limiting behavior for @xmath51 .
one then hopes to combine these two facts with ( [ ecoupl ] ) to get the desired result .
let @xmath52 be the event that @xmath18 survives indefinitely .
note that this implies that on @xmath52 , both @xmath26 and @xmath39 survive indefinitely by their definitions as solutions of the equations above .
let @xmath53 be the time at which @xmath18 goes extinct .
the next two theorems settle the question of what happens on the extinction set @xmath54 .
[ textinc ] on @xmath54 , @xmath55 and @xmath56 converge as @xmath57 , @xmath58-a.s .
, for any @xmath59 . [ textconv ] on the extinction set , @xmath54 , @xmath60 as @xmath61 a.s . , where @xmath62 and @xmath63 are @xmath14-valued random variables such that @xmath64 this last theorem is an analogue of the result of tribe @xcite for ordinary critical superprocesses
note that here it does not matter whether there is attraction or repulsion from the com .
the following three theorems for the attractive case ( @xmath11 ) form the main results of this work .
[ tsurvival ] on @xmath52 the following hold : if @xmath65 @xmath66 and this integral is finite almost surely . if @xmath67 , then @xmath68 and this quantity converges almost surely .
this says that the coms of ordinary and interacting sou process converge and is the result that allows us to fruitfully use the correspondence of ( [ ecoupl ] ) to show convergence of the interacting sou process .
the next theorem shows that the mass normalized sou process converges almost surely , which is a new result among superprocesses .
englnder and winter in @xcite have shown that this process converges in probability , and before them , englnder and turaev in @xcite shown convergence in distribution .
one expects a result of this sort to hold since in the particle picture conditional on survival , at large time horizons , there are a very large number of particles that move as independent ou processes , each of which are located in the vicinity of the origin .
thus , we expect that in the limit the mass will be distributed according to the limiting distribution of an ou process .
[ tsouconv ] suppose @xmath11 .
then on @xmath52 , @xmath69 where @xmath70 is the semigroup of an ornstein uhlenbeck process with attraction to @xmath71 at rate @xmath72 and @xmath0 is the vasserstein metric on the space of finite measures on @xmath14 ( see definition [ dvas ] ) .
it is possible to show that theorem [ tsouconv ] holds for a more general class of superprocesses .
if the underlying process has an exponential rate of convergence to a stationary distribution , then the above theorem goes through .
one can appeal to , for example , theorem 4.2 of tweedie and roberts @xcite for a class of such continuous time processes . using the correspondence of ( [ ecoupl ] )
, one can then show that the mass normalized interacting sou process with attraction converges to a gaussian distribution , centered at the limiting value of the com , @xmath73 .
[ tisouconv ] suppose @xmath11 . then on @xmath52 , @xmath74 where @xmath75 is the ou - semigroup at infinity , with the origin shifted to @xmath76 .
when there is repulsion , matters become more difficult on the survival set .
it is no longer clear whether there exists a limiting random measure , or what the correct normalizing factor is .
we can show however that in some cases the com of the interacting sou process still converges . that there should be a limiting measure comes from the fact that the ordinary repelling sou process has been shown to converge in probability by englnder and winter in @xcite to a multiple of lebesgue measure .
one may expect something similar to hold for the interacting sou process , given ( [ ecoupl ] ) .
unfortunately , the correspondence is rendered ineffectual in this case by part ( b ) of the following theorem .
[ trcom ] the following hold on @xmath52 if @xmath77 : the process , @xmath55 converges almost surely .
however , @xmath56 diverges exponentially fast .
that is , there is a random variable @xmath78 , such that @xmath79 this reveals an interesting byplay between @xmath26 and @xmath39 in the repelling case .
that is , if one fixes a compact set @xmath80 , then for the ordinary sou process , @xmath30 , @xmath81 is exponentially distant from the com of the process .
however , the com of @xmath39 will possibly lie in the vicinity of @xmath81 for all time .
therefore , one might expect that @xmath81 is charged by a different amount of mass by @xmath82 than @xmath30 , and thus we might need to renormalize @xmath82 differently to get a valid limit .
it is also possible that the limit for each case is different ( and not simply connected by a random translation ) . the proofs for these theorems and more are contained in the following sections . in section [ sec2 ] , we give some background information for the historical process @xmath13 and some rigorous definitions . in section [ sexpre ] , we prove theorems [ tsourep ] and [ texuniq ] and state some important preliminary results regarding the nature of the support of a supercritical historical brownian motion . these are consequently used to get moment bounds on the center of mass processes @xmath55 and @xmath83 .
we also derive a martingale problem for @xmath51 and @xmath84 . in section
[ sconvergence ] , we give the proofs of the convergence theorems mentioned above and in the final section give the proofs of technical results that are crucial to prove these .
we collect some terms below : [ cols= " < , < " , ] we take @xmath13 to be a supercritical historical brownian motion . specifically , let @xmath13 be a @xmath85-historical superprocess ( here @xmath86 is the @xmath0-dimensional laplacian ) , where @xmath87 constant , on the probability space @xmath88 .
here @xmath34 corresponds to the branching bias in the offspring distribution , and the @xmath89 to the variance of the offspring distribution . a martingale problem characterizing @xmath13
is given below .
for a more thorough explanation of historical brownian motion than found here , see section v.2 of @xcite .
it turns out that @xmath18 is supported on @xmath90 a.s . and
typically , @xmath18 puts mass on those paths that are `` brownian '' ( until time @xmath20 and fixed thereafter ) .
as @xmath13 takes values in @xmath91 , @xmath92 will denote integration over the @xmath22 variable .
let @xmath93 be the path - valued process associated with @xmath94 taking values in @xmath95 .
then for @xmath96 , if @xmath97 let @xmath98 , where the right - hand side denotes expectation at time @xmath20 given that until time @xmath99 , @xmath100 follows the path @xmath22 . the weak generator , @xmath101 , of @xmath100 is as follows . if @xmath102 we say @xmath103 if and only if @xmath104 is bounded , continuous and ( @xmath105)-predictable , and for some @xmath106 with the same properties as @xmath107 @xmath108 is a ( @xmath105)-martingale under @xmath109 for all @xmath110 if @xmath111 , we will say @xmath13 satisfies the historical martingale problem , [ hmp ] , if and only if @xmath112 a.s .
and @xmath113 @xmath114 \eqntext{\displaystyle\mbox{is a continuous $ ( \f_t)$-martingale}}\\ \eqntext{\displaystyle\mbox{with } \langle m(\phi)\rangle_t = \int_0^t k_s(\phi_s^2 ) \,ds\ \forall t\ge0 , \mbox { a.s.}}\end{aligned}\ ] ] using the martingale problem [ hmp ] , one can construct an orthogonal martingale measure @xmath115 with the method of walsh @xcite .
denote by @xmath116 , the @xmath117-field of @xmath118-predictable sets in @xmath119 .
if @xmath120 is @xmath121-measurable and @xmath122 then there exists a continuous local martingale @xmath123 with quadratic variation @xmath124 .
if the expectation of the term in ( [ emmeas ] ) is finite , then @xmath123 is an @xmath125-martingale .
[ dcamp ] let @xmath126 .
let @xmath127 denote the universal completion of @xmath128 . if @xmath129 is a bounded @xmath118-stopping time , then the normalized campbell measure associated with @xmath129 is the measure @xmath130 on @xmath131 given by @xmath132 where @xmath133 .
we denote sample points in @xmath134 by @xmath135 . therefore , under @xmath130 , @xmath136 has law @xmath137 and conditional on @xmath136 , @xmath22 has law @xmath138 .
[ dkae ] for two @xmath139-measurable processes @xmath140 and
@xmath141 we will say that @xmath142-a.e . if @xmath143 for all fixed times @xmath144 .
we say that @xmath24 is a solution to the strong equation [ se1 ] if both ( a ) and ( b ) of that equation are satisfied and where @xmath28 is an @xmath139-predictable process and @xmath30 is an @xmath145-predictable process .
we say @xmath146 is the solution to the stochastic equation [ se2 ] if it satisfies ( a ) and ( b ) of that equation and where @xmath147 is an @xmath139-predictable process , @xmath82 is an @xmath145-predictable process . unless stated otherwise
, @xmath24 will refer to a solution of [ se1 ] and @xmath146 to the solution of [ se2 ] . for
an arbitrary @xmath127-adapted , @xmath14-valued process @xmath148 , define the centre of mass ( com ) , @xmath149 , with respect to @xmath18 as follows : @xmath150 we also let @xmath151 note that as @xmath13 is supercritical , @xmath18 survives indefinitely on a set of positive probability @xmath52 , and goes extinct on the set of positive probability @xmath152 hence , we can only make sense of @xmath153 for @xmath154 where @xmath53 is the extinction time . let the process @xmath155 be defined for @xmath156 where @xmath157 possibly random .
@xmath158 is called a local martingale on its lifetime if there exist stopping times @xmath159 such that @xmath160 is a martingale for all @xmath161 .
the interval @xmath162 is called the lifetime of @xmath158 .
the following definition introduces a metric on the space of finite measures , which is equivalent to the metric of convergence in distribution on the space of probability measures .
[ dvas ] let @xmath0 denote the vasserstein metric on the space of finite measures on @xmath14 . that is , for @xmath163 finite measures , @xmath164 where @xmath165 .
proof of theorem [ tsourep ] although closely related , this does not follow automatically from theorem 4.10 of perkins @xcite where it is shown that equations like @xmath166 [ but with more general , interactive , drift and diffusion terms in ( a ) ] have solutions if @xmath13 is a _ critical _ historical brownian motion .
note that @xmath167 defines a @xmath168-historical superprocess .
let @xmath169 be the campbell measure associated with @xmath170 ( note that if @xmath129 is taken to be a random stopping time , this measure differs from @xmath130 ) .
the proof of theorem v.4.1 of @xcite with minor modifications shows that @xmath171 has a pathwise unique solution .
this is because ( k3 ) of theorem 2.6 of @xcite shows that under @xmath169 , @xmath172 is a brownian motion stopped at time @xmath129 and proposition 2.7 of the same memoir can be used to replace proposition 2.4 and remark 2.5(c ) of @xcite for the setting where the branching variance depends on time .
once this is established , it is simple to deduce that if @xmath173 is the solution of @xmath171 , and we let @xmath174 then @xmath175 is the pathwise unique solution of [ se1 ] .
the only thing to check this is that @xmath176 @xmath13-a.e . , but this follows from the fact that @xmath177 .
it can be shown by using by theorem 2.14 of @xcite that @xmath26 satisfies the following martingale problem : for @xmath178 , @xmath179 is a martingale where @xmath180 .
then by theorem ii.5.1 of @xcite this implies that @xmath26 is a version of a sou process , with initial distribution given by @xmath181 .
[ rsouhrep ] ( a ) under the lipschitz assumptions of section v.1 of @xcite , one can in fact uniquely solve [ se1 ] where @xmath13 is a supercritical historical brownian motion . the proof above can be extended with minor modifications .
the proof of theorem [ tsourep ] essentially shows that under @xmath130 , @xmath129 fixed , the path process @xmath182 such that @xmath183 is a @xmath0-dimensional brownian motion stopped at @xmath129 . under @xmath130 , @xmath129 fixed , @xmath28 can be written explicitly as a function of the driving path : for @xmath184 , @xmath185 \\[-8pt ] \nonumber & = & z_0 + \int_0^t e^{\gamma s } \,dy_s + \int_0^t e^{\gamma s } ( -\gamma z_s ) \,ds + \int_0^t z_s ( \gamma e^{\gamma s } ) \,ds,\end{aligned}\ ] ] where we have used a differential form of [ se1](a ) for the second equality .
hence , @xmath186 next , we show that there exists a unique solution to [ se2 ]
. proof of theorem [ texuniq ] suppose there exists a solution @xmath187 satisfying [ se2 ] .
then under @xmath130 , @xmath147 can be written as a function of the driving path @xmath22 and @xmath188 .
using integration by parts gives @xmath189 and hence @xmath190 if @xmath24 is the solution to [ se1 ] where @xmath191 , then note that by remark [ rsouhrep](c ) , @xmath192 by taking the normalized measure @xmath193 on both sides of the above equation , we get @xmath194 hence , @xmath195 is seen to satisfy a volterra integral equation of the second kind ( see equation ( 2.2.1 ) of @xcite ) and therefore can be solved pathwise to give @xmath196 which is easily verified using integration by parts . also , if @xmath197 is a second process which solves ( [ ecorr ] ) , then @xmath198 by gronwall s inequality , this implies @xmath199 , for all @xmath20 and @xmath136 .
pathwise uniqueness of @xmath200 follows from the uniqueness of the solution to [ se1 ] and the uniqueness of the process @xmath195 solving ( [ ecorr ] ) .
we have shown that if there exists a solution to [ se2 ] , then it is necessarily pathwise unique .
turning now to existence to complete the proof , we work in the opposite order and define @xmath187 and @xmath200 as functions of the pathwise unique solution to [ se1 ] where @xmath201 : @xmath202 then @xmath55 satisfies the integral equation ( [ ecorr ] ) , and hence @xmath203 therefore @xmath204 by equation ( [ ezpath ] ) , and so @xmath205 multiplying by @xmath206 and using integration by parts shows @xmath207 which holds for @xmath13-a.e .
@xmath22 , thereby showing @xmath208 satisfies [ se2 ] .
[ rcomrel ] some useful equivalences in the above proof are collected below .
if @xmath209 , then for @xmath154 , @xmath210 these equations intimately tie the behaviour of the interacting and ordinary sou processes .
part ( a ) says that the interacting sou process with attraction to the center of mass is the same as the ordinary sou process pushed by the position of its center of mass .
we now consider the martingale problem for @xmath39 . for @xmath211 , recall that @xmath212 and that the lifetime of the process @xmath213 is @xmath214 .
then the following theorem holds : [ tito ] for @xmath215 and @xmath216 , @xmath217 where @xmath218 and @xmath219 is a continuous local martingale on its lifetime such that @xmath220 and hence has quadratic variation given by @xmath221_t = \int_0^t\frac{\overline{\phi_s^2 } - ( \phib_s)^2}{k_s(1)}\,ds.\ ] ] similarly , the following is true .
[ rito2 ] the method of theorem [ tito ] can be used to show that for the @xmath34-supercritical sou process , x , for @xmath222 and @xmath223 , @xmath224 where @xmath225 and @xmath219 is a continuous local martingale on its lifetime such that @xmath226 and hence has quadratic variation given by @xmath221_t = \int_0^t\frac{\kb_s(\phi^2(z_s ) ) - \kb_s(\phi ( z_s))^2}{k_s(1)}\,ds.\ ] ] proof of theorem [ tito ] the proof is not very difficult ; one need only use it s lemma followed by some slight modifications of theorems in chapter v of @xcite to deal with the drift introduced in the historical martingale problem due to the supercritical branching .
let @xmath129 be a fixed time and @xmath184 . recall that under @xmath130 , @xmath22 is a stopped brownian motion by remark [ rsouhrep](b ) , and hence @xmath227 is a stopped ou process ( attracting to @xmath55 ) .
therefore , under @xmath130 : @xmath228_s \\ & = & \phi(y_0 ) + \int_0^t \nabla\phi(y_s)\cdot dy_s+ \int_0^t \gamma \nabla \phi(y_s)\cdot(\yb_s - y_s)+ \frac{1}{2}\delta\phi(y_s)\,ds\end{aligned}\ ] ] by the classical it s lemma .
then @xmath229\,dm(s , y)\\ & & { } + \beta\int_0^t k_s\biggl [ \int_0^s \nabla\phi(y_r)\cdot dy_r\biggr]\,ds\\ & & { } + \int_0^t\int b_s\ , dm(s , y)+ \beta\int_0^t k_s(b_s)\,ds + \int _ 0^tk_s(b_s)\,ds\\ & = & k_0(\phi(y_0))+ \int_0^t\int\phi(y_s)\,dm(s , y)\\ & & { } + \beta\int_0^t k_s(\phi(y_s))\,ds + \int_0^tk_s(b_s)\,ds.\end{aligned}\ ] ] the equality in the third line to @xmath230 follows from proposition 2.13 of @xcite .
the equality of the fourth line to the last term in the first line follows from a generalization of proposition v.2.4 ( b ) of @xcite .
the last equality then follows by collecting like terms and using the definition of @xmath46 .
note that @xmath231 is feller s @xmath34-supercritical branching diffusion and hence @xmath232 where @xmath233 is a martingale such that @xmath234_t = \int_0^t k_s(1 ) \,ds.$ ] therefore for @xmath154 , it s formula and properties of @xmath231 and @xmath235 imply that @xmath236\,dm(s , y)+\int_0^t\frac{k_s(b_s)}{k_s(1)}\,ds.\end{aligned}\ ] ] since @xmath104 is bounded , the stochastic integral term can be localized using the stopping times @xmath237 and hence it is a local martingale on @xmath238 . it is easy to check that it has the appropriate quadratic variation .
the following lemmas will be used extensively in section [ sconvergence ] , but will be proven in section [ stech ] .
[ lextprob ] there is a nonnegative random variable @xmath239 such that @xmath240 and @xmath241 almost surely .
note that as @xmath26 and @xmath39 are defined as projections of @xmath13 , their mass processes are the same as that of @xmath13 , and thus grow at the same rate .
let @xmath242 where @xmath243 .
let @xmath244 let @xmath13 be a supercritical historical brownian motion , with drift @xmath34 , branching variance @xmath89 , and initial measure @xmath245
. for @xmath246 fixed , @xmath247 , there exists a.s .
@xmath248 such that @xmath249 for all @xmath20 .
further , given @xmath250 , @xmath251 where @xmath252 as @xmath253 and for any @xmath254 , @xmath250 can be chosen large enough so that @xmath255 for @xmath256 $ ] .
the following moment estimates are useful in establishing the convergence of @xmath257 .
recall that @xmath53 is the extinction time of @xmath13 .
[ lsecmom ] assume @xmath258 .
then , @xmath259 where @xmath260 [ rmoments ] ( a ) the proof of lemma [ lsecmom ] , under the same hypotheses ( if @xmath261 ) yields @xmath262 lemma [ lsecmom ] and its proof can be extended to show that for any positive integer @xmath263 if @xmath264 polynomial in @xmath20 if @xmath265 , exponential if @xmath12 such that @xmath266
we will henceforth , unless specified otherwise , assume that for a path @xmath267 , @xmath268 . recall
that , by the construction of solutions to @xmath269 @xmath270 also recall that our standing hypothesis is that @xmath271 has finite initial mass [ and hence @xmath272 . in this section
, we will first settle what happens to @xmath26 and @xmath39 when there is extinction , and then the case when there is survival , under the attractive regime and lastly address the interacting repelling sou process on the survival set .
proof of theorem [ textinc ] assume for now that @xmath273 ( the case where @xmath274 can be ignored without loss of generality ) .
by theorem [ tito ] , @xmath275 and therefore is a local martingale on its lifetime with reducing sequence @xmath276 as defined in the proof of the same theorem . using doob s weak inequality and lemma [ lsecmom ] , @xmath277 by the first borel
cantelli lemma , @xmath278 it follows that @xmath279 which implies that on the set @xmath280 , @xmath281 converges , by theorem iv.34.12 of rogers and williams @xcite .
this shows convergence on the extinction set as @xmath282 .
note that if @xmath283 .
theorem ii.8.3 of @xcite gives @xmath284 where @xmath285 is the law of a historical brownian motion with initial distribution @xmath286 .
hence , the a.s . convergence of @xmath55 in the case where @xmath287 is finite in mean imply @xmath288 if @xmath289-a.s .
finally , to get rid of the assumption that @xmath290 note that corollary 3.4 of @xcite ensures that if @xmath291 , then at any time @xmath292 , @xmath18 ( and hence @xmath293 ) is compactly supported .
therefore , letting @xmath294 we see that @xmath295 by ( [ egen1 ] ) since @xmath296 a.s .
compact implies that @xmath297 holds .
this completes the proof for the convergence of @xmath46 on @xmath54 in its full generality .
the convergence of @xmath56 now follows from the convergence of @xmath55 and equation ( [ ecorr ] ) .
proof of theorem [ textconv ] as in the previous proof , note that we need only consider the case that @xmath273 .
we will follow the proof of theorem 1 of tribe @xcite here .
define @xmath298 it is known by the work of konno and shiga @xcite in the case where @xmath299 , that @xmath300 homeomorphically ( recall that @xmath301 a.s . in that case ) .
this latter result also holds when @xmath87 on the extinction set @xmath54 by a girsanov argument .
define @xmath302 as the unique inverse of @xmath303 ( on @xmath54 , this defines the inverse on @xmath304 ) and for @xmath305 , let @xmath306 .
let @xmath307 and define @xmath308 let @xmath309 where @xmath310 . then note that @xmath311 and each @xmath312 is a @xmath313-stopping time . on @xmath54 for @xmath314 ,
theorem [ tito ] implies @xmath315 since @xmath316 and where @xmath317 .
it follows that @xmath318 is a @xmath313-local martingale .
then , by theorem [ tito ] , @xmath319_{t\wedge t_n } & = & \int_0^{d_{t\wedge t_n } } \frac { \xb ' _ s(\phi^2 ) - \xb'_s(\phi)^2}{x_s(1)}\,ds \\ & = & \int_0^{t\wedge t_n } \xbd_s(\phi^2 ) - \xbd_s(\phi)^2 \,ds,\end{aligned}\ ] ] which is uniformly bounded in @xmath161 . hence , sending @xmath320 , one sees that @xmath321 is a @xmath313-martingale .
note that on @xmath54 , @xmath322 and hence on that event , @xmath323 where in the second line we have used the definition of @xmath39 and the cauchy
schwarz inequality in the fourth . using the definition of @xmath324 yields
@xmath325 as @xmath326 and @xmath327 is continuous on @xmath328 ( which follows from theorem [ tito ] ) .
hence , this implies that for @xmath104 positive , on @xmath54 @xmath329 for all @xmath20 and hence by corollary iv.34.13 of @xcite , @xmath219 converges as @xmath330 .
therefore by ( [ eexteq ] ) and ( [ eextbd ] ) , @xmath331 converges a.s .
as well .
denote by @xmath332 the limit of @xmath333 .
it is immediately evident that @xmath334 is a probability measure on @xmath14 . to show that @xmath335 where @xmath63 is a random point in @xmath14 , we now defer to the proof of theorem 1 in tribe @xcite , as it is identical from this point forward .
similar ( but simpler ) reasoning holds to show @xmath336 a.s . on @xmath54 where @xmath62 is a random point in @xmath14 .
let @xmath337 .
note that @xmath338 is independent of @xmath22 and that @xmath339 a.s .
when @xmath340 because @xmath341 and both @xmath55 and @xmath56 converge a.s . by theorem [ textinc ] .
then for @xmath104 bounded and lipschitz , @xmath342 as @xmath340 .
therefore it is enough to note that since @xmath343 depends only on @xmath136 , the convergence of @xmath50 gives @xmath344 and hence @xmath345 by remark [ rcomrel](a ) , @xmath346 as @xmath340 . since there exists a countable separating set of bounded lipschitz functions @xmath347 , and the above holds for each @xmath348 ,
@xmath349 \(a ) theorem [ textconv ] holds in the critical branching case .
that is , if @xmath350 , @xmath351 where @xmath352 the convergence of the critical ordinary sou process to a random point follows directly from tribe s result . that this holds for the sou process with attraction to the com follows from the calculations above .
the distribution of the random point @xmath62 has been identified in tribe @xcite by approximating with branching particle systems .
in fact , the law of @xmath62 can be identified as @xmath353 , where @xmath354 is an ornstein uhlenbeck process with initial distribution given by @xmath355 and @xmath53 is the extinction time .
finding the distribution of @xmath63 remains an open problem however .
proof of theorem [ tsurvival ] let @xmath265 and as in the proof of theorem [ textinc ] , assume @xmath273 . also , without loss of generality , assume that @xmath356 for this proof . by theorem [ tito ] , @xmath55 is a continuous local martingale with decomposition given by @xmath357 where @xmath358_t = \int^t_0\frac{\overline{y^2_s}-\yb_s^2}{k_s(1)}\,ds = \int ^t_0\frac{v(y_s)}{k_s(1)}\,ds,\ ] ] with @xmath359 .
theorem iv.34.12 of @xcite shows that on the set @xmath360_\infty < \infty\}\cap s$ ] , @xmath361 a.s . converges
note that by lemma [ lextprob ] , for a.e .
@xmath362 , @xmath363 , recalling that @xmath364 .
hence , it follows that @xmath365_\infty<\infty$ ] on @xmath52 if @xmath366 then @xmath367 by cauchy schwarz and lemma [ lsecmom ] since @xmath265 .
therefore on @xmath52 , @xmath257 converges a.s . to some limit @xmath368 .
note that if @xmath369 , remark [ rcomrel](b ) gives @xmath370 and so ( b ) holds .
that @xmath56 converges on @xmath52 for @xmath11 follows from the fact that @xmath55 converges and equation ( [ ecorr ] ) by setting @xmath371 by remark [ rcomrel](b ) , we see that for @xmath11 since @xmath372 , @xmath373 and @xmath374 now argue by conditioning as in the end of theorem [ textinc ] to get the full result .
the next few results are necessary to establish the almost sure convergence of @xmath375 on the survival set .
this will in turn be used to show the almost sure convergence of @xmath376 using the correspondence of remark [ rcomrel](a ) .
let @xmath70 be the standard ornstein uhlenbeck semigroup ( with attraction to the origin ) .
note that @xmath377 in norm where @xmath378 which is independent of @xmath379 .
recall that @xmath380 and @xmath381 a.s . from lemma [ lextprob ] .
[ lsouconvl2 ] if @xmath11 , @xmath382 and @xmath383 , then on @xmath52 , for any @xmath384 , @xmath385 and @xmath386 where @xmath387 depends only on @xmath0 and @xmath245 , and @xmath303 is a positive constant dependent only on @xmath34 and @xmath42 .
[ rconv ] as the @xmath125 convergence in lemma [ lsouconvl2 ] is exponentially fast , it follows from the borel cantelli lemma and chebyshev inequality that for a strictly increasing sequence @xmath388 where @xmath389 , for @xmath384 , @xmath390 the idea is to use the above remark to bootstrap up to almost sure convergence in lemma [ lsouconvl2 ] with some estimates on the modulus of continuity of the process @xmath391 [ lsou - increments ] suppose @xmath11 , @xmath392 and @xmath393 . if @xmath384 and @xmath394 , then @xmath395 ^ 4\bigr ) \le c(t)h^2e^{-\zeta^*t},\ ] ] where @xmath396 is a positive constant depending only on @xmath34 and @xmath42 and @xmath387 is polynomial in @xmath20 , and depends on @xmath42 , @xmath34 and @xmath0 .
let @xmath397 , @xmath398 be positive , continuous functions .
further , suppose @xmath399 is symmetric about 0 and convex with @xmath400 and @xmath401 is increasing with @xmath402 .
the following is a very useful result of garsia , rodemich and rumsey @xcite .
[ pgrr ] if @xmath338 is a measurable function on @xmath403 $ ] such that @xmath404 ^ 2 } \psi\biggl(\frac{f(t)-f(s)}{p(|t - s|)}\biggr ) \,ds
\,dt = b < \infty\ ] ] then there is a set @xmath13 of measure 0 such that if @xmath405\setminus k$ ] then @xmath406 if @xmath338 is also continuous , then @xmath13 is in fact the empty set . with this result in hand , we can now bring everything together to prove convergence of @xmath375 .
proof of theorem [ tsouconv ] the strategy for this proof is simple : we use remark [ rconv ] to see that we can lay down an increasingly ( exponentially ) dense sequence @xmath407 which converges almost surely , and that we can use lemma [ lsou - increments ] to get a modulus of continuity on the process @xmath408 , which then implies that if the sequence is converging , then the entire process must be converging .
assume that @xmath409 and @xmath393 and argue as in theorem [ textinc ] in the general case .
let @xmath384 .
denote @xmath410 by @xmath411 for the remainder of the proof .
let @xmath412 , and let @xmath413 where @xmath414 .
let @xmath415 be the constant @xmath416 that appears in proposition [ pgrr ] , with aforementioned functions @xmath417 and @xmath418 , for the path @xmath419 $ ] . then note that @xmath420 ^ 2 } \psi\biggl(\frac { { { \accentset{\circ}{x}}}_{tt}-{{\accentset{\circ}{x } } } _ { ts}}{p(|t - s|)}\biggr ) \,ds \,dt\biggr ] \\ & = & \int\int_{[0 , 1]^2 } \frac { \p[\vert{{\accentset{\circ}{x}}}_{tt}-{{\accentset{\circ}{x } } } _ { ts}\vert^4]}{|t - s|^3\log^2({\lambda } /{|t - s| } ) } \,ds \,dt \\ & \le&\int\int_{[0 , 1]^2 } \frac { c(t(s\wedge t))e^{-\zeta ^*(s\wedge t)}t^2|t - s|^2}{|t - s|^3\log^2({\lambda}/{|t - s| } ) } \,ds \,dt \\ & = & 2t^2 \int_0 ^ 1\int_0^t \frac { c(ts)e^{-\zeta ^*(ts)}|t - s|^2}{|t - s|^3\log^2({\lambda}/{|t - s| } ) } \,ds \,dt \\ & \le&2c(t)t^2 \int_0 ^ 1\int_0^t \frac { 1}{|t - s|\log^2 ( { \lambda } /{|t - s| } ) } \,ds \,dt \\ & \le&\frac{c(t)t^2}{2e^4},\end{aligned}\ ] ] where @xmath387 is the polynomial term that appears in lemma [ lsou - increments ] . since @xmath421 is continuous , by garsia rodemech rumsey @xcite , for all @xmath422 , @xmath423 where @xmath81 is a constant independent of @xmath129 ( see corollary 1.2 of walsh @xcite for this calculation ) . rewriting the above , @xmath424 where @xmath425 . note that @xmath426 , which is polynomial in @xmath129 of fixed degree @xmath427 .
let @xmath428 be the set of probability 1 such that for all positive integers @xmath129 equation ( [ emc ] ) holds and @xmath429 for @xmath129 large enough .
to see , use borel cantelli : @xmath430 which is summable over all positive integers @xmath129 .
suppose @xmath431 .
let @xmath432 be such that @xmath433 .
then for all integral @xmath434 , and @xmath435 with @xmath436 , @xmath437 be a sequence of the form in remark [ rconv ] , with the additional condition that @xmath438 are evenly spaced within @xmath439 for each @xmath440 ( i.e. , @xmath441 for @xmath442 ) .
evidently @xmath443 converges a.s . to a limit @xmath444 . without loss of generality , assume convergence of the sequence on the set @xmath428 .
there exists @xmath445 such that for all @xmath446 , @xmath447 .
hence , for all @xmath20 such that @xmath448 there exists @xmath449 such that @xmath450 and hence @xmath451 sending @xmath330 gives almost sure convergence of @xmath452 to @xmath453 by theorem [ lsouconvl2 ] , since @xmath454 with @xmath20 . note that this implies for @xmath384 @xmath455 since on @xmath52 , @xmath456 a.s . by exercise 2.2 of @xcite on @xmath457 , the space of probability measures on @xmath14 ,
the prohorov metric of weak convergence is equivalent to the vasserstein metric .
it is easy to construct a class @xmath458 that is a countable algebra of lipschitz functions and that therefore is strongly separating ( see page 113 of @xcite ) . hence by theorem 3.4.5(b ) of @xcite , @xmath458 is convergence determining . since there exists a set @xmath459 with @xmath460 such that on @xmath461 , equation ( [ etestconv ] ) holds simultaneously for all @xmath462 , @xmath463 in the vasserstein metric , for @xmath464 because @xmath458 is convergence determining . to drop the dependence on the eighth moment
, we argue as in the proof of theorem [ textinc ] , where we make use of the markov property and the compact support property for historical brownian motion .
proof of theorem [ tisouconv ] this follows almost immediately from theorem [ tsouconv ] and the representation given in remark [ rcomrel](a ) .
let @xmath384 , then @xmath465 where @xmath466 .
remark [ rcomrel](b ) gives @xmath467 , and hence @xmath468 follows from theorem [ tsurvival ] .
note that @xmath469 since @xmath384 .
taking the supremum over @xmath104 and the previous theorem give @xmath470 much less can be said for the sou process repelling from its center of mass than in the attractive case .
we can , however , show that the center of mass converges , provided the rate of repulsion is not too strong , which we recall was the first step toward showing the a.s .
convergence of the normalized interacting sou process in the attractive case .
the situation here is more complicated since we prove that the com of the ordinary sou process with repulsion diverges almost surely , implying that results for convergence of @xmath84 will not simply be established through the correspondence .
we finish with some conjectures on the limiting measure for the repelling case . as in the previous section ,
assume @xmath471 , unless stated otherwise .
proof of theorem [ trcom ] assume that @xmath472 , like in the proof of theorem [ textinc ] . as in that theorem
, this condition can be weakened to just the finite initial mass condition using similar reasoning .
for part ( a ) , note that as in theorem [ tsurvival ] , @xmath55 will converge if @xmath473_t ) < \infty$ ] and which holds if the following quantity is bounded : @xmath474 by lemma [ lsecmom ] and by the conditions on @xmath42 . for ( b ) , we require the following lemma .
let @xmath475 and @xmath476 .
for a measure @xmath477 on @xmath14 , let @xmath478 be @xmath477 translated by @xmath479 .
that is , @xmath480 .
then : for all but at most countably many @xmath481 , @xmath482 for all but at most one value of @xmath481 @xmath483 we first note that , by the correspondence ( [ ecorr ] ) , we have that @xmath484 under our hypotheses @xmath55 converges by theorem [ trcom](a ) , and hence @xmath485 a.s . on @xmath52 .
therefore , on @xmath52 , @xmath486 note that one can build a solution of @xmath166 with initial conditions given by @xmath487 by seeing that if @xmath147 gives the solution of [ se2 ] , then @xmath488 gives the solution of @xmath489 , and that the projection @xmath490 gives the appropriate interacting sou process . by ( [ erep1 ] ) and ( [ erep2 ] ) , @xmath491
the random variable @xmath492 is finite a.s . and
so only a countable number of values @xmath481 exist with the latter expression positive , implying the first result .
the second result also follows as well since the last expression in the above display can be 1 for at most 1 value of @xmath481 . to complete the proof of theorem [ trcom](b ) ,
choose a value @xmath479 such that ( [ enzlim ] ) holds .
by theorem iii.2.2 of @xcite and the fact that @xmath493 for all @xmath494 , for the ou semigroup @xmath70 , we have that for all @xmath494 @xmath495 by our choice of @xmath481 , @xmath496 holds , and hence by ( [ eabscon ] ) we have @xmath497 recalling from ( [ erep2 ] ) that @xmath498 exists a.s . , we are done .
note that for @xmath499 , this implies that even if mass is repelled at rate @xmath42 , the com of the interacting sou process still settles down in the long run .
that is , driving @xmath147 away from @xmath55 seems to have the effect of stabilizing it .
one can think of this as a situation where the mass is growing quickly enough that the law of large numbers overcomes the repelling force .
more surprising is that the com of the ordinary sou process diverges exponentially fast , even while the com of the interacting one settles down .
this follows from the correspondence @xmath500 and the cancellation that occurs in it due to the exponential rate of @xmath56 .
the next lemma shows that theorem 1 of englnder and winter @xcite can be reformulated to yield a result for the sou process with repulsion at rate @xmath42 ( where @xmath42 is taken to be a negative parameter in our setting ) .
[ lreform ] on @xmath52 , for the sou process , @xmath501 with repulsion rate @xmath502 and compactly supported initial measure @xmath503 , and any @xmath504 @xmath505 where @xmath506 is a positive random variable on the set @xmath52 . note that by example 2 of pinsky @xcite it is shown that the hypotheses of theorem 1 of @xcite hold for the sou process with repulsion from the origin at rate @xmath507 .
the theorem says that there is a function @xmath508 such that @xmath509 where @xmath239 is as in lemma [ lextprob ] .
example 2 also shows that for @xmath510 , @xmath511 where @xmath477 is lebesgue measure on @xmath14 .
hence , manipulating the expression in ( [ eew ] ) by using the previous equation and lemma [ lextprob ] gives @xmath512 this lemma indicates that on the survival set , when @xmath12 , one can not naively normalize @xmath30 by its mass since the probability measures @xmath513 are not tight .
that is , a proportion of mass is escaping to infinity and is not seen by compact sets .
note that the lemma above implies that for @xmath30 , the right normalizing factor is @xmath514 .
we say a measure - valued process @xmath515 goes locally extinct if for any finite initial measure @xmath516 and any bounded @xmath517 , there is a @xmath518-a.s .
finite stopping time @xmath519 so that @xmath520 for all @xmath521 a.s .
[ rlocextinc ] example 2 of pinsky @xcite also shows that for @xmath522 the sou undergoes local extinction ( all the mass escapes ) .
hence for @xmath523 , there is no normalization where @xmath524 can be expected to converge to something nontrivial . from remark [ rito2 ] , one can show that @xmath525 where @xmath161 is a martingale .
therefore , you can think of the com of @xmath26 , the sou process repelling from origin , as being given by an exponential drift term plus fluctuations .
the correspondence of remark [ rcomrel](b ) implies then that @xmath526 or in other words , the center of mass of the sou process repelling from its com is given by simply the fluctuations .
we finish with some conjectures .
[ crepulse ] on the survival set , if @xmath527 is fixed and compactly supported , then the following is conjectured to hold : if @xmath528 , then there exists constant @xmath529 so that for @xmath530 , @xmath531 where @xmath532 is a random measure depending on @xmath76 .
if @xmath533 , then @xmath82 undergoes local extinction . we expect that @xmath534 simply because of the repulsion from the com built in to the model results in a proportion of mass being lost to infinity . one would expect that the limiting measure @xmath532 is a random multiple of lebesgue measure as in the ordinary sou process case , due to the correspondence , but it is conceivable that it is some other measure which has , for example , a dearth of mass near the limiting com . as stated earlier , it is difficult to use lemma [ lreform ] to prove this conjecture as the correspondence becomes much less useful in the repulsive case .
the problem is that while the equation @xmath535 still holds for @xmath20 finite , the time integral of @xmath536 now diverges .
we now prove the lemmas first stated in section [ sexpre ] .
proof of lemma [ lextprob ] we first note that @xmath167 is a @xmath537-historical process .
the martingale problem then shows that @xmath538 is a nonnegative @xmath118-martingale and therefore converges almost surely by the martingale convergence theorem to a random variable @xmath239 .
it follows that @xmath539 , since 0 is an absorbing state for @xmath231 .
exercise ii.5.3 in @xcite shows that @xmath540 the same exercise also shows @xmath541 now sending @xmath542 gives @xmath543 and sending @xmath544 gives @xmath545 since @xmath546 .
therefore , @xmath241 almost surely .
proof of lemma [ lcsp ] we follow the proof of theorem iii.1.3(a ) in perkins @xcite .
first , note that if @xmath547 is another supercritical historical brownian motion starting at time @xmath548 with initial measure @xmath477 under @xmath549 and @xmath81 a borel subset of @xmath387 , then the process defined by @xmath550 is also a supercritical historical brownian motion starting at time @xmath548 with initial measure @xmath551 given by @xmath552 under @xmath549 .
then using the extinction probabilities for @xmath553 ( refer , e.g. , to exercise ii.5.3 of @xcite ) we have @xmath554 using the markov property for @xmath13 at time @xmath555 and ( [ eext ] ) gives @xmath556 \nonumber\\[-2pt ] & & \qquad=\p\bigl[1- \exp\bigl\{-\bigl ( 2\beta k_{{j}/{2^n } } \bigl(\bigl\{y\dvtx \bigl\vert y ( { j}/{2^n } ) - y\bigl({(j-1)}/{2^n}\bigr)\bigr\vert\nonumber\\[-2pt ] & & \hspace*{169pt}\qquad{}>c ( { j}/{2^n})h(2^{-n } ) \bigr\ } \bigr)\bigr)\nonumber\\[-2pt ] & & \hspace*{197pt}\qquad{}\big/ ( 1 - e^{-\beta/{2^n } } ) \bigr\}\bigr ] \nonumber \\[-8pt ] \\[-8pt ] \nonumber & & \qquad\le\p\biggl[\frac { 2\beta k_{{j}/{2^n } } ( \{y\dvtx \vert y ( { j}/{2^n } ) - y({(j-1)}/{2^n})\vert > c({j}/{2^n})h(2^{-n } ) \ } ) } { 1 - e^{-\beta/{2^n } } } \biggr ] \\[-2pt ] & & \qquad\le\p\biggl[\frac { 2\beta k_{{j}/{2^n } } ( \{y\dvtx \vert y ( { j}/{2^n } ) - y({(j-1)}/{2^n})\vert > c({j}/{2^n})h(2^{-n } ) \ } ) } { ( { \beta}/{2^n } ) - ( { \beta^2}/{2^{2n+1 } } ) } \biggr]\nonumber\\[-2pt ] & & \qquad\le\frac{2^{n+1}}{1 - ( { \beta}/{2^{n+1 } } ) } \nonumber\\[-2pt ] & & \qquad\quad{}\times\p(k_{{j}/{2^n}}(1))\hat{\p}_{{j}/{2^n}}\biggl(\biggl\vert y\biggl(\frac{j}{2^n}\biggr ) - y\biggl(\frac{j-1}{2^n}\biggr)\biggr\vert > c\biggl(\frac{j}{2^n}\biggr)h(2^{-n } ) \biggr),\nonumber\end{aligned}\ ] ] where in the last step we have simply rearranged the constants and multiplied and divided by the mean mass at time @xmath557 and used the definition of the campbell measure ( definition [ dcamp ] ) .
since under the normalized mean measure , @xmath22 is a stopped brownian motion by remark [ rsouhrep](b ) , we use tail estimates to see that the last quantity in ( [ ebc ] ) is @xmath558 & \le&2^{n+2}2^{{\beta j\ln2}/{2^n}}\p(x_0(1))c_d n^{d/2 - 1}2^{-n(({j}/{2^n})+c_0)/2}.\end{aligned}\ ] ] hence , summing over @xmath559 from @xmath89 to @xmath560 gives @xmath561 & & \quad k_t\biggl(\biggl\{y\dvtx \biggl\vert y\biggl(\frac{j}{2^n}\biggr ) - y\biggl(\frac { j-1}{2^n}\biggr)\biggr\vert > c\biggl(\frac { j}{2^n}\biggr)h(2^{-n } ) \biggr\}\biggr)>0 \biggr ] \\[-2pt ] & & \qquad\le\p(x_0(1))c_d n^{d/2 - 1}2^{n+2- nc_0/2}\sum_{j=1}^{n2^n } 2^{{\beta j\ln2}/{2^n } - n({j}/{2^n})}\\[-2pt ] & & \qquad\le\p(x_0(1))c_d n^{d/2 - 1}2^{-2n+2}\sum_{j=1}^{n2^n } 2^{- { j}/{2^n } } \\[-2pt ] & & \qquad\le\p(x_0(1))c_d n^{d/2 - 1}2^{-n+2},\end{aligned}\ ] ] where we have used the fact that @xmath246 .
hence , the sum over @xmath7 of the above shows by borel
cantelli that there exists for almost sure @xmath136 , @xmath562 such that for all @xmath563 , for all @xmath564 , for all @xmath565 , for @xmath18-a.a .
@xmath22 , @xmath566 letting @xmath567 , note that on the dyadics , by above , we have that @xmath568\\ & \le&\sum_{n= \lfloor-{\ln\lambda}/{\ln2 } \rfloor } c'(d , c_0)n^{d/2 - 1}2^{2n - nc_0/2 } \\ & \le & c(d , c_0 , \varepsilon)\lambda^{({c_0}/{2})-\varepsilon},\end{aligned}\ ] ] where @xmath569 can be chosen to be arbitrarily small ( though the constant @xmath387 will increase as it decreases ) .
the rest of the proof follows as in theorem iii.1.3(a ) of @xcite , via an argument similar to levy s proof for the modulus of continuity for brownian motion .
proof of lemma [ lsecmom ] assume that @xmath201 and @xmath154 .
recall @xmath570 note that below `` @xmath571 '' denotes less than up to multiplicative constants independent of @xmath20 and @xmath22 .
suppose that @xmath572 , where @xmath573 is the same as in the previous lemma .
then , as @xmath574 @xmath575 by cauchy
schwarz and jensen s inequality .
therefore , integrating with respect to the normalized measure gives @xmath576 and therefore we need only find the appropriate bounds for expectation of @xmath577 to get the result .
after another few applications of cauchy schwarz and integrating by parts , @xmath578 as @xmath579 , @xmath580 e^{-2\gamma(t - s)}\,ds \\ & \lesssim & e^{-2\gamma t}(|z_0|^2+|y_0|^2 ) + |y_0|^2 + |y_0|^2\gamma t(1-e^{-2\gamma t})/2 \\ & & { } + c(t)^2 \biggl(\frac{h(\delta)}{\delta}\biggr)^2\bigl(t^2 + \gamma t^3(1-e^{-2\gamma t})\bigr ) \\ & \lesssim&(1+\vert\gamma\vert t)(1+e^{-2\gamma t})(|z_0|^2+|y_0|^2 ) \\ & & { } + c(t)^2 \biggl(\frac{h(\delta)}{\delta}\biggr)^2\bigl(t^2 + \gamma t^3(1-e^{-2\gamma t})\bigr).\end{aligned}\ ] ] integrating by the normalized measure @xmath193 , @xmath581 then using ( [ ecombd ] ) and using the above bound on @xmath582 gives @xmath583 note that @xmath584 and @xmath585 are @xmath586 measurable . by applying it^ o s formula to @xmath587 and using the decomposition @xmath588 ( which follows from proposition 2.7 of @xcite ) , we get @xmath589 where @xmath590 is a local martingale until time @xmath53 , for each @xmath7 .
in fact , the sequence of stopping times @xmath276 appearing in theorem [ tito ] can be used to localize each @xmath590 . applying first the monotone convergence theorem and
then localizing gives @xmath591 where we have used the positivity of @xmath348 to get the fourth line and the monotone convergence theorem in the last line .
further , note that @xmath592 by the calculation immediately above .
thus , taking expectations in ( [ el2bd ] ) and plugging in @xmath593 gives @xmath594 now let @xmath250 be chosen so that @xmath595 and @xmath596 for @xmath256 $ ] . note that @xmath597 by choosing the constant @xmath250 so that @xmath598 is large ( @xmath599 is enough ) .
remark [ rmoments ] follows from the above proof , after noting that the exponent @xmath598 in lemma [ lcsp ] can be made arbitrarily large by choosing a sufficiently large constant @xmath250 .
hence by choosing @xmath598 appropriately , we can show that @xmath600 < \infty,\ ] ] which can then be used to adapt the proof above .
recall that @xmath601 .
we will , with a slight abuse of notation , allow @xmath158 to denote the orthogonal martingale measure generated by the martingale problem for @xmath26 .
let @xmath81 be the infinitesimal generator for an ou process , and hence recall that for @xmath602 , @xmath603
the next two proofs are for lemmas stated in section [ sssurv ] .
proof of lemma [ lsouconvl2 ] let @xmath384 . by the extension of the martingale problem for @xmath26 given in proposition ii.5.7 of @xcite , for functions @xmath604\times\r ^d\rightarrow\r$ ] such that @xmath605 satisfies the definition before that proposition , @xmath606 where @xmath158 is the orthogonal martingale measure derived from the martingale problem for the sou process .
it is not difficult to show that @xmath607 where @xmath104 as above satisfies requirements for proposition ii.5.7 of @xcite .
plugging this in gives @xmath608 since @xmath609 .
multiplying by @xmath610 and integrating by parts gives @xmath611 note that as the @xmath612-process has a stationary distribution @xmath613 where @xmath377 in norm .
when @xmath99 is large in ( [ exblim ] ) , @xmath614 does not contribute much to the stochastic integral and hence we expect the limit of @xmath615 to be @xmath616 which is a well defined , finite random variable as @xmath617_\infty < \| \phi\|^2 \int_0^\infty e^{-2\beta s}x_s(1)\,ds,\ ] ] which is finite in expectation
. as @xmath618 does not depend on @xmath379 , it follows that @xmath619 given this decomposition for @xmath620 , we write @xmath621 if @xmath148 is a @xmath0-dimensional ou process satisfying @xmath622 where @xmath15 is a @xmath0-dimensional brownian motion , then @xmath623 and hence @xmath148 is gaussian , with mean @xmath624 and covariance matrix @xmath625 . evidently , @xmath626 is also gaussian , mean 0 and variance @xmath627 .
we use a simple coupling : suppose that @xmath628 is a random variable independent of @xmath148 such that @xmath629 ( i.e. , @xmath628 is gaussian with mean @xmath630 and covariance @xmath631 ) . then using the fact that @xmath384 and the cauchy schwarz inequality , followed by our coupling with @xmath632 gives @xmath633 taking expectations and using cauchy
schwarz and the assumptions on @xmath245 gives exponential rate of convergence for the above term .
since we can think of @xmath634 as a martingale in @xmath635 up until time @xmath20 , various martingale inequalities can be applied to get bounds for the terminal element , @xmath636 .
note that this process is not in general a martingale in @xmath20 .
we have @xmath637 \nonumber\\ & & \qquad= \p\biggl[\biggl(\int_0^t\int e^{-\beta s}\bigl(p_\infty \phi ( x ) - p_{t - s}\phi(x)\bigr ) \,dm(s , x ) \biggr)^2 \biggr ] \\ & & \qquad\le\p\biggl[\int_0^t e^{-2\beta s}\int\bigl(p_\infty \phi ( x ) - p_{t - s}\phi(x)\bigr)^2x_s(dx)\,ds\biggr ] . \nonumber\end{aligned}\ ] ] then as @xmath104 lipschitz , by the coupling above , @xmath638 \\ & = & \int_0^t e^{-2\beta s-2\gamma(t - s)}\p\biggl[\int|x|^2 + \frac { d}{2\gamma } x_s(dx)\biggr]\,ds \\ & = & \int_0^t e^{-2\beta s-2\gamma(t - s)}\p\biggl[k_s(|z_s|^2 ) + \frac { d}{2\gamma}x_s(1)\biggr]\,ds.\end{aligned}\ ] ] applying the cauchy
schwarz inequality followed by remark [ rmoments](b ) gives @xmath639 where the last line follows by first noting that @xmath640 is a martingale .
that is , @xmath641 & = & 2\p\biggl(x_0(1)^2+\biggl[\int_0^\cdot e^{-\beta r}\,dm(r , x)\biggr]_s\biggr)\nonumber \\[-2pt ] & = & 2\p\biggl(x_0(1)^2 + \int_0^s e^{-2\beta r}x_r(1)\,dr \biggr ) \nonumber \\[-8pt ] \\[-8pt ] \nonumber & = & 2\p(x_0(1)^2)+ 2\int_0^s e^{-\beta r } \p(e^{-\beta r}x_r(1 ) ) \,dr\\[-2pt ] & = & 2\p(x_0(1)^2)+ 2\int_0^s e^{-\beta r } \p(x_0(1 ) ) \,dr\nonumber \nonumber\\[-2pt ] & \le&2\p\biggl(x_0(1)^2 + \frac{1}{\beta}x_0(1)\biggr).\nonumber\end{aligned}\ ] ] therefore , @xmath642 \,ds\\[-2pt ] & & { } + \int_0^te^{-2\beta s-2\gamma(t - s ) } e^{\beta s}\frac { d}{2\gamma}\p(x_0(1))\,ds \\[-2pt ] & \le&\int_0^t e^{-\beta s-2\gamma(t - s)}\biggl[b(s , \gamma , 4)^ { { 1}/{2 } } \p\biggl(x_0(1)^2 + \frac{1}{\beta}x_0(1)\biggr)^{{1}/{2}}\\[-2pt ] & & \hspace*{155pt}\qquad { } + \frac { d}{2\gamma}\p(x_0(1 ) ) \biggr]\,ds \\[-2pt ] & < & c e^{-\zeta_1 t},\end{aligned}\ ] ] where @xmath643 where @xmath569 is arbitrary small and comes from the polynomial term in the integral .
finally , @xmath644 & \le&(p_\infty\phi)^2\int_t^\infty e^{-\beta s}\p(e^{-\beta s}x_s(1))\,ds\\ & \le&\frac{(p_\infty\phi)^2}{\beta}e^{-\beta t}\p(x_0(1)),\end{aligned}\ ] ] since @xmath645 is a martingale . therefore , since @xmath646 , we see that @xmath647 gives the correct exponent .
proof of lemma [ lsou - increments ] the proof will follow in a manner very similar to the proof of the previous lemma . from the calculations above , we see that @xmath648 using the cauchy
schwarz inequality , we can find bounds for @xmath649 , separately : @xmath650 ^ 2.\end{aligned}\ ] ] recalling the simple coupling in the previous lemma to see that @xmath651 where @xmath652 is as above , an ou process started at @xmath379 , and @xmath653 is independent of @xmath654 but such that @xmath655 .
hence , @xmath653 is gaussian with mean @xmath656 and covariance matrix @xmath657
. therefore , @xmath658 hence , @xmath659\\ & \le & c_1(d , \gamma ) h^2 e^{-4\gamma t},\end{aligned}\ ] ] where @xmath660 is a constant that is finite by assumptions on the initial measure . to get bounds on the expectation of @xmath661 , we use martingale inequalities .
note that @xmath662 is a martingale until time @xmath20 .
therefore , using the burkholder davis gundy inequality and the coupling above gives @xmath663_t^2 ) \\ &
= & c\p\biggl[\biggl(\int_0^t\int e^{-2\beta s } \bigl((p_{t+h - s}-p_{t - s})\phi ( x ) \bigr)^2x_s(dx ) \,ds \biggr)^2\biggr]\\ & \le & c\p\biggl[\biggl ( \int_0^t e^{-2\beta s } \bigl(e^{-\gamma ( t+h - s)}-e^{-\gamma(t - s)}\bigr)^2\int|x|^2x_s(dx)\,ds\biggr)^2 \\ * & & { } + \biggl(\int_0^t\frac{de^{-2\beta s } } { 2\gamma } \bigl(e^{-2(t - s)\gamma}-e^{-2(t+h - s)\gamma}\bigr)x_s(1 ) \,ds \biggr)^2\biggr]\\ & \le & c\p\biggl[\biggl ( t\int_0^t e^{-4\beta s } \bigl(e^{-\gamma ( t+h - s)}-e^{-\gamma(t - s)}\bigr)^4\biggl(\int|x|^2x_s(dx)\biggr)^2\,ds\biggr ) \\ * & & { } + t\biggl(\frac{d}{2\gamma}\biggr)^2\int_0^t e^{-4\beta s } \bigl(e^{-2(t - s)\gamma}-e^{-2(t+h - s)\gamma}\bigr)^2x_s(1)^2 \,ds \biggr]\\ & = & c t\int_0^t e^{-2\beta s } \bigl(e^{-\gamma(t+h - s)}-e^{-\gamma ( t - s)}\bigr)^4\p\biggl[\biggl(\int|x|^2e^{-\beta s}x_s(dx)\biggr)^2\biggr]\,ds \\ & & { } + t\biggl(\frac{d}{2\gamma}\biggr)^2\int _ 0^te^{-2\beta s}\bigl(e^{-2(t - s)\gamma}-e^{-2(t+h - s)\gamma}\bigr)^2\p[(e^{-\beta s}x_s(1))^2 ]
\,ds.\end{aligned}\ ] ] since @xmath664 , by remark [ rmoments](b ) , @xmath665\\ & \le&\p ( \overline{z_{s}^2}^4 ; s<\tx)^{1/2}\p ( e^{-4\beta s}x_{s}(1)^4)^{1/2 } \\ & \le&\p ( \overline{z_{s}^8 } ; s<\tx)^{1/2 } \p ( e^{-4\beta s}x_{s}(1)^4)^{1/2}\\ & \le & cb(\gamma , s , 8)^{1/2}\bigl ( \p\bigl(x_0(1)^4 + sx_0(1)^2 + sx_0(1)\bigr ) \bigr)^{1/2}.\end{aligned}\ ] ] the bound on the expectation of @xmath666 follows by an application of the bdg inequality to @xmath667 and similar calculations used to determine the bound on ( [ emassl2 ] ) .
therefore , @xmath668 where @xmath669 is polynomial in @xmath20 . by another application of the bdg inequality , and
noting that @xmath670 , @xmath671\\ & \le & c\p\biggl[\biggl(\int_t^{t+h}\int\bigl(e^{-\beta s}p_{t+h - s}\phi ( x)\bigr)^2x_s(dx ) \,ds\biggr)^2 \biggr ] \\ & \le & c h \p\biggl[\int_t^{t+h}e^{-4\beta s}x_s(1)^2 \,ds \biggr]\\ & = & c he^{-2\beta t } \int_t^{t+h } \p[e^{-2\beta s}x_s(1)^2 ] \,ds \\ & \le & c he^{-2\beta t } \int_t^{t+h } \p[x_0(1)^2+x_0(1)/\beta]\,ds \\ & \le & c_3(\beta ) h^2e^{-2\beta t},\end{aligned}\ ] ] where the second last line follows from the same calculations performed in estimating moments of @xmath661 .
note that the constant @xmath672 does not depend on @xmath20 here .
putting the pieces together shows that there exists a function @xmath387 polynomial in @xmath20 and a positive constant @xmath396 such that ( [ eincrbd ] ) holds .
i am greatly indebted to my advisor , ed perkins , for his numerous ( very helpful ) suggestions and generous support during the course of this work .
i am particularly grateful for his help with part ( b ) of theorem [ trcom ] . | we construct a supercritical interacting measure - valued diffusion with representative particles that are attracted to , or repelled from , the center of mass . using the historical stochastic calculus of perkins ,
we modify a super ornstein uhlenbeck process with attraction to its origin , and prove continuum analogues of results of englnder [ _ electron .
j. probab . _
* 15 * ( 2010 ) 19381970 ] for binary branching brownian motion .
it is shown , on the survival set , that in the attractive case the mass normalized interacting measure - valued process converges almost surely to the stationary distribution of the ornstein uhlenbeck process , centered at the limiting value of its center of mass . in the same setting
, it is proven that the normalized super ornstein uhlenbeck process converges a.s . to a gaussian random variable , which strengthens a theorem of englnder and winter [ _ ann .
inst .
henri poincar probab .
stat . _
* 42 * ( 2006 ) 171185 ] in this particular case . in the repelling setting
, we show that the center of mass converges a.s .
, provided the repulsion is not too strong and then give a conjecture .
this contrasts with the center of mass of an ordinary super ornstein uhlenbeck process with repulsion , which is shown to diverge a.s .
a version of a result of tribe [ _ ann . probab . _ * 20 * ( 1992 ) 286311 ] is proven on the extinction set ; that is , as it approaches the extinction time , the normalized process in both the attractive and repelling cases converges to a random point a.s . . . |
universal features of the electronic wave function @xmath0 are of continued interest to physicists and chemists , as they guide the construction of highly accurate wave functions , @xcite explicitly correlated anstze within f12 theory @xcite and accurate jastrow factors for quantum monte carlo ( qmc ) calculations .
@xcite the coulombic singularity at short interelectronic distances @xmath1 dominates all other terms and , near the two - particle coalescence ( 2pc ) point @xmath2 , the behavior of @xmath0 becomes independent of other details of the system .
early work by kato , @xcite and elaborations by pack and byers - brown , @xcite showed that , as two opposite - spin electrons approach , @xmath0 has the form @xmath3 similar 2pc conditions are known for triplet and unnatural parity singlet states .
@xcite to remove divergences in the local energy @xmath4 at the 2pc points , cusp conditions such as must be satisfied .
these divergences are especially harmful in diffusion qmc calculations , where they can lead to population - control problems and significant biases .
@xcite recently , several authors have further probed the coalescence behavior of the wave function , extending the analysis of pack and byers brown to higher - order 2pc conditions .
@xcite surprisingly , imposing these additional conditions in qmc calculations was not found to be beneficial .
@xcite a wave function with correct 2pc behavior nonetheless yields a discontinuous local energy at three - particle coalescences ( 3pc ) , i.e. where @xmath5 .
unfortunately , a theoretical understanding of 3pc behavior has been much more elusive .
@xcite in 1954 , fock @xcite rediscovered a result derived earlier by gronwall @xcite and bartlett .
@xcite using hyperspherical coordinates , he showed that , when both electrons approach the nucleus in a helium atom , the wave function takes the form @xmath6 such logarithmic terms are characteristic of 3pc behavior @xcite and using a wave function based on yields a continuous local energy for the helium atom .
@xcite the present work aims to elucidate the form of the wave function at three - electron coalescences for three electrons , in various spin states , confined to a harmonic potential in @xmath7 or @xmath8 dimensions . eschewing the usual hyperspherical approach
, we develop a new technique to show that , in each case , the wave function has non - analytical behavior characterized by a logarithmic term .
atomic units are used throughout .
a system of three electrons , interacting coulombically but confined within a harmonic well , is variously called hookium , harmonium or hooke s law atom .
it has been studied at high and low densities by taut _
_ @xcite and cioslowski _ et al .
_ @xcite the previous work of white and stillinger , @xcite and other work showing that the kato cusp conditions are independent of the external potential , @xcite imply that our 3pc results will also hold for any external potential .
the hamiltonian of the system is @xmath9 where @xmath10 is the @xmath11-dimensional laplacian for electron @xmath12 and @xmath13 is the interelectronic distance between electrons @xmath12 and @xmath14 .
the strength of the interelectronic repulsion is measured by @xmath15 . without loss of generality ,
we choose the harmonic force constant @xmath16 .
we consider only @xmath17 states ( i.e. @xmath18 ) and , after adopting jacobi coordinates , the center - of - mass coordinate @xmath19 separates and one can write @xmath20 where @xmath21 are the remaining jacobi coordinates and @xmath22 are the wave function and energy of a harmonic oscillator .
if we express the rest of the hamiltonian in terms of the interelectronic distances instead of @xmath23 , @xmath24 and @xmath25 , then separation of variables reveals that the wave function @xmath26 in is a solution of the schrdinger - like equation @xmath27 where @xmath28 , the kinetic operator is @xmath29 and the internal and external potentials are @xmath30 we also define the symmetric polynomials [ eq : sympol ] @xmath31 as well as the usual `` hyperradius '' @xmath32 and @xmath33 which is proportional to the area of the triangle defined by the three interelectronic distances .
these quantities will be helpful for the remainder of this communication .
we are interested in the behavior of @xmath26 when the @xmath34 are all small and , because the laplacian is @xmath35 , we can treat @xmath15 as a perturbation parameter . expanding @xmath26 in ascending powers of @xmath34 yields @xmath36 where the zeroth- , first- and second - order wave functions satisfy @xmath37 the external potential @xmath38 does not contribute up to second order in @xmath34 ( or fourth - order perturbation theory ) .
following pauncz @xcite and matsen , @xcite the wave function of an @xmath39 state is given by @xmath40 , \end{split}\ ] ] where the vertical bar separates the spin - up and spin - down electrons , and the spatial wavefunction satisfies @xmath41 equation ensures that the pauli principle is satisfied and ( which is found in appendix c of ref . )
ensures that there is no quartet contamination .
the zeroth - order wave function is the lowest solution of eq . which satisfies and and , using the frobenius method , @xcite one finds that @xmath42 in the same way
, one finds the first - order wave function @xmath43 solving eq .
is difficult but it can be shown that @xmath44 where @xmath45 these zeroth- , first- and second - order wave functions agree with white and stillinger s hyperspherical 3d results .
@xcite the general second - order coefficient @xmath46 can be found by integrating over the hypersphere @xmath47 of unit radius .
@xcite in the 2d case , we find @xmath48 \notag \\ & \approx 3.344854,\end{aligned}\ ] ] where @xmath49 and @xmath50 are the complete elliptic integrals of the first and second kind .
@xcite in the 3d case , we find @xmath51 which disagrees with white and stillinger s value .
the 3pc behavior of quartet state wave functions has not been studied before .
the wave function of an @xmath53 , @xmath52 state is @xmath54 \ ^4\psi(\bm{r}_1,\bm{r}_2,\bm{r}_3)\end{aligned}\ ] ] where the spatial wave function is antisymmetric , i.e. @xmath55 using the same approach as above , we seek antisymmetric zeroth- and first - order wave functions and find @xmath56 and @xmath57 where @xmath58 and @xmath59 , @xmath60 and @xmath61 are given by .
the second - order wave function is @xmath62 where @xmath63 and @xmath64 the differences between and are interesting . in particular
, the second - order wave function of the quartet , unlike that of the doublet , is dimension - dependent , and the logarithmic singularity for the quartet appears at order @xmath65 , rather than @xmath66 .
this agrees with the prediction by white and stillinger @xcite that the non - analytic terms for the quartet state are of higher order . in the 2d case
, we use to find @xmath67 \\ & \approx 0.131306 .
\end{split}\end{aligned}\ ] ] in the 3d case , we find @xmath68 the wave function of an @xmath53 , @xmath69 state is given by @xcite @xmath70 where @xmath71 and @xmath72 is a symmetric solution of in dimension @xmath73 .
this `` interdimensional degeneracy '' has been used many times in the past , especially to calculate the energy of excited states in atomic systems . @xcite
see refs . for more details
. using the frobenius method as before , one finds that the zeroth- and first - order wave functions are @xmath74 the second - order wave function has the form @xmath75 where @xmath76 and @xmath77 though similar , the logarithmic singularities in the quartet states depend on @xmath78 via the constants in and . in the 2d case ,
we use to find @xmath79 \approx 1.834021 . } } \ ] ] in the 3d case , we find @xmath80
in this communication , we have shown that the exact wave function at the three - electron coalescence point for various spin states of three electrons in a two- or three - dimensional harmonic well diverges logarithmically .
our results should be valuable for explicitly correlated calculations and for qmc methods where the local energy discontinuity at the three - electron coalescence point could be removed by including logarithmic terms .
thanks the australian research council for a discovery early career researcher award ( grant no .
de130101441 ) and a discovery project grant ( dp140104071 ) .
p.m.w.g .
thanks the australian research council for funding ( grants no . dp120104740 and dp140104071 ) .
p.f.l . and p.m.w.g .
also thank the nci national facility for generous grants of supercomputer time .
49ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _
( , , ) @noop * * , ( ) , , , and , eds . ,
@noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) | the form of the wave function at three - electron coalescence points is examined for several spin states using an alternative method to the usual fock expansion .
we find that , in two- and three - dimensional systems , the non - analytical nature of the wave function is characterized by the appearance of logarithmic terms , reminiscent of those that appear as both electrons approach the nucleus of the helium atom .
the explicit form of these singularities is given in terms of the interelectronic distances for a doublet and two quartet states of three electrons in a harmonic well . |
the atlas detector @xcite is a general purpose detector built at one of the interaction points of the large hadron collider ( lhc ) where proton on proton collisions with a center of mass energy of 14 tev are expected .
the inner tracking system of atlas is made up of silicon detectors and straw drift tubes . while these detectors were placed with very high precision of the order of 100 @xmath0 , the precision required for physics necessitates determining the positions of the tracking elements to a few microns .
this is only achievable by doing a track based alignment .
the lhc has not yet started proton - proton collisions , however , the atlas detector is fully operational and has been collecting cosmic ray data .
this paper describes the alignment achieved using this cosmic ray data and explores the expected impact of the alignment on the physics performance in the early data .
the atlas detector consists of several systems . an inner tracker ( inner detector ) , electromagnetic and hadronic calorimeters and a muon spectrometer .
the inner detector is located within a solenoidal magnetic field of about 2 tesla and is made up of three subsystems as shown in figure [ fig : innerdetector ] .
the innermost subsystem is the pixel detector .
it is made up of three barrel layers and three disks in each endcap giving at least 3 space points per track .
there are a total of 1744 modules ( 1456 in the barrel and 144 in each endcap ) .
the pixel cell size is 50 @xmath0 @xmath1 400 @xmath0 with a corresponding resolution of 10 @xmath0 @xmath1 115 @xmath0 .
the more precise measurement is in the @xmath2 direction ( in the bending direction of the magnetic field ) and the less precise direction measures @xmath3 in the barrel and @xmath4 in the endcap . in the local frame of the module
the directions are referred to as local @xmath5 and local @xmath6 respectively .
the next subsystem is the sct ( semi - conductor tracker ) which consists of silicon microstrip detectors . in the sct
there are four barrel layers and 9 disks in each endcap giving 4 space points per track .
it has a total of 4088 modules ( 2112 in the barrel and 988 in each endcap ) .
the strip pitch is about 80 @xmath0 giving a resolution of 17 @xmath0 in the @xmath2 ( local @xmath5 ) measurement direction .
the modules are made of two back to back sides which are rotated 40 mrad with respect to each other to give a stereo measurement .
this results in a space point resolution of about 580 @xmath0 in @xmath3 ( barrel ) or @xmath4 ( endcap ) .
the outer subsystem is the trt ( transition radiation tracker ) .
it is made up of straw drift tubes which have a diameter of 4 mm .
the straws are embedded in a material that produces transition radiation photons which facilitates electron identification . on average 36 straws
are crossed per track .
the barrel is segmented into 96 modules arranged in three rings .
each endcap is made up of 20 wheels , where each wheel consists of two four - plane structures .
the resolution is 130 @xmath0 in the @xmath2 measurement direction .
the data used to obtain the alignment presented here was collected from september to december 2008 using cosmic rays . in total 2.6 million inner detector tracks
were recorded with the magnetic field on and 5 million with the magnetic field off . a number of these tracks , however
, do not pass through all three subsystems .
requiring at least one hit in the sct , results in 880k and 2 million tracks with field on and off respectively . requiring at least one hit in the pixel detector , results in 190k tracks with field on and 230k tracks with field off .
the hit residual is the distance between the track prediction and the measured hit .
the basic approach to align the detector is to reduce the residuals .
the main methods build a @xmath7 that is to be minimized .
this @xmath7 is defined as @xmath8 where @xmath4 is a vector of the residuals and @xmath9 is the corresponding covariance matrix . the minimum is obtained by solving the condition : @xmath10 where @xmath11 is a vector containing the alignment constants .
these are generally the 3 translations and 3 rotations of each alignable structure ( e.g. a module or layer ) .
the solution to the minimization is of the form : @xmath12 where a full derivation can be found in @xcite .
the matrix @xmath13 is a @xmath14 matrix where @xmath15 is the number of degrees of freedom . if aligning all modules this will be 6 ( three translation and three rotations ) times the number of modules . in the case of the silicon detectors
this results in a 35k @xmath1 35k matrix .
the solution of this large system of linear equations can be obtained by full diagonalization ( for example with lapack @xcite or scalapack @xcite ) which is computationally intensive or by fast solving techniques ( e.g. ma27 @xcite ) which can be performed on a standard workstation .
the fast solving methods rely on the matrix being sparse which is generally the case . solving this large matrix
is referred to as the global @xmath7 approach @xcite .
the correction of one module will be correlated to the movement of other modules and these correlations are taken into account in one go when solving this matrix .
a few iterations are generally required due to non linearities .
a second approach is the local @xmath7 approach @xcite which ignores the correlations between modules and so one only needs to invert a @xmath16 matrix for each module .
correlations are taken into account by iterating several times .
the global @xmath7 is currently the baseline approach but both approaches are implemented in atlas and give consistent results .
in addition to the @xmath7 minimization techniques , a robust alignment algorithm @xcite has also been developed .
this works by shifting modules according to their observed average residual offsets in an iterative fashion .
in particular it takes advantage of the regions where modules overlap .
the alignment sequence was as follows .
first the silicon detector ( both pixel and sct together ) was aligned internally .
a more detailed sequence of the silicon alignment is described below . for the pixel detector ,
survey information was available and was used as a starting point for the alignment .
next the trt was aligned internally .
after that , the trt was aligned with respect to the silicon detectors .
finally a `` center - of - gravity '' correction was made which adjusts the overall translation and rotation of the entire inner detector such that the aligned detector has the same center of gravity as the nominal detector .
this is needed as the minimization is insensitive to the overall translation and rotation of the inner detector .
the alignment of the detector was done in a number of steps at different levels closely following the structural assembly of the detector .
the placement of the larger structures are less precisely known than that of the precision of the assembly of the modules within their substructures .
the first level of alignment ( referred to as level 1 ) was at the level of major subsystems which were installed as separate items .
these are the 2 trt endcaps and the trt barrel , the 2 sct endcaps and the sct barrel and the whole pixel system ( however , for the silicon internal alignment the trt was not included ) .
the next level of alignment was at the disk and layer level .
each of the three pixel barrel layers were constructed from two semi - circular half shells . at this alignment level the 6 pixel layer half shells plus
the 4 sct barrel layers were aligned . due to the poor illumination of the endcaps (
since most cosmic ray muons travel predominantly in the vertical direction ) the disks within the endcap were not aligned separately but rather the endcaps were aligned as a whole ( 2 sct and 2 pixel endcaps ) . finally a module level alignment was done .
before this module level alignment was made , it was observed that the pixel staves have a significant bow lateral to the module plane which was also expected from the mechanical construction of the detector . to remain conservative
not all degrees of freedom were aligned at the module level but rather they were restricted to the two degrees of freedom which were able to correct for the stave bow .
these degrees of freedom were the angle about the normal of the module and the translation in the precision measurement direction ( local @xmath5 ) .
at the module level only barrel modules were aligned , the endcaps were again kept as a whole .
since the alignment works by minimizing residuals , the distributions of the residuals are a key test of the performance of the alignment . figures [ fig : pixbarresx ] - [ fig : trtbarresx ] show the residuals for the different detectors .
the @xmath17 quoted in the figures is the @xmath17 of the gaussian describing the core after doing a fit to a double gaussian distribution . in the case of the pixel , both
the high precision measurement direction ( local @xmath5 ) and the direction orthogonal to this ( local @xmath6 ) are shown .
the figures show the distributions before alignment where one sees wide distributions which are not centered around zero .
after the alignment the residual widths are reduced significantly and well centered on zero .
the widths are approaching that of an ideal alignment .
measurement direction of the pixel detector.,width=302 ] measurement direction of the pixel detector.,width=302 ] measurement direction of the sct detector.,width=302 ] if one takes the quadratic difference one estimates that the remaining misalignment is equivalent to a random displacement of the modules less than 20 @xmath0 .
this was further tested by generating a residual misalignment set ( labeled as `` day 1 '' ) with modules misplaced randomly with gaussian width of 20 @xmath0 in the local @xmath5 and @xmath6 directions which approximately reproduces what is seen in data as shown in figure [ fig : day1-pixel ] .
it is expected , however , that the remaining misalignment is not just random misplacement , but quite likely also includes systematic distortions .
this is discussed further in section [ sec : systematicdistortions ] .
direction for the pixel for perfect alignment , the aligned data and monte - carlo with `` day 1 '' residual misalignment of the order 20 @xmath0.,width=302 ] data from cosmic ray runs in 2009 were also recently processed using the alignments obtained with the 2008 data . while the residual widths were slightly increased the mean positions were still well centered on zero indicating that the detector has been quite stable over an extended period of time .
the reduction in width and the centering of residual is a necessary condition to demonstrate a good alignment .
however , it is not sufficient as a number of systematic distortions are insensitive or weakly constrained by the minimization of the @xmath7 .
cosmic ray tracks have the unique feature that many cross the upper and lower parts of the detector .
one can take tracks that pass close to the origin ( i.e. where beam particles collide in collision events ) and split the track into a lower and upper segment and then refit these as two independent tracks .
it is possible then to compare the track parameters .
this was done for tracks with @xmath18 gev , @xmath19 mm and @xmath20 mm , where @xmath21 is the transverse impact parameter measured with respect to the origin and @xmath22 is the @xmath3 position at the point of closest approach to the origin .
this cut ensures that the track has at least gone through the first pixel layer .
the result of this procedure for the impact parameter , where one looks at the difference of the impact parameter of the two tracks , is shown in figure [ fig : deltad0 ] . before alignment
there is a large shift away from zero and a very broad distribution .
after alignment the width is significantly reduced .
a shift of 11 @xmath0 is however observed , indicating there is still some improvement needed which is still under investigation .
since there are two tracks , an estimate of the impact parameter resolution can be obtained from the width of this distribution divided by @xmath23 .
this result in a resolution of 35 @xmath0 . for comparison , the impact parameter resolution from collision events as estimated from monte - carlo is 20 @xmath0 for tracks with @xmath24 gev .
figure [ fig : deltaphi ] shows the azimuthal angle , @xmath2 , of the track and again good improvement is seen after alignment and the distribution is well centered around zero .
figure [ fig : qoverpt ] shows the resolution of @xmath25 ( the charge over momentum ) as a function of the transverse momentum , @xmath26 , as obtained with this track splitting method .
both tracks reconstructed with the full inner detector ( i.e. including the trt ) and those reconstructed with only the silicon detector are shown .
the trt , due to its large lever arm , is seen to significantly improve the resolution , especially at large momentum . comparing tracks reconstructed with the full inner detector using perfectly aligned monte - carlo and
tracks reconstructed in data , one can see that at low momentum the agreement is very good . at low momentum multiple scattering
dominates and as expected the alignment is not as crucial for these tracks .
however , for high momentum tracks there is less multiple scattering and the impact of the remaining residual misalignment can be seen .
as function of @xmath26 obtained from a cosmic ray track which is split into an upper and lower track and refit.,width=302 ]
as was mentioned above , the residual distributions from the cosmic ray data can be reproduced in monte - carlo by introducing random residual misalignments of the order of 20 @xmath0 ( see figure [ fig : day1-pixel ] ) .
the random residual misalignment set used in simulation was referred to as `` day 1 '' being an estimate of what is achievable on the first day of collisions given a starting alignment obtained solely from the cosmic ray data .
while it is expected that we will rapidly improve the alignment with collision data , it is interesting to investigate the effects that this `` day 1 '' alignment has on physics observables . therefore some physics channels were simulated with this misalignment set .
in addition , a set was produced with a smaller amount of misalignment as an estimate of the alignment that might be achievable after a few months of running on collision data and is labeled as `` day 100 '' alignment .
it should be cautioned , however , that there are many uncertainties to what will actually be achieved on this timescale . table [ tab : day1day100levels ] shows the size of misalignment included in these two sets .
.[tab : day1day100levels]size of misalignment for `` day 1 '' and `` day-100 '' residual misalignment sets .
modules were moved randomly with a gaussian distribution with sigma as tabulated . [ cols="^,^,^,^,^ " , ] the impact on the @xmath27 mass resolution from @xmath28 events reconstructed using only the inner detector with the `` day 1 '' and `` day 100 '' alignment sets is shown in figure [ fig : zmumu - random ] @xcite . as can be seen in the figure
the impact of misalignment in the `` day 1 '' scenario shows large degradation with a contribution to the mass resolution ( in quadrature ) of 2.2 gev . for the `` day 100 '' set
there is still some degradation , however it is much reduced with a contribution from misalignment of about 1 gev .
mass resolution in @xmath28 events for perfect alignment and for different random misalignment scenarios.,width=302 ] the impact on observables of interest to @xmath29 physics was also investigated @xcite by studying the impact on the mass resolution of @xmath30 ( see figure [ fig : jspidayxmisalign ] ) and @xmath31 in @xmath32 events where @xmath33 and @xmath34 . while there is some degradation in the `` day 1 '' sample , the dominance of the lower @xmath26 tracks in these samples results in more of a contribution from multiple scattering and less from misalignment .
the effects of misalignment from the `` day 100 '' set are reduced even further to almost an insignificant amount .
mass resolution for perfect alignment and for different random alignment scenarios.,width=302 ] a study of the impact of misalignment on @xmath35-tagging was made with alignment sets different from those described above .
the results of this study @xcite are summarized in figure [ fig : b - tagging ] which shows the rejection of light jets ( jets originating from gluons , up , down and strange quarks ) for a fixed @xmath35-tagging efficiency .
the light jet rejection is defined as the inverse of the efficiency of a light jet being tagged as a @xmath35-jet .
different @xmath35-tagging algorithms were used .
the ip2d algorithm is based on the transverse impact parameter significance ( the impact parameter divided by its error ) .
the ip3d algorithm combines transverse impact parameter significance with the longitudinal impact parameter significance
. the sv1 algorithm uses properties of secondary vertices such as the vertex mass and the number of tracks used to make the vertex .
the ip3d - sv1 combines the information from the ip3d and sv1 taggers . in this
study four alignment sets were considered .
for the first set , random10 , modules were displaced and rotated randomly with a distribution with width of about 10 @xmath0 .
layers and disks and the whole pixel layer were also displaced by a similar amount . only misalignment in the pixel subsystem
was considered .
taking into account the misplacements at several levels , the overall misalignment is considered to be comparable to the `` day 1 '' set above .
the second set , random5 , introduces misalignment about half the size .
the third set is the case of a perfectly aligned detector .
the forth set , aligned , was produced using a simulation where large misalignments were introduced which were typical of what is expected when building the detector and then running the actual atlas alignment algorithms with a mixture of collision - like monte - carlo and cosmic ray monte - carlo .
for the random10 set the degradation for the ip3d - sv1 algorithm is significant ( about 50% ) but still one is able to achieve reasonable rejection rates . for the more realistic alignment case ( aligned ) the degradation is only marginal with a 15% degradation .
the impact parameter based taggers appear to be more sensitive to misalignment than the secondary vertex based taggers . -tagging efficiency of 60% for different tagging algorithms and different alignment scenarios in @xmath36 events .
the different @xmath35-tagging algorithms and alignment scenarios are described in the text.,width=302 ] while random misalignments cause some degradation to the performance of physics , it is also likely there will be systematic distortions which could be difficult for the alignment to completely eliminate .
any deformation which will still allow helical tracks to be reconstructed will leave residuals unchanged .
the minimization of the @xmath7 is not sensitive to such deformations .
these deformations are also referred to as weak modes .
one such deformation is a twist of the detector where the detector is systematically rotated as a function of the global @xmath3 coordinate .
another such deformation is a curl deformation where each subsequent layer is rotated progressively by larger amounts as a function of the radius .
this particular deformation creates a momentum bias .
tracks with one charge get a larger momentum and tracks with the opposite charge get a smaller momentum .
such a deformation was introduced into an alignment set and the results on the mass reconstruction of @xmath27 in @xmath28 events was investigated .
the size of the curl introduced was such that the outer most silicon layer was shifted around 300 @xmath0 .
this set is labeled as `` curl large '' .
in addition , the atlas alignment algorithms were run starting with a geometry with this curl misalignment .
it was seen that the alignment procedures were able to remove much of the deformation indicating that the deformation was not a perfect weak mode .
the alignment set obtained after running the alignment is labeled as `` curl small '' .
the `` curl large '' is seen to give both a degradation in the resolution and introduces a bias in the mass .
the `` curl small '' shows improvement but still some degradation with respect to the perfect alignment .
the mass bias however is mostly removed . when running the alignment to produce the ` curl small ' set , only collision - like monte - carlo was used .
the addition of events from cosmic rays is expected to improve the alignment further .
mass resolution for curl systematic misalignment scenarios.,width=302 ]
the atlas inner detector has successfully taken a large amount of cosmic ray data which has allowed a first full inner detector alignment to be achieved using the standard atlas alignment algorithms .
the residuals obtained after alignment show much reduced widths and are well centered on zero .
the resulting alignment is equivalent to less than a 20 @xmath0 residual misalignment in the barrel .
the results from comparing track parameters from upper and lower segments also show good performance .
the alignment is already at a level where it is possible to analyse low @xmath26 physics channels such as those of interest for @xmath29 physics and the alignment is expected to rapidly improve with collision data .
the tackling of systematic deformation is still expected to be a challenge although combining collision data with cosmic ray data is expected to help .
+ 99 the atlas collaboration , g. aad , _ et al .
, _ _ the atlas experiment at the cern large hadron collider , _ 2008 jinst 3 s08003 .
p. brckman , a. hicheur , s. haywood , _ global @xmath37 approach to the alignment of the atlas silicon tracking detectors _ , atl - indet - pub-2005 - 002 ( 2005 ) .
e. anderson , _ et al . , _ _ lapack users guide _ , philadelphia pa , society for industrial and applied mathematics , 1999 .
blackford , _ et al . ,
_ _ scalapack users guide _ , philadelphia pa , society for industrial and applied mathematics , 1997 .
duff and j.k .
reid , rep .
aere r10533 , hmso , london , 1982 .
a. bocci , w. hulsbergen , _ trt alignment for sr1 cosmics and beyond , _ atl - indet - pub-2007 - 009 ( 2007 ) .
r. hrtel , diploma thesis , tu mnchen , 2005 .
t. gttfert , diploma thesis , universitt wrzburg , 2006 . f. heinemann , _ track based alignment of the atlas silicon detectors with the robust alignment algorithm , _ atl - indet - pub-2007 - 011 ( 2007 ) . the atlas collaboration , _ the impact of inner detector misalignments on selected physics , _ atl - phys - pub-2009 - 080 ( 2009 ) .
the atlas collaboration , _ expected performance of the atlas detector , trigger and physics , _ cern - open-2008 - 020 , geneva , 2008 . | the atlas detector , built at one of the interaction points of the large hadron collider , is operational and has been collecting data from cosmic rays .
this paper describes the track based alignment of the atlas inner detector tracker which was performed using cosmic rays collected in 2008 .
the alignment algorithms are described and the performance of the alignment is demonstrated by showing the resulting hit residuals and comparing track parameters of upper and lower segments of tracks .
the impact of the alignment on physics measurements is discussed . |
high @xmath2 superconductivity in intermetallic layered structures attracted much attention since the discovery of superconductivity in magnesium diboride ( mgb@xmath0 ) in 2001 @xcite .
however , mgb@xmath0 keeps being the intermetallic material with the highest superconducting temperature ( @xmath7 k ) and even is well known that this property is based on the strong coupling of its covalent b sigma bands with b bond - stretching modes @xcite , it has not been possible to predict a new diboride type to enhance it . on the other hand , it is widely accepted that the superconductivity on intercalated systems as cac@xmath8 is dominated by the interaction of free - electron - like interlayer states with soft intercalant modes . lately , in this context , new proposals for the quest of layered intercalated superconductors are being merged .
the most recent theoretical study reports that , by decorating graphene with li atoms , a superconductor at a much higher temperature with respect to ca - covered graphene is obtained @xcite .
the li - be layered compounds with a low - dimensional electronic structure and strong electron - phonon ( e - ph ) coupling , are another family of li - based compounds with possible superconducting properties under high pressure @xcite .
more recently , there has been considerable interest in the structural stability , electronic band structure and chemical bond of li - b compounds under high pressures @xcite .
layered li - based materials with a quasi - two - dimensional electronic structure and high dynamical energy scales resulting from the li - atom motion are potential candidates for high @xmath2 superconductors . in that direction ,
first - principles calculations using data - mining approach have predicted new hypothetical intermetallic materials @xcite called metal - sandwich ( ms ) structures ms1(mb ) and ms2(m@xmath0b@xmath0 ) , which also have @xmath9-layers of boron as in mgb@xmath0 but separated by two metal layers _
particularly , lib and li@xmath0b@xmath0 resulted marginally stable under ambient conditions and they become favored over the known stoichiometric compounds under pressure @xcite . both compounds resulted remarkly interesting since the @xmath4- and @xmath10-bands derived from its boron @xmath11- and @xmath12-orbitals were similars to those in mgb@xmath0 @xcite and therefore a similar superconductor mechanism based on the electron - phonon coupling between the @xmath4- and @xmath10-bands and the @xmath5 phonon mode in the boron planes was expected .
recent calculations revealed that the electron - phonon coupling ( @xmath6 ) in li@xmath0b@xmath0 was weaker than in mgb@xmath0 , estimated about @xmath13 @xcite in comparison to those reported for magnesium diboride ranging from @xmath14 to @xmath15 @xcite .
thus , the transition temperature of li@xmath0b@xmath0 was reported to be around 7 k @xcite .
now it is also widely accepted that in mgb@xmath0 the electron - phonon coupling take a large contribution from @xmath4-bands but also a small participation from @xmath10-bands@xcite .
therefore the seemingly unimportant @xmath10-electrons at @xmath16 could be determinant to increase the transition temperature of li@xmath0b@xmath0 .
thus , an appropiate doping of li@xmath0b@xmath0 , in which @xmath10-electrons are increased , could result in a higher @xmath2 than in mgb@xmath0 .
this hypothesis was tested by calandra _
_ @xcite substituting li atoms with mg or al .
nevertheless , the tests showed that it was difficult to induce a significant amount of @xmath10-states at @xmath16 with small doping because the band crossing in li@xmath0b@xmath0 happens to be exactly at @xmath16 , around @xmath17 ev lower than in mgb@xmath18 .
furthermore , even from a thermodynamical stand point , doping li2b2 might be as difficult as it is for mgb@xmath19 . from the mcmillan equation @xcite ( eq .
[ eq : lambda ] ) , li@xmath0b@xmath0 would be optimistic compared to mgb@xmath0 , because li@xmath0b@xmath0 ( 0.48 states / ev ) has a higher density of states of b atoms than mgb@xmath0(0.39 states / ev ) at @xmath16 and this could enhance the electron - phonon coupling constant . however , this is not enough for arising t@xmath20 .
one interesting case is represented by cabesi @xcite , a material with @xmath4- and @xmath10-bands at the fermi level and density of states twice higher than in mgb@xmath0 , but with a very low critical temperature t@xmath20= @xmath21 k. thus , while the band structure can present strong similarities with both @xmath4- and @xmath10-bands crossing the fermi level , the phonon structure and the e - ph interaction could differ substantially . in this paper , we have analyzed why the t@xmath20 of li@xmath0b@xmath0(7 k ) has been reported lower than for mgb@xmath0(39 k ) and if it was possible to enhance the e - ph coupling of li@xmath0b@xmath0 and consequently , the superconducting temperature by applying hydrostatic pressure . since the in - plane motion of the boron atoms changes the boron orbitals overlap , an important electron - phonon coupling could be expected mainly for the @xmath4- bands at @xmath16 and the @xmath5 phonon mode . after a careful analysis of the pressure effects on the band structure and the @xmath5 phonon frequency
, we found that the electron - phonon coupling and t@xmath20 decrease with pressure as a result of the strong reduction of the boron-@xmath11 states contribution to the density of states at fermi level and the hardening of the @xmath5 phonon frequency as pressure was applied .
we studied the li@xmath0b@xmath0 compound ( crystal structure : ms2 ) as model of layered lithium borides .
li@xmath0b@xmath0 has eight atoms in the primitive unit cell and a space group p6@xmath22/mmc ( no.194 ) .
wickoff positions : li1(4f ) ( @xmath23 ) , b1(2b ) ( @xmath24 ) and b2(2d ) ( @xmath25 ) , see fig .
[ fig : fig1 ] .
the structural parameters were fully relaxed through a molecular dynamics scheme : @xmath26=@xmath27=@xmath28 , @xmath29=@xmath30 ; @xmath31=@xmath32=@xmath33 , @xmath34=@xmath35 .
the total energy calculations were performed with siesta ( spanish initiative for electronic simulations with thousands of atoms ) code @xcite based on density functional theory with exchange - correlation functionals as parametrized by perdew , burke , and ernzerhof @xcite for the generalized gradient approximation ( gga ) @xcite .
this code is implemented with pseudopotentials to describe electron - ion interactions and numerical atomic orbitals to expand the valence wavefunctions .
pseudopotentials were generated according to the troullier - martins procedure @xcite using the atomic configurations 1@xmath362@xmath37 for li and 1@xmath362@xmath362@xmath38 for b with a core radii of 2.49 and 1.24 atomic units ( a.u . ) for li and b , respectively . as the basis set for the valence wavefunctions we have employed a double-@xmath39 basis .
to calculate the integrals of the khon - sham hamiltonian , we have used a * k*-space sampling via the monkhorst - pack matrix ( @xmath40 ) with a meshcutoff of @xmath41 ry @xcite .
all calculations were numerically converged when interatomic forces were less than @xmath42 mev / .
li@xmath0b@xmath0 phase was calculated as a function of cell volume @xmath43 and @xmath44 ratio .
we performed total energy calculations for fifteen volumes and seven @xmath44 ratios in order to optimize both .
subsequently , the total energy was fitted to the murnaghan equation of state @xcite to obtain the equilibrium volume @xmath45 and the @xmath46 equation of state ( eos ) .
thus , from the @xmath46 eos the unit cell volume and @xmath44 ratio were estimated from @xmath47-@xmath48 gpa .
the @xmath5-phonon mode was calculated using the frozen - phonon model @xcite that , even is a very simple model , it is still a suitable and useful approach @xcite .
this approach is a direct method in which the distorted li@xmath0b@xmath0 is treated as a crystal of a lower symmetry than the non - distorted li@xmath0b@xmath0 .
thus , total energy calculations of distorted crystals were performed to simulate the dynamics of the @xmath5 phonon mode .
the oscillating pattern of the @xmath5 mode corresponds to boron ions in opposite directions along @xmath49- or @xmath50-axis , with li ions stationary ( fig .
[ fig : fig2 ] ) . for this , we have simulated the @xmath5-phonon mode as a composition of two modes : @xmath5(a ) and @xmath5(b ) .
figure 2 represents the vibration pattern of these modes . to implement the displacement pattern of these modes , we defined an adimensional distortion parameter @xmath51 related to the boron bond , where @xmath52 is the magnitude of the distortion and @xmath26 the in - plane lattice parameter , using the unit cell volume and @xmath44 ratio for @xmath47 , @xmath53 and @xmath48 gpa previously optimized .
these distortions simulated expansions and contractions of @xmath5 phonon mode as a function of the @xmath52 parameter .
thus , to calculate the effective potential due to boron atoms oscillation we use a maximum adimensional displacement of @xmath54=@xmath55 and steps of @xmath56=@xmath57 .
eight different displacements were calculated to simulate the vibrational dynamics of the @xmath58 and @xmath59 component modes and the total energy values of these configurations were substracted to the total energy of non - distorted li@xmath0b@xmath0 when its boron atoms are on its equilibrium positions .
the energy differences determined the restitutive potential of each oscillator and they were used to solve numerically the schr@xmath60dinger equation and obtain the characteristic frequency of each mode .
the superconducting critical temperature of li@xmath0b@xmath0 under pressure was evaluated using the mcmillan s relation @xcite , @xmath61,\ ] ] where @xmath62 is the coupling constant between @xmath4-band and the @xmath5 mode , @xmath63 is the average frequency of the @xmath58 and @xmath59 component modes and @xmath64 is an adimensional empirical value which represents the repulsive force between cooper pairs . to determine the coupling constant ( @xmath62 ) quantitatively , we evaluated the deformation potential @xmath65 for @xmath5(a ) and @xmath5(b ) modes under each distortion and used the isotropic limit from the eliashberg theory @xcite with the mcmillan s formula @xcite , @xmath66d^2,\ ] ] where , @xmath67 , @xmath68 , @xmath69 and @xmath70 are the average of the phonon frequency of the @xmath5(a ) and @xmath5(b ) modes , the deformation potential of the @xmath4-band due to @xmath5 mode , the density of states of boron atoms at fermi level and total mass of boron atoms inside unit cell , respectively .
in fig . [ fig : fig3 ] , we show the results for the @xmath26 and @xmath29 lattice parameters , and the @xmath71 ratio for li@xmath0b@xmath0 as a function of pressure for @xmath72 gpa .
it can be seen that @xmath26 parameter decreases from 3.087 to 3.071 while @xmath29 parameter from 11.187 to 8.0 in the range of 20 gpa .
although both , @xmath26 and @xmath29 parameters decreases , it is important to note that the slope for @xmath29 parameter as a function of pressure is higher than for @xmath26 parameter , that is , it would be easier to compress li@xmath0b@xmath0 through @xmath73-direction than in plane direction .
conversely , @xmath71 decreases non - linearly and li@xmath0b@xmath0 could be compressed about @xmath74 when it is applied over 20 gpa . the band structure of mgb@xmath0 and li@xmath0b@xmath0
are shown in fig .
[ fig : fig4 ] .
both structures are as similar as it was reported before @xcite .
li@xmath0b@xmath0 presents the @xmath10- and @xmath4-bands which are coupled with @xmath5 phonon mode to induce superconductivity in mgb@xmath0 .
however , in li@xmath0b@xmath0 the sigma band is shifted up by 0.23 ev and the pi - band is shifted down by 1.31 ev , with respect to mgb@xmath0 .
this shifting produces that the @xmath10-@xmath75 crossing occurs practically at fermi level , resulting in a small density of states from @xmath10-band at @xmath16 .
hence , to evaluate the effects of pressure on the electronic band structure , in fig .
[ fig : fig5 ] the evolution of the @xmath4- and @xmath10-bands with pressure is shown . from that , it can be observed the energy position relative to @xmath16 of the @xmath4- bands at @xmath76 @xmath77 and @xmath10-bands at @xmath3 @xmath78 as a function of pressure .
these results demonstrate that both energies , @xmath79 and @xmath80 , decrease as pressure increases
. in particular , @xmath79 changes from 1.03 ev at 0 gpa to 0.21 ev at 20 gpa , and @xmath80 changes from 0.11 ev at 0 gpa to @xmath81 ev at 20 gpa . to provide an explanation of the energy changes of @xmath4- and @xmath10-bands in terms of the charge redistribution induced by the pressure
, we estimated the charge population of the valence orbitals by using mulliken populations @xcite , as shown on table [ tab : tab1 ] . a charge transfer from b to li atoms
is observed .
the b-@xmath11 and b-@xmath82 orbitals loose charge , transferring it to the li orbitals , mainly to li-@xmath83 orbital .
it should be pointed out that the values provided in the mulliken analysis are strongly dependent on the basis set , but trends observed in the charge transfer process are reliable .
+ @l*15@l atom & pressure [ gpa ] & q & @xmath83 & @xmath84 & @xmath85 & @xmath82 & @xmath86 + li & 0 & 1.021 & 0.334 & 0.256 & 0.256 & 0.175 & - + & 10 & 1.070 & 0.381 & 0.279 & 0.279 & 0.131 & - + & 20 & 1.180 & 0.461 & 0.299 & 0.299 & 0.120 & - + b & 0 & 5.957 & 1.819 & 1.459 & 1.459 & 0.999 & 0.218 + & 10 & 5.859 & 1.835 & 1.441 & 1.441 & 0.930 & 0.213 + & 20 & 5.641 & 1.765 & 1.366 & 1.366 & 0.958 & 0.185 + the total and partial density of states of b , b-@xmath11 and b - p@xmath87 orbitals as a function of pressure inside unit cell are shown in fig .
[ fig : fig6 ] . from that , it is clear that at 0 gpa the main contribution at @xmath16 is coming from the @xmath4-band . for 0 gpa ,
the contributions of the b-@xmath11 and b-@xmath82 orbitals to the density of states ( dos ) at @xmath88 are @xmath89 and @xmath90 states / ev , respectively .
we find that pressure induce a strong reduction of the b-@xmath11 and b-@xmath82 orbitals contribution to the dos at @xmath88 .
thus , pressure of 20 gpa induces a reduction of 50% and 37% in the b-@xmath11 and b-@xmath82 orbitals contributions , respectively , with respect to the 0 gpa pressure values .
these results are summarized on table [ tab : tab2 ] . from these results
, we find that the behavior of the @xmath4- and @xmath10-bands with pressure correlates with the charge transfer and the reduction of states at fermi level of the partial dos corresponding to the b-@xmath11 and b-@xmath82 orbitals .
thus , as the coupling constant @xmath6 is directly proportional to @xmath69 and this one is weighted by contributions of b-@xmath11 and b-@xmath12 , the present result did not represent an advance in our quest to increase @xmath6 in comparison with mgb@xmath91 .
from eq .
( [ eq : lambda ] ) , we can see that the other parameters controlling @xmath6 and therefore @xmath2 are : the phonon frequency @xmath92 $ ] and the deformation potential @xmath68 .
@l*15@l pressure [ gpa ] & b(@xmath11 ) [ states / ev ] & b(@xmath82 ) [ states / ev ] + 0 & 0.352 & 0.075 + 10 & 0.271 & 0.036 + 20 & 0.178 & 0.028 + to further investigate @xmath5 phonon frequency value , in fig .
[ fig : fig7 ] is plotted the energy as the atoms are displaced by an adimensional amount @xmath51 according to @xmath5(a ) and @xmath5(b ) modes .
usually , phonons are harmonic for small displacements of the atoms with respect to equilibrium , but anharmonic behavior is observed for large displacements . a different source of anharmonicity
take place for atomic displacement patterns with non symmetric atomic environment , that is , for the cases when the atom experiences a different force field for positive and negative displacements with respect to the equilibrium position .
thus , in the present case , we find that the total - energy as a function of the distortions for the @xmath5(a ) and @xmath5(b ) phonon modes show an asymmetric and symmetric behavior , respectively ( see fig . [ fig : fig7 ] ) . for the @xmath5(a ) mode , the distortion energy could be reasonably approximated by the anharmonic relation @xmath93 . while for the @xmath5(b ) mode , it could be fitted to @xmath94 . using the calculated well potential for each pressure and numerically solving the schr@xmath60dinger equation
, we obtained @xmath95=79.48 $ ] mev ( 641.06 @xmath96 ) when li@xmath0b@xmath0 is free of pressure and @xmath95=93.95 $ ] mev ( 757.78 @xmath96 ) and @xmath95=104.39 $ ] mev ( 841.98 @xmath96 ) for 10 and 20 gpa , respectively .
therefore , the obtained results show that the hydrostatic pressure causes an increment in @xmath5(a ) phonon frequencies , and the rising does not depend linearly on the increasing pressure .
similarly , we obtained @xmath97=83.04 $ ] mev ( 669.78 @xmath96 ) when li@xmath0b@xmath0 is non - distorted and @xmath97=93.95 $ ] mev ( 757.78 @xmath96 ) and @xmath97=104.78 $ ] mev ( 845.13 @xmath96 ) for 10 and 20 gpa , respectively .
harmonic and anharmonic frequencies were estimated for @xmath5(b ) mode but they resulted to be equal and linearly dependent of pressure .
thus , because the frequency is directly proportional to the superconducting critical temperature , these relationship could be a good way to enhance @xmath2 . despite that
, it was quite important to quantify the deformation potential from the in - plane motions of the @xmath5 mode changes the boron orbital overlap and therefore the pairing superconductor . to determine the deformation potential due to vibrations of the @xmath5(a ) and @xmath5(b ) modes phonons
, we considered that only @xmath4-band is coupled to @xmath5 phonon , even that in mgb@xmath0 it has been well established that @xmath4- and @xmath10-bands are coupled with the @xmath5 mode .
we made this approximation because of two reasons : 1 ) lackness of @xmath10-states , and 2 ) @xmath4-states are the main contribution to the electron - phonon coupling in mgb@xmath0 @xcite . thus , by estimating the average energy of the @xmath4-band through @xmath76-@xmath98 path for each distortion , we determined the deformation potential as suggested by khan - allen @xcite , @xmath99 . as it is shown on table [ tab : tab3 ] , the deformation potential increased slightly as a result of an increase of electron concentration when pressure arise .
however , most remarkly is that average values of @xmath68 for @xmath5 in 0@xmath10020 gpa range are quite close to the reported value of 12.88 ev / for mgb@xmath0 @xcite and then this contribution would be at least comparable with magnesium diboride .
finally , we evaluated the electron - phonon coupling @xmath6 of the @xmath4-band and the @xmath5 phonon whithin isotropic limit of the eliashberg theory with the mcmillan s equation and the values of @xmath92 $ ] , @xmath69 and @xmath68 calculated previously .
these results are also shown on table [ tab : tab3 ] . from these data , it is suggested that the reduced density of states of b-@xmath11 at @xmath16 and the enhanced frequency of the @xmath5 phonon mode are the main causes for the decreasing of @xmath6 . to establish a connection between @xmath2 and @xmath6 , we considered the strong coupling approach from eliashberg theory , eq .
[ eq : tc ] . by adopting the conventional value of coulomb repulsion @xmath101 and two additional of @xmath102 and @xmath103 @xcite , we estimated @xmath2 at different pressures .
the results are summarized on table [ tab : tab4 ] .
@xmath2 was found to decrease from 45 to 3.6 k with a null coulomb repulsion , from 18.6 to 0.01 k with @xmath103 and from 10.95 to 0.07 k with @xmath104 . with all choices of @xmath105
, the resulting @xmath2 has a continously reduced trend and practically vanishes at 20 gpa .
@l*15@l p[gpa ] & @xmath69[@xmath106 & @xmath107[mev ] & @xmath108 $ ] & @xmath62 + 0 & 0.48 & 81.26 & 12.88 & 0.57 + 10 & 0.35 & 93.95 & 13.17 & 0.32 + 20 & 0.29 & 104.58 & 13.21 & 0.22 + @l*15@l p[gpa ] & @xmath102 & @xmath103 & @xmath104 + 0 & 45 & 18.6 & 10.95 + 10 & 12 & 0.98 & 0.11 + 20 & 3.6 & 0.01 & 0.07 +
in conclusion , we have presented a first - principle investigation of the electronic structure , phonon frequency of the @xmath5 mode , the deformation potential of the @xmath4-band and the electron - phonon coupling @xmath6 for li@xmath0b@xmath0 under pressure from 0@xmath10020 gpa within the framework of density functional theory .
we predicted a reduction of @xmath6 from 0.57 at 0 gpa to 0.22 at 20 gpa .
even the lattice dynamics of the @xmath5 phonon mode resulted similar to mgb@xmath0 , the reduced density of states of b-@xmath11 at fermi level and the raising of the phonon frequency corresponding to the @xmath5 mode with pressure were determinant to decrease @xmath6 and consequently @xmath2 .
one of the authors ( e.m.g . ) gratefully acknowledges financial support from the consejo nacional de ciencia y tecnologa ( conacyt - mxico ) through a posdoctoral scholarship ( no .
also , e.m.g . would like to thank to ramiro quijano quiones and omar de la pea seaman for technical assistance and useful discussions .
this research was partially funded by conacyt - mxico under grant no .
83630 and no .
200 nagamatsu j , nakagawa n , muranaka t , zenitari y and akimitsu j 2001 _ nature _ * 410 * 63 an j m and picket w e 2001 _ phys .
lett . _ * 86 * 4366 kortus j , mazin i i , belashchenko k d , antropov v p and boyer l l 2001 _ phys
lett . _ * 86 * 4656 baron a q r , uchiyama h , tsusui s , tanaka y , ishikawa d , sutter j p , lee s , tajima s , heid r and bhonen k p 2007 _ physica c _ * 456 * 83 floris a , sanna a , l@xmath109ders m , profeta g , lathiotakis n n , marques m a l , franchini c , gross e k u , continenza a and massidda s 2007 _ physica c _ * 456 * 45 kortus j 2007 _ physica c _ * 456 * 54 de la pea o , aguayo a and de coss r 2002 _ phys . rev .
b _ * 66 * 012511 chepulskii r v and curtarolo s 2009 _ phys . rev .
b _ * 79 * 134203 bersier c , floris a , sanna a , profeta g , continenza a , gross e k u and massidda a 2009 _ phys . rev . b _ * 79 * 104503 profeta g , calandra m and mauri f 2012 _ nat . phys . _
* 8 * 131 feng j , henning rg , ashcroft nw and hoffmann r 2008 _ nature _ * 451 * 445 xu y , chen c and wu b 2012 _ solid state commun . _ * 152 * 151 hermann a , suarez - alcubilla a , gurtubay ig , yang lm , bergara a , ashcrof nw and hoffmann r 2012 _ phys . rev .
b _ * 86 * 144110 hermann a , mcsorley a , ashcrof nw and hoffmann r 2012 _ j. am .
soc . _ * 134 * 18606 peng f , miao m , wang h , li q and ma y 2012 _ j. am .
soc . _ * 134 * 18599 kolmogorov a n and curtarolo s 2006 _ phys .
b _ * 73 * 180501(r ) kolmogorov a n and curtarolo s 2006 _ phys . rev . b _ * 74 * 224507 calandra m , kolmogorov a n and curtarolo s 2007 _ phys . rev . b _ * 75 * 144506 liu a y and mazin i i 2007 _ phys . rev . b _ * 75 * 064510 wang y , plackowski t and junod a 2001 _ physica c _ * 355 * 179 bouquet f , fisher r a , phillips n e , hinks d g and jorgensen j d 2001 _ phys
_ * 87 * 047001 yildrim t , g@xmath110lseren o , lynn j w , brown c m , udovic t j , huang q , rogado n , regan k a , hayward m a , slusky j s , he t , hass m k , khalifah p , inumaru k and cara r j 2001 _ phys .
* 87 * 037001 choi h j , roundy d , sun h , cohen m l and louie s g 2002 _ nature _ * 418 * 758 de la pea - seaman o , de coss r , heid r , and bohnen kp 2010 _ phys . rev .
b _ * 82 * 224508 soler j m , artacho e , gale j d , garca a , junquera j , ordejn p and snchez portal d 2002 _ j. phys . : condens . matter _ * 14 * 2745 ordejn p , artacho e and soler j m 1996 _ phys .
b _ * 53 * 10441 perdew j p , burke k and ernzerhof m 1996 _ phys . rev .
* 77 * 3865 langreth d c and mehl m j 1983 _ phys .
b _ * 28 * 1809 becke a d 1988 _ phys .
a _ * 38 * 3098 perdew j p , chevary j a , vosko s h , jackson k a , pederson m r , singh d j and fiolhais c 1992 _ phys . rev .
b _ * 46 * 6671 perdew j p , chevary j a , vosko s h , jackson k a , pederson m r , singh d j and fiolhais c 1993 _ phys .
b _ * 48 * 4978(e ) troullier n and martins j l 1991 _ phys . rev .
b _ * 43 * 1993 monkhorst h j and pack j d 1976 _ phys . rev .
b _ * 13 * 5188 murnaghan d f 1944 _ proc
usa _ * 30 * 244 kunc k and martin r m 1983 _ physica b&c _ * 117 * 511 eliashberg g m 1960 _ sov .
. jetp _ * 11 * 696 mcmillan w l 1968 _ phys
_ * 167 * 331 mulliken r s 1955 _ j. chem .
* 23 * 1833 kong y , dolgov o v , jepsen o and andersern o k 2001 _ phys .
b _ * 64 * 020501(r ) liu a y , mazin i i and kortus j 2001 _ phys . rev .
_ * 87 * 087005 golubov a a , kortus j , dolgov o v , jepsen o , kong y , andersen o k , gibson b j , ahn k and kremmer r k 2002 _ j.phys . : condens . matter _ * 14 * 1353 khan f s and allen p b 1984 _ phys .
b _ * 29 * 3341 de la pea - seaman o , de coss r , heid r , and bohnen k.p .
2007 _ j. phys . : condens
. matter _ * 19 * 476216 | we present the structural , electronic and superconducting properties of li@xmath0b@xmath0 under pressure within the framework of the density functional theory .
the structural parameters , electronic band structure , phonon frequency of the @xmath1 phonon mode and superconducting critical temperature @xmath2 were calculated for pressures up to 20 gpa .
we predicted that the superconducting critical temperature of li@xmath0b@xmath0 is about 11 @xmath3 and this decreases as pressure increases .
we found that even though the lattice dynamics of the @xmath1 phonon mode is similar to mgb@xmath0 , the reduction of the @xmath4-band density of states at fermi level and the raising of the @xmath5 phonon frequency with pressure were determinant to decrease @xmath6 and consequently @xmath2 . |
since the discovery of the optical counterpart of grb 970228 ( van paradijs et al .
@xcite ) the accurate localizations provided by the _ beppo sax _ satellite have led to the detection of the optical afterglow for more than ten gamma
ray bursts ( hereafter grbs ) .
the most spectacular result of these observations is to have provided a direct proof of the cosmological origin of grbs .
the detection of absorption lines at @xmath0 in the spectrum of grb 970508 ( metzger et al .
@xcite ) followed by other redshift determinations ( between @xmath1 and @xmath2 ) confirmed the indications which were already available from the _ batse _ data showing a grb distribution perfectly isotropic but non homogeneous in distance ( fishman and meegan @xcite and references therein ) .
the energy release of grbs with known redshifts extends from @xmath3 to @xmath4 .
the solid angle @xmath5 in which the emission is beamed is quite uncertain .
a small @xmath5 should reveal itself by a break after a few days in the afterglow light curve .
a break is indeed observed in a few cases such as grb 990510 ( harrison et al .
@xcite ) where @xmath6 could be as small as 0.01 . however , most afterglows do not show any break which means that @xmath5 is usually not very small ( @xmath7 ? ) .
+ the source of cosmic grbs must therefore be able to release a huge energy in a very short time .
possible candidates include the coalescence of two neutron stars ( eichler et al .
@xcite ; paczyski @xcite ) , the disruption of the neutron star in a neutron star black hole binary ( narayan et al .
@xcite ; mochkovitch et al .
@xcite ) or the collapse of a massive star ( woosley @xcite , paczyski @xcite ) . in all these cases
the resulting configuration is expected to be a stellar mass black hole surrounded by a thick disc . since the power emitted by grbs is orders of magnitude larger than the eddington limit it can not be radiated by a static photosphere .
the released energy generates a fireball which then leads to the formation of a wind .
moreover , this wind has to become highly relativistic in order to avoid the compactness problem and produce gamma
rays ( baring @xcite ; sari & piran @xcite ) .
values of the lorentz factor as high as @xmath8@xmath9 are required , which limits the allowed amount of baryonic pollution to a remarkably low level .
only a few mechanisms have been proposed to produce a wind under such severe constraints : ( _ i _ ) magnetically driven outflow originating from the disc or powered by the blandford
znajek ( @xcite ) process ( thomson @xcite ; mszros & rees @xcite ; daigne & mochkovitch @xcite ; lee et al .
@xcite ) ; ( _ ii _ ) reconnection of magnetic field lines in the disc corona ( narayan et al .
@xcite ) ; ( _ iii _ ) neutrino antineutrino annihilation in a funnel along the rotation axis of the system ( mszros & rees @xcite ; mochkovitch et al .
@xcite , @xcite ) .
mechanisms ( _ i _ ) and ( _ ii _ ) require that the magnetic field in the disc reaches very high values @xmath10 . our preliminary study ( daigne & mochkovitch @xcite ) of the wind emitted from the disc shows that it can avoid baryonic pollution only if some very severe constraints on the dissipation in the disc and the field geometry are satisfied . some recent works ( ruffert et al .
@xcite ) have also shown that mechanism ( _ iii _ ) is probably not efficient enough to power a gamma ray burst , except may be for the shortest events .
when the wind has reached its terminal lorentz factor , the energy is mainly stored in kinetic form and has to be converted back into gamma rays .
two main ideas have been proposed to realize this conversion .
the first one is the so - called external shock model ( rees & mszros @xcite ; mszros & rees @xcite ) .
the wind is decelerated by the external medium , leading to a shock .
gamma rays are emitted by the accelerated electrons in the shocked material through the synchrotron and/or inverse compton mechanisms .
this model has been studied in details ( fenimore et al .
@xcite ; panaitescu et al .
@xcite ; panaitescu & mszros @xcite ) and seems unable to reproduce some important features of grbs such as their strong temporal variability ( see however dermer & mitman @xcite ) .
conversely , the external shock model reproduces very well the delayed emission at lower energy from the afterglows ( mszros & rees @xcite ; wijers et al .
@xcite ) .
the second proposal is the internal shock model ( rees & mszros @xcite ) where the wind is supposed to be formed initially with a very inhomogeneous distribution of the lorentz factor .
rapid parts of the wind then catch up with slower ones leading to internal shocks where gamma
rays are again produced by synchrotron or inverse compton radiation .
we have started a study of this model in a previous paper ( daigne & mochkovitch @xcite , hereafter dm98 ) where the wind was simply made of a collection of `` solid '' shells interacting by direct collisions only ( all pressure waves were suppressed ) .
the very encouraging results we obtained had to be confirmed by a more detailed study .
we have therefore developed a relativistic hydrocode to follow the evolution of the wind .
we present the code and the main results in this paper .
we write in sect .
2 the lagrangian equations of hydrodynamics in special relativity . in sect .
3 we describe the numerical method we use to solve them and we present the tests we performed to validate the method .
we display our results in sect . 4 and sect .
5 is the conclusion .
we write in a fixed frame the equations of mass , momentum and energy conservation in spherical symmetry : @xmath11 @xmath12 @xmath13 where @xmath14 and @xmath15 are respectively the spatial and temporal coordinates in the fixed frame .
the following quantities appear in eqs .
[ eqvm][eqem ] : @xmath16 is the pressure in the fluid local rest frame , @xmath17 is the fluid velocity in the fixed frame and @xmath18 , @xmath19 and @xmath20 are the specific volume , momentum density and energy density ( including mass energy ) in the fixed frame .
these three quantities are related to quantities in the fluid local rest frame : @xmath21 @xmath22 @xmath23 where @xmath24 is the lorentz factor , @xmath25 is the rest - mass density and @xmath26 is the specific enthalpy density ( @xmath27 being the specific internal energy density ) .
the lagrangian mass coordinate @xmath28 is defined by @xmath29 @xmath30 being the radius of the back edge of the wind .
the system of eqs [ eqvm][eqem ] is completed by the equation of state @xmath31 the adiabatic index @xmath32 being a constant . in the non - relativistic limit ( @xmath33 and @xmath34 ) , the quantities @xmath18 , @xmath19 and @xmath20 become equal to their newtonian counterparts @xmath35 , @xmath17 and @xmath36 and eqs .
[ eqvm][eqem ] then reduce to the classical equations of lagrangian hydrodynamics in spherical symmetry .
the great similarity between the relativistic and classical equations will allow us to use the powerful numerical methods which have been developed in classical hydrodynamics to follow the evolution of a fluid with shocks .
an extension of the piecewise parabolic method of colella and woodward ( @xcite ) to 1d eulerian relativistic hydrodynamics in planar symmetry has been already presented by mart and mller ( @xcite ) .
we follow exactly the same procedure to extend the ppm to the lagrangian case in spherical symmetry .
we adopt @xmath37 as the set of variables .
if the mass - averaged values @xmath38 of @xmath39 at time @xmath40 in each cell @xmath41 $ ] extending from @xmath42 to @xmath43 are known , the values @xmath44 at time @xmath45 are computed in four steps : + the variables @xmath46 are obtained from @xmath39 in each cell by solving the following equation in @xmath47 : @xmath48 once @xmath47 is known , the other quantities are easily computed since @xmath49 and @xmath50 .
these mass - averaged values are then interpolated by polynomials in the way described in the original paper by colella and woodward ( @xcite ) .
we use the same modifications of the coefficients of the interpolation polynomials , leading to a steeper representation of discontinuities and a monotone representation of smoother parts .
+ at each interface @xmath51 , two effective states @xmath52 and @xmath53 ( @xmath54 and @xmath14 denote the left and right sides of the interface ) are constructed by mass - averaging these quantities in the region of cells @xmath55 and @xmath56 connected to the interface @xmath51 by a characteristic line during the time step .
+ at each interface @xmath51 , the two effectives states @xmath52 and @xmath53 define a riemann problem ( two constant states separated by a discontinuity surface ) , which is known to give rise , like in newtonian hydrodynamics , to two new states @xmath57 and @xmath58 separated by a contact discontinuity and related to the initial states @xmath52 and @xmath53 either by a shock or a rarefaction wave .
the common values of the pressure and the velocity of the two new intermediate states are given by the implicit equation @xmath59 an analytic expression ( for a polytropic gas ) of @xmath60 ( where @xmath61 either refers to the @xmath54 or @xmath14 state ) has been worked out by mart and mller ( @xcite ) .
equation ( 11 ) is solved using brent s method to obtain the pressure @xmath62 and the velocity @xmath63 of the intermediate states at each interface .
+ the quantities @xmath64 are calculated with numerical fluxes at each interface obtained from @xmath62 and @xmath63 in the same way than colella and woodward ( @xcite ) .
we have sucessfully checked our code against two usual tests in relativistic hydrodynamics .
the first one is the shock tube problem which is simply a riemann problem in which the initial states are at rest .
we present in fig.[figtest1 ] the results for @xmath65 , @xmath66 , @xmath67 and @xmath68 , @xmath69 , @xmath70 .
the adiabatic index is @xmath71 and the discontinuity is initially located at @xmath72 .
the figure is plotted at a time @xmath73 for a grid of 1000 zones initially equally spaced .
the agreement between the exact and numerical profiles is satisfactory .
the positions of the contact discontinuity and the shock are very accurate
. the density , pressure and velocity of the post - shock state are also exact .
however , the value of the density in the immediate vicinity of the contact discontinuity shows a small non - physical increase , which is more pronounced when the shock is stronger and disappears when the shock is weak ( as in the shock tube problem with @xmath74 , @xmath75 and @xmath68 , @xmath76 , which has been considered by several authors ) . in the context of the internal shock model for grbs , the shocks are only mildly relativistic and
we do not observe any unexpected increase of the density in the results presented below .
this test consists in a cold fluid , which is initially homogeneous ( @xmath77 ) and enters a sphere of radius 1 at constant velocity @xmath78 .
the fluid bounces at @xmath79 and is heated up .
we present in fig.[figtest2 ] the results for @xmath80 , @xmath81 and @xmath82 at @xmath83 .
the adiabatic index is @xmath84 and we used a grid of 1000 zones initially equally spaced . in the considered case of a cold homogeneous fluid , an analytical solution is known .
the shocked state is at rest with a density @xmath85 and a pressure @xmath86 . at time
@xmath15 the unshocked cold fluid of velocity @xmath78 has a distribution of density @xmath87 .
the numerical profiles appear accurate except in the vicinity of the origin .
the shock propagates with the correct velocity and the post - shock values of density and pressure are well reproduced .
we therefore conclude that the treatment of the geometrical terms in eqs [ eqvm][eqem ] is correct .
we have not tried to improve the computation near the center , which is not of major importance in the context of the internal shock model for grbs where most of the emission takes place far from the origin .
we have used our code to follow the evolution of a relativistic wind with a very inhomogeneous initial distribution of the lorentz factor .
the first results have been already presented for small values of the lorentz factor @xmath88 ( daigne & mochkovitch @xcite ) . here
we describe our results for the large lorentz factors ( @xmath89 ) which are relevant for the study of grbs .
we first consider the case of a simple single - pulse burst .
whatever the initial event leading to a grb may be ( ns ns or ns bh merger , `` hypernova '' , etc ) , the system at the end of this preliminary stage is probably made of a stellar mass black hole surrounded by a thick disc ( the `` debris '' torus ) .
we consider that @xmath90 , a substantial fraction of the available energy of the system , is injected at a typical radius @xmath91 into a wind emitted during a duration @xmath92 with a mass flow @xmath93 .
we do not discuss here the physical processes controlling @xmath93 , @xmath92 and @xmath90 but we assume that the baryonic load @xmath94 is very small .
the wind converts its internal energy into kinetic energy during its free expansion in the vacuum ( the effect of the interstellar medium is negligible at this early stage ) and accelerates until it reaches a lorentz factor @xmath95 at a typical radius @xmath96 ( mszros et al .
this is where our simulation starts .
more precisely , we define our initial state as follows .
we consider that from @xmath97 to @xmath98 , a wind with a distribution of the lorentz factor defined by @xmath99 has been produced by the source and that its back edge has reached @xmath100 km ( we adopt @xmath101 ) .
we suppose that energy is injected at a constant rate @xmath102 ( we adopt @xmath103 in the following ) , so that the total energy injected into the wind simply equals @xmath104 and the injected mass flux is @xmath105 .
the density profile in this initial state can be calculated if we assume that the internal energy is very small compared to the kinetic energy , which is indeed the case when the wind has reached its terminal lorentz factor ( @xmath106 ) .
the eulerian and lagrangian profiles of @xmath107 in the wind at @xmath98 are shown in fig.[figinitial ] for @xmath108 and @xmath109 .
we have adopted @xmath110 and have checked that the results do not depend on this small value .
the fast part of the wind catches up with the slower one .
the matter is strongly compressed is the collision region , the velocity gradient becomes very steep and at @xmath111 , two shocks appear : a forward shock reaching the front edge at @xmath112 and a reverse shock reaching the back edge at @xmath113 .
the hot and dense matter behind these two internal shocks radiates and produces the observed burst .
the radiation losses are not taken into account in the dynamics , which is probably not a too severe approximation since the dissipated energy represents about 10% of the total kinetic energy of the wind .
when the two shocks have reached the edges , the evolution becomes unimportant regarding the emission of gamma - rays : two rarefaction waves develop at each edge and the wind continues to expand and cool .
in fact , at this stage , the interstellar medium should absolutely be included in the calculation .
an external shock propagates into the ism which produces the afterglow and a reverse shock crosses the wind which can also lead to an observable emission .
all these effects are not included in the present simulation , which is stopped when the two internal shocks have reached the wind edges .
we present in fig.[figdynamique ] ( left panel ) the paths in a @xmath15 @xmath114 plot ( @xmath115 is the arrival time of photons emitted at time @xmath116 on the line of sight at a distance @xmath14 from the source ) of the two shocks and of the two edges of the wind . in the right panel the corresponding distributions of @xmath107 and @xmath25 are plotted at different times .
consider an internal shock located at a distance @xmath117 from the source at a time @xmath118 in the fixed frame .
the density @xmath119 , the lorentz factor @xmath120 and the specific internal energy @xmath121 of the shocked ( @xmath122 ) and unshocked ( @xmath61 ) material are known from our hydrodynamical simulation .
this shock will produce a contribution to the grb which will be observed at an arrival time @xmath123 and which will last @xmath124 where @xmath125 is the lorentz factor of the emitting material for which we adopt @xmath126 .
the luminosity of the shock is estimated by @xmath127 where @xmath128 is the mass flux across the shock and @xmath129 is the dissipated energy per unit mass in the frame of the shocked material .
our code detects all the internal shocks present in the wind at a given time and saves their parameters in order to sum all the contributions to the emission and produce a synthetic gamma - ray burst . in a recent paper ( dm98 )
we presented a simple model where the wind was idealized by a collection of `` solid '' shells interacting by direct collision only ( i.e. all pressure waves were neglected ) .
we detailed in this previous paper our assumptions to treat the emission of a given shock .
we adopt here the same assumptions .
the magnetic field in the shocked material is supposed to reach equipartition values @xmath130 with @xmath131 .
the lorentz factor of the accelerated electrons is calculated using the expression given by bykov & mszros ( @xcite ) who consider the scattering of electrons by turbulent magnetic field fluctuations : @xmath132^{1/(3-\mu)}\ , \label{eqbykov}\ ] ] where @xmath133 @xmath134 is the fraction of the dissipated energy which goes into the magnetic fluctuations ; @xmath135 is the fraction of the electrons which are accelerated and @xmath136 ( @xmath137 ) is the index of the fluctuation spectrum . for @xmath138 and @xmath139 ,
( 17 ) corresponds to the usual equipartition assumption , leading to @xmath140 of a few hundreds . in this case
, the emission of gamma - rays could result from inverse compton scattering on synchotron photons .
bykov and mszros however suggests that only a small fraction @xmath141 of the electrons may be accelerated , leading to @xmath140 values of several thousands . in this last case , synchrotron radiation can directly produce gamma - rays of typical energy @xmath142 we now present a detailed comparison of the results of our hydrodynamical code with those previously obtained with the simple model ( dm98 ) for a single pulse burst .
we have plotted in fig .
[ figquantiteshydro ] the values of @xmath143 , @xmath107 , @xmath144 and @xmath25 as function of @xmath114 for the forward and reverse shocks .
we observe an overall similarity between the two calculations , despite the crude approximations of the simple model .
not surprisingly , the worst estimated quantities are the post - shock density and the dissipated energy per proton , which are underestimated by a factor of @xmath145 .
conversely , the emission time and the lorentz factor of the emitting material are correctly reproduced .
the emission starts earlier in the simple model where there is no preliminary phase of compression before the formation of shocks ( this leads to a larger underestimate of the density at the very beginning of the simulation ) , and ends later .
the total efficiency of the dissipation process is also smaller @xmath146 instead of @xmath147 for the detailed model .
the other quantities @xmath148 , @xmath140 and @xmath149 are not directly given by the hydrodynamical simulation but are parametrized by @xmath150 , @xmath151 , @xmath135 and @xmath136 , whose values are unknown .
to make a useful comparison between the two series of results , we take the same @xmath150 and @xmath136 in the two cases but adjust @xmath152 so that the typical synchrotron energy is the same .
the corresponding values of @xmath148 , @xmath140 and @xmath149 are represented in fig .
[ figequipartition ] with @xmath153 , @xmath154 and @xmath155 for the hydrocode and @xmath9 for the simple model . as expected because of the differences in density and dissipated energy , the magnetic field
is underestimated by a factor of @xmath156 in the simple model .
this is corrected by our choice of parameters for @xmath140 and the resulting synchrotron energies are very similar in the two cases . also notice that the efficiency of the synchrotron process is smaller in the simple model due to a poor estimate of the mass flux accross the shock .
the agreement between the two calculations is satisfactory and allows to be quite confident in the results of the simple model .
compared to the hydrodynamical code , the simple model has very short computing times and enables a detailed exploration of the temporal and spectral properties of synthetic bursts which was presented in our previous paper ( dm98 ) .
we show in the next section the detailed results obtained with the hydrocode in the case of a single pulse burst .
the contributions of the forward and the reverse shocks are added to construct the synthetic burst .
we assume that the photons emitted from @xmath15 to @xmath157 by an internal shock of current luminosity @xmath158 are distributed according to a simple power - law spectrum @xmath159 where we adopt @xmath160 or @xmath161 ( the two extreme low energy index that are expected for a synchrotron spectrum ) for @xmath162 and @xmath163 for @xmath164 ( @xmath165 in the following ) .
we take into account cosmological effects ( time dilation and redshift ) assuming that the burst is located at @xmath166 .
we have plotted in fig.[figc23_10s]a the photon flux observed in batse bands 2 + 3 for the initial distribution of lorentz factor shown in fig.3 , calculated either with the hydrocode ( with @xmath167 ) or the simple model ( with @xmath168 ) .
the two profiles look similar but the hydrodynamical code leads to a slower decay . with @xmath169 ( resp .
@xmath170 ) being the time when 5% ( resp 95 % ) of the total fluence has been received , we obtain a duration @xmath171 instead of @xmath172 with the simple model .
figure [ figc23_10s]b illustrates that the exponential decay of the burst is also nicely reproduced with the detailed calculation .
however , if we define @xmath173 as the time of maximum count rate and @xmath174 and @xmath175 as the rise and the decay times , we get a ratio @xmath176 .
dm98 found that a larger value of @xmath177 and a corresponding profile closer to the characteristic `` fast rise exponential decay '' ( fred ) shape is obtained by assuming that the fraction @xmath135 of accelerated electrons increases with the dissipated energy per proton @xmath144 .
as in dm98 we adopt @xmath178 , so that @xmath140 is independent of @xmath144
. figures [ figc23_10s]c and [ figc23_10s]d show the resulting profiles with @xmath179 for the hydrodocode and @xmath180 for the simple model .
the profile then better reproduces a typical fred shape . the observed tendency of short bursts to become symmetric ( norris et al .
@xcite ) has been tested in dm98 with the simple model .
the basic behaviour was reproduced but the effect was even exagerated since , for @xmath181 , @xmath177 was smaller than unity i.e. the decline was faster than the rise .
as can be seen in fig .
[ figsymmetry ] and [ figevolution ] , the situation is improved with the hydrocode since now @xmath182 for @xmath183 . however
, the shortest bursts are still asymmetric with @xmath184 for @xmath185 .
figure [ figsymmetry ] also shows that the ratio @xmath177 is limited to a maximum value of @xmath186 for the longest bursts which appears to be in contradiction with the short rise times observed in some cases .
as discussed in dm98 , an initial distribution of the lorentz factor with a steeper gradient than the one used here ( eq . ( 12 ) ) can indeed increase @xmath177 but extreme values ( such as @xmath177 possibly larger than @xmath187 in grb 970208 ) might still be difficult to reproduce .
in dm98 we presented a complete study of the global and instantaneous spectral properties of synthetic bursts calculated with the simple model .
since these spectral properties are hardly different when calculated with the hydrocode , we do not present them in detail again .
we just show in fig .
[ figspectrum ] the shape of the global spectrum calculated for the single pulse burst .
despite the very simple form adopted for the instantaneous spectrum ( eq . [ eqspectrum ] ) , the sum of all the elementary contributions produces an overall spectrum with a more complex shape , which is well reproduced with band s formula ( band et al .
@xcite ) @xmath188^{\alpha-\beta } \exp{\left(\alpha-\beta\right ) } \left(\frac{e}{100\
{ \rm kev}}\right)^{\beta}\,\nonumber\\ & & { \rm for}\ e \ge ( \alpha-\beta ) e_{0}\ .\end{aligned}\ ] ] we find values of the parameters comparable to those observed in real bursts . the best fits in fig .
[ figspectrum ] correspond to @xmath189 , @xmath190 and @xmath191 in case ( a ) ( @xmath192 ) and @xmath193 , @xmath194 and @xmath195 in case ( b ) ( @xmath196 ) .
as @xmath197 is limited to the range @xmath198 , we can not get spectra with low energy slopes flatter than @xmath199 as they are observed in several bursts ( preece et al .
@xcite ) . a more detailed description of the radiative processes
is then needed to reproduce these extreme slopes ( an attempt to solve this problem is proposed by meszaros & rees @xcite ) .
+ however , even with the crude modelization of the instantaneous spectra which is used here , it has been shown in dm98 that several spectral properties of grbs are reproduced . in particular ,
the hard to soft evolution during a pulse and the change of pulse shape as a function of energy as well as the duration
hardness ratio relation which appears as a natural consequence of the internal shock model .
these important spectral features are confirmed in our detailed hydrodynamical calculation .
an important property of the internal shock model is its ability to produce a great variety of temporal profiles .
norris et al .
( @xcite ) have shown that complex bursts can generally be analysed in terms of a series of ( possibly overlapping ) simple pulses .
this result is readily interpreted in the context of the internal shock model .
a wind made of a succession of fast and slow shells will produce a succession of pulses which will add to form a complex burst .
we present such examples of complex bursts in fig .
11 and 12 . the first one ( fig .
[ figcomplex1 ] ) is produced by an initial distribution of the lorentz factor made of five consecutive identical patterns .
each pattern made of a slow and a rapid part produces its own individual pulse and the resulting burst has a complex shape with five , very similar , pulses .
our second example ( fig .
12 ) uses the same type of initial distribution of the lorentz factor but the slow parts now have non equal @xmath107 values .
the resulting burst is more realistic with four pulses of different intensities .
we did not treat with the hydrocode a large number of cases as we did with the simple model .
nevertheless we confirmed the essential result that the variability introduced in the initial distribution of the lorentz factor is present in the burst profile with the same time scale .
the profile therefore appears as a direct indicator of the activity of the central engine .
this paper is the continuation of our study of the internal shock model started in dm98 .
we developed a 1d lagrangian relativistic hydrocode ( in spherical symmetry ) to validate our previous simpler approach where all pressure waves were neglected in the wind .
our code is an extension of the classical _ ppm _ method of colella and woodwards ( @xcite ) in the spirit of the work by mart & mller ( @xcite ) for the eulerian case in planar symmetry .
a detailed comparison has been made between the hydrocode and the simple model in the case of a single pulse burst .
it appears that the dynamical evolution of the wind is well reproduced by the simple model , which is not too surprising because the wind energy is largely dominated by the kinetic part so that the effect of pressure waves is small .
only one physical quantity the density of the shocked material is strongly underestimated in the simple model . in order to make valuable comparisons between the two calculations
we have therefore adjusted the equipartition parameters so that the mean value of the synchrotron energy is the same in the two cases
. the synthetic bursts which are then obtained are very similar which proves that our first approach was essentially correct and confirm our previous results .
a similar conclusion was reached by panaitescu and mszros ( @xcite ) who performed a comparable study .
the internal shock model can easily explain the great temporal variability observed in grbs .
the main characteristic features of individual pulses are well reproduced : ( 1 ) pulses have typical asymmetric `` fred '' profiles ; ( 2 ) the pulse width decreases with energy following a power
law @xmath200 with @xmath201 ; ( 3 ) short pulses show a tendency to become more symmetric .
our model still gives very short pulses which decay faster than they rise but the hydrodynamical simulation improves the situation compared to the simple model .
spectral properties of grbs are also well reproduced .
we obtain synthetic spectra which can be nicely fitted with band s function with parameters comparable to those observed in real grbs . the spectral hardness and the count rate are correlated during the evolution of a burst with the hardness usually preceeding the count rate . as also pointed in dm98 , the duration
hardness relation is a natural consequence of the internal shock model .
these results are very encouraging and the main difficulty which remains is the low efficiency ( about @xmath202 ) of the internal shock model . as long as the energetics of grbs and the mechanism initially operating in the central engine are not precisely identified , we can not say if this is a critical problem or not .
we still believe that the internal shock model is at present the most convicing candidate to explain the gamma ray emission from grbs .
next steps in this work will address the following questions .
we first want to extend our hydrodynamical code to a non
adiabatic version in order to include the radiative losses in the dynamical calculation .
we have already developed an `` isothermal rieman solver '' for that purpose ( daigne & mochkovitch @xcite ) .
we would also like to study the effects of the external medium , with a special attention to the reverse shock which propagates into the wind and possibly interacts with the internal shocks .
preliminary results with the simple method using `` solid layers '' have already been obtained ( daigne & mochkovitch @xcite ) but they have to be confirmed by a hydrodynamical calculation . finally , we would like to investigate the details of the emission process during internal shocks to solve some of the problems encountered by the synchrotron model .
band d. et al . , 1993 , apj , 413 , 281 baring m. , 1995 , apss , 231 , 169 blandford r.d . , znajek r. , 1977 , mnras , 179 , 433 bykov a. , mzros p. , 1996 ,
apj , 461,l37 colella p. , woodward p.r . , 1984 , j. comput .
54 , 174 daigne f. , mochkovitch r. , 1997 , proceedings of the fourth huntsville gamma - ray burst symposium daigne f. , mochkovitch r. , 1998 , mnras , 296 , 275 ( dm98 ) daigne f. , mochkovitch r. , 1999 , to appear in the proceedings of the 19th texas symposium , astro ph 9907289 daigne f. , mochkovitch r. , 1999 , a&ass , 138 , 523 dermer c.d .
, mitman k.e . , 1999 , apj , 513 , l5 eichler d. , livio m. , piran t. , schramm d. , 1989 , nat , 340 , 126 fenimore e. , madras c. , nayakshin s. , 1997 , apj , 473 , 998 fishman g.j . , meegan c.a .
, 1995 , ara&a , 33 , 415 harrison f.a .
et al . , 1999 ,
apj , 523 , l21 lee h.k . , wijers r.a.m.j . , brown g.e . , 1999 ,
ph 9906213 mart j.m . , mller e. , 1994 , j. fluid mech .
258 , 317 mart j.m . , mller e. , 1996 , j. comput
123 , 1 mszros p. , laguna p. , rees m.j . , 1993 ,
apj 415 , 181 mszros p. , rees m.j . , 1997 ,
apj , 482 , l29 mszros p. , rees m.j . , 1992 , mnras , 257 , 29 mszros p. , rees m.j . , 1993 , apj , 405 , 278 mszros p. , rees m.j . , 1997 ,
apj , 476 , 232 mszros p. , rees m.j .
, 2000 , apj , 530 , 292 metzger m.r .
, djorgovski s.g . ,
kulkarni s.r . , steidel c.c . ,
adelberger k.l . ,
frail d.a . ,
costa e. , frontera f. , 1997 , nat , 387 , 878 mochkovitch r. , hernanz m. , isern j. , martin x. , 1993 , nat , 361 , 236 mochkovitch r. , hernanz m. , isern j. , loiseau s. , 1995 , a&a , 293 , 803 narayan r. , paczyski b. , piran t. , 1992 , apj , 395 , l83 norris j. , nemiroff r. , bonell j. , scargle j. , kouveliotou c. , paciesas w. , meegan c. , fishmann g. , 1996 , apj , 459 , 393 paczyski b. , 1991 , acta astron . , 41 , 257 paczyski b. , 1998 , apj , 494 , l45 panaitescu a. , wen l. , laguna p. , mszros p. , 1997 ,
apj , 482 , 942 panaitescu a. , mszros p. , 1998 ,
apj , 492 , 683 panaitescu a. , mszros p. , 1999 , apj , astro
ph 9808106 preece r.d . , briggs m.s . ,
mallozzi r.s . ,
pendleton g.n .
, paciesas w.s . ,
band d.l . ,
2000 , apjs , 126 , 19 rees m.j .
, mszros p. , 1992 ,
mnras , 258 , 41p rees m.j . , mszros p. , 1994 ,
apj , 430 , l93 ruffert m. , janka h.t . ,
takahashi k. , schaeffer g. , 1997 , a&a , 319 , 122 sari r. , piran t. , 1997 , mnras , 287 , 110 thomson c. , 1994 , mnras , 270 , 480 van paradijs j. , et al . , 1997 , nat , 386 , 686 wijers r. , rees m.j .
, mszros p. , 1997
, mnras , 288 , l51 woosley s.e . , 1993 ,
apj , 405 , 273 | the internal shock model for gamma - ray bursts involves shocks taking place in a relativistic wind with a very inhomogeneous initial distribution of the lorentz factor .
we have developed a 1d lagrangian hydrocode to follow the evolution of such a wind and the results we have obtained are compared to those of a simpler model presented in a recent paper ( daigne & mochkovitch @xcite ) where all pressure waves are suppressed in the wind so that shells with different velocities only interact by direct collisions .
the detailed hydrodynamical calculation essentially confirms the conclusion of the simple model : the main temporal and spectral properties of gamma - ray bursts can be reproduced by internal shocks in a relativistic wind . |
steam reforming is an important industrial process , where natural gas ( ch@xmath7 ) is converted into synthesis gas ( co and h@xmath2 ) according to the overall reaction @xmath8 the synthesis gas is subsequently transformed into more valuable chemicals , such as ammonia , methanol , and diesel . since steam reforming acts as a source of hydrogen , it is also potentially important for any emerging hydrogen economy . for a detailed review of the steam reforming process and its applications the reader can consult ref . .
commercially the steam reforming reaction is conducted over a ni - based catalyst due to the relatively low cost and good activity of nickel .
this reaction has been studied in detail on the close - packed ni(@xmath4 ) surface and on the stepped ni@xmath9 ) surface by means of density - functional theory.@xcite the major technological challenge for ni catalysts is the formation of carbonaceous deposits , termed coke , that lead to catalyst deactivation . in refs .
it was established that the step edges on ni surfaces act as growth centers for graphite .
other transition metals ( tm ) , such as ru , rh , pd , ir , and pt also show high activity and selectivity towards steam reforming@xcite and have a high resistance against carbon formation .
however , these materials are scarce in nature and expensive
. therefore new materials that are resistant to carbon formation are needed .
transition - metal carbides ( tmc s ) have gained quite some attention since levy and boudart reported that they have " platinum - like behavior for certain reactions.@xcite the starting material for the production of carbides is cheap and abundant and therefore it has been suggested that they can replace the noble metals in catalysis .
the main problem with the carbides has been to produce materials with sufficiently high surface area for them to be interesting for catalytic applications .
this problem has been overcome and several studies report on carbides with surface areas as high as 200 m@xmath10g.@xcite in refs . and , it was shown that carbides of molybdenum and tungsten are stable and extremely active catalysts for not only the steam reforming but also the dry reforming and the partial oxidation of methane at elevated pressure . moreover ,
no macroscopic carbon was deposition on the catalysts during the catalytic reactions .
the relative activity of a number of steam reforming catalysts was established as : ru @xmath11 rh @xmath11 ir @xmath12 mo@xmath2c @xmath11 wc@xmath11 pd @xmath11 pt .
several density - functional theory studies concerning adsorption on tmc s have been conducted , for example , atomic adsorption,@xcite o@xmath2 adsorption,@xcite co adsorption,@xcite nh@xmath13 ( @xmath14 ) adsorption,@xcite methanol adsorption,@xcite ch@xmath13 ( @xmath15 ) and c@xmath2h@xmath7 adsorption.@xcite in addition , the water - gas - shift reaction has been studied in refs . . to our knowledge , a theoretical understanding of the steam reforming reaction on tmc s is missing in the literature .
the aim of this paper is to investigate how the tmc s are suited for the steam reforming reaction by performing a set of density - functional theory ( dft ) calculations on different surfaces of @xmath1-mo@xmath2c , tic , vc , and @xmath6-moc in order to catch the trends between tmc s composed of different tm atoms and of different structures .
the influence of the surfaces being covered with oxygen is also investigated .
the outline of this paper is as follows . to begin with the systems to be investigated
are defined in sec .
[ sec : systems ] .
this is followed by sec .
[ sec : calculations ] , which describes the calculation details .
our results are presented in sec .
[ sec : results ] and discussed in section [ sec : discussion ] .
the paper ends with a conclusion in section [ sec : conclusion ] .
the mo c system can occur in several crystalline forms , see for example the phase diagrams in refs . and
. there are two types of mo@xmath2c phases , one orthorhombic ( @xmath1 ) and one hexagonal ( @xmath16 ) , which is a high temperature phase . in this study
, we focus on the @xmath1 phase . for this phase
the mo atoms are only slightly distorted from an hcp arrangement and the carbon atoms occupy one half of the octahedral interstitial sites . in the present molybdenum carbide literature
there seems to be a confusion when it comes to the attribution of the symbols to the different phases , especially regarding the @xmath1 and @xmath16 phases .
this is probably because the @xmath1-mo@xmath2c phase was originally assigned a hexagonal structure but refined experiments identified the structure to be of orthorhombic type@xcite , with lattice parameters @xmath17 , @xmath18 and @xmath19 .@xcite the close - packed surface of @xmath1-mo@xmath2c is the ( @xmath5 ) surface .
it consists of alternating mo and c layers and can be either mo- or c - terminated .
experimental studies of @xmath1-mo@xmath2c(@xmath5 ) show that the surface termination and structure depend strongly on the cleaning procedure .
both the mo- and the c - terminated surfaces have been observed .
@xcite therefore we consider both the tm- and the c - terminated carbide surfaces . the tic , vc and @xmath6-moc
adopt the nacl structure .
the ( @xmath5 ) surface of these carbides consists of layers of equal amounts of tm and c atoms .
the ( @xmath4 ) surface , however , is polar and consists of alternating tm and c layers , similar to the @xmath1-mo@xmath2c(@xmath5 ) surfaces .
one important difference is that the c - terminated @xmath1-mo@xmath2c(@xmath5 ) surface has half the amount of c atoms as compared to a c - terminated tmc(@xmath4 ) surface . in this paper , we report the results from dft calculations on ( i ) the tm- and c - terminated surfaces of @xmath1-mo@xmath2c(@xmath5 ) , tic(@xmath4 ) , vc(@xmath4 ) , and @xmath6-moc(@xmath4 ) ; ( ii ) the o - covered surfaces in ( i ) ; and ( iii ) the ( @xmath5 ) surfaces of tic , vc and @xmath6-moc .
all the dft calculations in this paper are performed with the computer code dacapo@xcite , which uses plane - waves and ultra - soft pseudopotentials . for the optimization of the bulk lattice parameters of the considered tmc s a plane - wave cutoff of @xmath20 ev , a @xmath21 @xmath22-point sampling , and the rpbe exchange - correlation ( _ xc _ ) functional@xcite
the bulk structure of @xmath1-mo@xmath2c was obtained by minimizing the total energy of the unit cell with respect to the length of the lattice vectors using a newton - raphson scheme and allowing for complete internal relaxation at each step .
the calculated lattice parameters @xmath23 , @xmath24 , and @xmath25 differ by @xmath26 from the experimental values.@xcite for tic , vc and @xmath6-moc , the equilibrium bulk structure was obtained utilizing the murnaghan equation of state.@xcite the calculated lattice constants are @xmath27 , @xmath28 , and @xmath29 for tic , vc , and @xmath6-moc , respectively .
a quantitative agreement is found between the theoretical and experimental lattice constants.@xcite the surfaces are modeled using the supercell approach , with the dft lattice parameters given above . a @xmath20 ev energy cutoff for the plane - wave expansion of the one - electron orbitals , a @xmath30 @xmath22-point grid , and the rpbe _ xc _ functional are used .
each tmc(@xmath4 ) and tmc(@xmath5 ) surface is represented by a slab with the geometry defined in fig .
[ fig : structure ] and a thickness of four atomic layers .
repeated slabs are separated by at least a @xmath31 thick vacuum region .
the mo- and c - terminated @xmath1-mo@xmath2c(@xmath5 ) slabs have a surface geometry as defined in fig .
[ fig : structure ] and consist of four layers mo and four layers c separated by @xmath32 vacuum . during structural relaxations ,
the two bottom atomic layers are constrained to the corresponding bulk geometry .
figure [ fig : structure ] shows the relaxed structures of the mo- and c - terminated @xmath6-moc(@xmath4 ) surfaces , the @xmath6-moc(@xmath5 ) surface , and the mo- and c - terminated @xmath1-mo@xmath2c(@xmath5 ) surfaces .
these relaxed surfaces and the corresponding vc and tic surfaces are used as starting points for the modeling of the steam reforming reaction .
the investigated steam reforming reaction is assumed to have the following nine elementary steps : @xmath33 where @xmath34 ( 2@xmath34 ) denotes one ( two ) empty surface site and @xmath35 denotes the adsorbate @xmath36 bonded to this site .
adsorption of all the intermediates ( ch@xmath37 , ch@xmath2 , ch , c , oh , o , co , and h ) in eqs . ( [ eq : reactionstep1])([eq : reactionstep9 ] ) are considered . in all adsorption calculations ,
the adsorbate coverage is one - quarter of a monolayer ( ml ) .
adsorption is allowed on one of the two slab surfaces at the time .
adsorbates together with the surfaces are allowed to relax in all directions until the forces are less than @xmath38 ev / . the energies in the calculated potential energy diagrams are expressed relative to the clean surface ( o - covered surfaces if the surface is oxygenated ) , a ch@xmath7 molecule , and a h@xmath2o molecule in the gas phase .
the steam reforming reaction on the oxygenated tmc surfaces is modeled by adsorbing the reaction intermediates on a 1ml o - covered tmc surface . according to dft calculations in ref . , the o covering of the mo- and c - terminated mo@xmath2c(@xmath5 ) surfaces is energetically favorable up to 1ml .
our calculations give the same structure for the 1ml o - covered mo- and c - terminated mo@xmath2c(@xmath5 ) surfaces as in ref . .
this section focuses on the results obtained from the screening calculations study .
we provide the calculated potential energy diagrams for the steam reforming reaction on all the considered tmc surfaces .
these are compared with existing results on ni(@xmath4 ) and ni(@xmath39 ) @xcite . calculated potential energy diagram for the steam reforming reaction on the tm - terminated ( circles ) and c - terminated ( squares ) tmc(@xmath4 ) and
@xmath1-mo@xmath2c(@xmath5 ) surfaces . for comparison the ni(@xmath4 ) and ni(@xmath39 ) data , adapted from ref . ,
are given . ] the calculated potential energy diagram for the steam reforming reaction on the tm- and c - terminated tmc s is shown in fig .
[ fig : tmc111 ] .
first of all we notice that the span of the potential energies of the investigated surfaces is huge . compared to the ni surfaces @xcite ,
all the tmc surfaces are much more reactive because they bind the intermediates too strongly .
therefore none of these surfaces is suitable for the steam reforming reaction . in the following
we present the trends in reactivity that are found when changing the tm atom in the tmc compound . for each of the considered tmc s
, the c - terminated surface is found to be more reactive than the tm - terminated one .
this is to be expected since the tmc(@xmath4 ) surfaces are found to be tm terminated under uhv conditions.@xcite all the tm - terminated tmc(@xmath4 ) surfaces have similar potential energy shape as a function of the reaction intermediates .
the same is found for the c - terminated group of tmc surfaces . for the tm - terminated tmc surfaces , the reactivity can be ordered as tic(@xmath4 ) @xmath11 vc(@xmath4 ) @xmath11 @xmath1-mo@xmath2c(@xmath5 ) @xmath11 @xmath6-moc(@xmath4 ) , while for the c - terminated tmc surfaces we find that the reactivity order is tic(@xmath4 ) @xmath11 vc(@xmath4 ) @xmath11 @xmath6-moc(@xmath4 ) @xmath11 @xmath1-mo@xmath2c(@xmath5 ) .
hence different trends are observed for the different terminations . the considered @xmath1-mo@xmath2c(@xmath5 )
surfaces are less reactive than the c - terminated @xmath6-moc(@xmath4 ) surface but more reactive than the mo - terminated @xmath6-moc(@xmath4 ) . for a given carbide , the smallest difference in reactivity between the tm- and
the c - terminated surfaces is found for @xmath1-mo@xmath2c(@xmath5 ) .
this can be due to the fact that the c - terminated @xmath1-mo@xmath2c(@xmath5 ) surface has @xmath40 less c than the c - terminated tmc(@xmath4 ) surfaces .
the largest minimum in the potential energy surface is found upon adsorption of o - carrying intermediates , that is , a strong o surface bond is formed .
this implies that the tmc(@xmath4 ) surfaces are easily oxidized .
experimental studies show that the o@xmath2 adsorption on tic(@xmath4)@xcite and @xmath1-mo@xmath2c(@xmath5)@xcite are dissociative .
therefore we investigate how the steam reforming reaction is influenced when the tmc(@xmath4 ) and the @xmath1-mo@xmath2c(@xmath5 ) surfaces are oxygenated .
calculated potential energy diagram for the 1ml o - covered tm - terminated ( circles ) and c - terminated ( squares ) tmc(@xmath4 ) and @xmath1-mo@xmath2c@xmath3 surfaces .
also given are the ni(@xmath4 ) and ni(@xmath39 ) data , adapted from ref . . ]
figure [ fig : otmc111 ] shows the calculated potential energy diagram for the steam reforming reaction on the 1ml o - covered tm- and c - terminated tmc surfaces .
we find that the o - covered c - terminated @xmath1-mo@xmath2c(@xmath5 ) has a potential energy profile similar to that reported for the ni(@xmath4 ) surface , which is the preferred ni facet for the steam reforming reaction according to the dft calculations in ref . .
the other o - covered surfaces are unsuitable as catalysts for the steam reforming either because several of the reaction steps are significantly uphill energetically , or because the surface gets poisoned by some of the intermediates .
some of the trends that are found for this class of surfaces are here described .
each of the o - covered tmc surfaces is less reactive than its corresponding non - oxidized surface .
for a given o - covered tmc surface the tm - terminated surface has higher reactivity than the c - terminated one , in contrast to the finding on the non - oxidized surfaces . for the o - covered tm - terminated tmc surfaces , the reactivity can be ordered as follows @xmath1-mo@xmath2c(@xmath5 ) @xmath11 @xmath6-moc(@xmath4 ) @xmath11 vc(@xmath4 )
@xmath11 tic(@xmath4 ) , while for the o - covered c - terminated tmc surfaces we find that @xmath1-mo@xmath2c(@xmath5 ) @xmath11 tic(@xmath4 ) @xmath11 vc(@xmath4 ) @xmath11 @xmath6-moc(@xmath4 ) .
the order in reactivity between the different surfaces is thus different for the tm- and for the c - terminated o - covered surfaces .
calculated potential energy diagram for tmc(@xmath5 ) .
the ni(@xmath4 ) and ni(@xmath39 ) data given here are adapted from ref . .
] the calculated potential energy diagram for the steam reforming reaction on the tmc(@xmath5 ) surfaces is shown in fig .
[ fig : tmc100 ] . the @xmath6-moc(@xmath5 ) surface is found to have a potential energy curve similar to the one of the ni(@xmath4 ) surface .
we find the same shape of the potential energy curves for all the tmc(@xmath5 ) surfaces .
a local minimum is found for all the tmc(@xmath5 ) surfaces upon adsorption of co. the reactivity of the surfaces follows the order @xmath6-moc(@xmath5 ) @xmath11 tic(@xmath5 ) @xmath11 vc(@xmath5 ) , that is , a non linear variation when moving from right to left in the periodic table .
compared to the tmc(@xmath4 ) surface the tmc(@xmath5 ) surfaces are far less reactive . in ref
it was suggested that the tic(@xmath5 ) surface is a suitable catalyst for the water - gas - shift reaction .
however , according to our calculation it is not the surface of choice for the steam reforming reaction since it is inert ( see fig .
[ fig : tmc100 ] ) .
the conducted computational screening study clearly shows that only two surfaces , the o - covered c - terminated @xmath1-mo@xmath2c(@xmath5 ) and the @xmath6-moc(@xmath5 ) , are suitable for the steam reforming reaction .
these two surfaces are found to have potential energy profiles similar to the one of ni(@xmath4 ) , which is the preferred ni surface for steam reforming according to ref . .
the obtained potential energy diagrams show that there is a large variety in the reactivities of the different carbides ( see figs .
[ fig : tmc111 ] , [ fig : otmc111 ] and [ fig : tmc100 ] ) .
however , besides the two candidate surfaces none of the other considered tmc surfaces are suitable for the steam reforming reaction since they either get self - poisoned by the intermediates or they show a significant energetical uphill for some of the elementary steps in eqs .
( [ eq : reactionstep1])([eq : reactionstep9 ] ) .
an understanding of the reactivity can be achieved by studying the electronic structure of the surface . in refs .
a thorough analysis of the electronic structure established that the tm - terminated tmc(@xmath4 ) surfaces possess surface resonances ( sr s ) .
these were identified to be responsible for strong adsorbate - surface bonds .
therefore the high reactivities of the tm - terminated tmc(@xmath4 ) surfaces ( see fig . [
fig : tmc111 ] ) are attributed to the presence of these sr s .
preliminary electronic structure calculations show that similar resonances are present also on the layered polar @xmath1-mo@xmath2c(@xmath5 ) surfaces , which would explain their high reactivities .
the nacl tmc(@xmath5 ) surfaces , on the other hand , do not posses any srs@xcite and show a much lower reactivity ( see fig . [
fig : tmc100 ] ) compared to the tmc(@xmath4 ) surfaces . since a suitable candidate , the @xmath6-moc(@xmath5 ) surface ,
is found among these surfaces we have tried to change the reactivity of the tmc(@xmath4 ) surfaces by weakening their sr s .
the lower reactivities of the o - covered tmc(@xmath4 ) and @xmath1-mo@xmath2c(@xmath5 ) surfaces ( see fig .
[ fig : otmc111 ] ) , compared to the corresponding clean surfaces can be understood as follows .
adsorption of oxygen on the tmc(@xmath4 ) results in a quenching of the sr s .
thus , the surfaces are passivated , that is , their ability to form strong bonds with other adsorbates decreases due to the strong o surface bond .
this result shows that the reactivity of carbides surfaces can be tuned .
an important question for catalytic applications is the stability of the candidate materials .
steam reforming operates over a large range of working conditions with temperatures @xmath41
@xmath42c and pressures @xmath43 bar .
the most recent phase diagram for mo
c systems is the one by velikanova _ _ et al.__@xcite it shows that the bulk @xmath1-mo@xmath2c phase exists for temperatures @xmath44 @xmath42c , while the @xmath6-moc is a high temperature phase which exists above @xmath45 @xmath42c .
based only on bulk data , @xmath1-mo@xmath2c is preferred over @xmath6-moc . when it comes to the stability of a surface the situation can of course differ from the bulk . according to york and coworkers@xcite , mo@xmath2c was active and selective for the stoichiometric steam reforming of methane to synthesis gas .
the catalyst deactivated at atmospheric pressure , but was very stable when elevated pressures were employed , and no carbon deposition was observed on the catalysts . in those studies
the surface characteristics was , however , not addressed .
a surface characterization of @xmath1-mo@xmath2c(@xmath46 has been performed in ref . .
the surface preparation consisted of ar ion bombardment followed by annealing at successively higher temperatures .
the surface was found to change from being mo terminated to c terminated at annealing temperatures above @xmath47 k. despite that these surfaces are stable at temperatures relevant for the steam reforming , our study shows that they are too reactive .
adsorption of o@xmath2 on both the mo- and the c - terminated @xmath1-mo@xmath2c(@xmath46 surfaces were found to be dissociative and the surfaces were covered with oxygen at @xmath48 k.@xcite this supports the stability of the o - covered @xmath1-mo@xmath2c(@xmath5 ) surface .
in addition , stable crystalline molybdenum carbide films prepared by chemical vapor deposition ( cvd ) and physical vapor deposition ( pvd ) have been found to contain either mo@xmath2c and/or the @xmath49-moc@xmath50 phase.@xcite hence it seems plausible that both the o - covered c - terminated @xmath1-mo@xmath2c(@xmath5 ) and the @xmath6-moc(@xmath5 ) surfaces may be stable at relevant steam reforming conditions , however , further studies are needed .
in this paper , a computational screening study of the steam reforming reaction on transition - mental carbides has been presented . by using a screening approach we are able to map out the huge reactivity space , which ranges from very reactive to inert , provided by the carbides .
a small set of surfaces , consisting of the @xmath6-moc(@xmath5 ) and the o - covered c - terminated @xmath1-mo@xmath2c(@xmath5 ) , is identified as promising candidate for the steam reforming reaction . with this
thermochemical data at hand we lay a basis for a thermodynamic and micro - kinetic treatment of the carbides .
this study illustrates the versatility and tunabilty of the tmc systems for the steam reforming reaction .
the results can be understood in terms of the previously proposed model of chemisorption on tmc surfaces , in which surface resonances play crucial roles . however ,
further studies of the surface electronic structure are needed to establish what surface characteristics that are required of a carbide to make is a suitable catalyst for steam reforming .
the calculations were performed at hpc2n via the swedish national infrastructure for computing .
the author thanks anders hellman and carlo ruberto for constructive discussions and comments on the manuscripts .
g. jones , j. g. jakobsen , s. s. shimb , j. kleis , m. p. andersson , j. rossmeisl , f. abild - pedersen , t. bligaard , s. helveg , b. hinnemann , j. r. rostrup - nielsen , i. chorkendorff , j. sehested , and j. k. nrskov , j. catal . * 259 * , 147 ( 2008 ) . | a screening study of the steam reforming reaction ( @xmath0 ) on early transition - metal carbides ( tmc s ) is performed by means of density - functional theory calculations .
the set of considered surfaces includes the @xmath1-mo@xmath2c@xmath3 surfaces , the low - index ( @xmath4 ) and ( @xmath5 ) surfaces of tic , vc , and @xmath6-moc , and the oxygenated @xmath1-mo@xmath2c@xmath3 and tmc(@xmath4 ) surfaces .
it is found that carbides provide a wide spectrum of reactivities towards the steam reforming reaction , from too reactive via suitable to too inert .
the reactivity is discussed in terms of the electronic structure of the clean surfaces .
two surfaces , the @xmath6-moc@xmath3 and the oxygen passivated @xmath1-mo@xmath2c@xmath3 surfaces , are identified as promising steam reforming catalysts .
these findings suggest that carbides provide a playground for reactivity tuning , comparable to the one for pure metals . |
the detection and analysis of large samples of cool white dwarfs over the past decade has led to a significant improvement in our knowledge of the chemical composition and evolution of their atmospheres @xcite .
the difficulties in interpreting the spectra of some very cool white dwarfs thought to posses helium - rich atmospheres with small amounts of h @xcite are most probably related to the extreme physical conditions found in their atmospheres @xcite . on the other hand ,
the physical conditions in hydrogen - rich atmospheres are much less extreme and the current models should be much more reliable . nonetheless , models systematically overestimate the flux for @xmath2 of hydrogen - rich white dwarfs of @xmath3 @xcite .
@xcite suggested that the mechanism responsible for this phenomenon could be the pseudo - continuum bound - free opacity from hydrogen atoms perturbed by collisions in a relatively dense medium .
however , a realistic calculation of this opacity source shows that it is too weak to correct the observed blue flux excess in the models @xcite .
in this paper we consider the possibility that the missing opacity is the red wing of the pressure broadened hydrogen lyman@xmath4 line .
this absorption mechanism results from the perturbations of hydrogen atoms by their interaction with other particles , primarily @xmath5 and @xmath6 .
such perturbations result in the lowering of the @xmath1 transition energy and the possibility of a bound - bound transition for photon energies that are smaller than that of an isolated hydrogen atom ( @xmath7 ) .
we present a new , semi - classical calculation of this process based on accurate quantum mechanical calculations and a model that has no free parameters .
the inclusion of broadening by collisions with @xmath6 is the key element of this calculation .
the resulting pure hydrogen models successfully reproduce the spectral energy distributions ( sed ) and colors of very cool da and dc white dwarfs .
we are interested in the @xmath1 opacity far from the line center , at wavelengths @xmath8 .
a lowering of the @xmath1 transition energy by more than @xmath9 is required for a bound - bound absorption from the ground state of hydrogen atom at these wavelengths .
this can occur in rare , close - range collisions that strongly perturb the bound states of the atom . because the ground state interactions are strongly repulsive at short range , the probability of such close - range collisions is much smaller than unity .
multi - particle collisions that result in a similar decrease in the @xmath1 transition energy are therefore insignificant @xcite and we consider the interaction between a @xmath5 atom and its closest neighbor only . this approach was used to successfully explain the complex shape of the lyman @xmath10 line wings detected in the spectra of white dwarfs with @xmath11 @xcite , the uv flux deficiency observed in the white dwarf star @xmath12 @xcite , and to model resonance broadening of the @xmath1 line in the solar spectrum @xcite . for a given colliding
pair , the change in the @xmath1 transition energy results from the formation of a temporary dimer , whose first transition energy @xmath13 differs from that of an isolated hydrogen atom , @xmath14 .
@xmath13 is given by the differences in the energies of the ground state and first excited rydberg state of a @xmath5-perturber dimer calculated at a fixed inter - particle separation @xmath15 : @xmath16 , where @xmath17 is the frequency of the absorbed photon .
this picture is in the spirit of the franck - condon principle , which states that the nuclei remain in fixed positions during a radiative transition @xcite .
the differential probability of finding such a dimer with an inter - particle separation between @xmath18 and @xmath19 is given by @xcite @xmath20 where @xmath21 is a number density of perturbers , @xmath22 is the interaction energy between a hydrogen atom and the perturber localized at the position @xmath23 in relation to the hydrogen atom .
the potential for the interaction of a hydrogen atom with h ( @xmath24 dimer ) is from @xcite and that for perturbations by h@xmath25 ( @xmath26 dimer ) is from @xcite ( fig .
1 ) . for the @xmath27 dimer ,
the allowed bound - bound transition from the ground state with the smallest transition energy ( i.e. greatest broadening ) at all values of @xmath18 is from @xmath28 to @xmath29 .
satellite features @xcite do not appear for this transition because the difference between the two energy curves has no extrema ( fig .
contributions from transitions to other excited levels of the @xmath27 dimer ( such as @xmath30 and @xmath31 ) are negligible in the far red wing ( @xmath32 ) because they always lie above the @xmath33 state by at least 2.5@xmath34ev in the relevant range of @xmath18 and hence only provide absorption for photons with higher energies than the @xmath33 state . for the same reason , transitions from the binding @xmath27 ground state ( @xmath35 ) are also negligible in the far red wing .
we use the @xmath36 energy curve from @xcite . for the energy of the first excited rydberg state of the @xmath26 system
, we used the h@xmath37 calculations of @xcite . to obtain the h
h@xmath25 potential from the h@xmath37 potential surface , we assumed a fixed internuclear separation for @xmath6 of @xmath38 this approach is justified by the fact that the atoms in @xmath6 vibrating in the ground state spend most of the time at the equilibrium internuclear separation . at the temperatures of interest ( @xmath39k ) ,
the vibrational excitation is small , and so is the amplitude of vibration ( @xmath40 ) .
we have verified that such a small change in the intermolecular separation does not affect our results significantly .
the line profile is calculated in the quasi - static approximation , which is suitable for a calculation of the far line wings perturbed by classical particles @xcite .
the profile of the line @xmath41 is given by ( @xcite , eqs .
13 & 61 ) : @xmath42 where @xmath43 is the @xmath18-dependent transition dipole moment between the ground state and the first excited rydberg state of the colliding pair , and @xmath44 is the frequency .
the dipole transition moments , which vary with separation @xmath18 , are from @xcite for @xmath27 and @xcite for @xmath26 . for a given @xmath1 transition energy ,
the @xmath26 ground state interaction is much less repulsive than the @xmath27 ground state interaction ( fig .
1 ) , and the probability of a close @xmath26 interaction is much larger than that of a close @xmath27 interaction ( @xmath45 ) . hence ,
the @xmath26 collisions are the dominant contributors to the @xmath1 line red wing opacity in the atmospheres of cool h - rich white dwarfs with a significant abundance of @xmath6 ( @xmath46 ) .
bpm 4729 ( wd 0752@xmath47676 ) is the only cool da white dwarf observed in the near uv @xcite . when complemented with the bvrijhk photometry of @xcite and the u magnitude @xcite the complete spectral energy distribution ( sed ) of this cool white dwarf
is obtained .
the uv spectrum extends well into the wing of the @xmath1 line and thus provides an excellent measure of the strength and profile of the line .
our fit of a pure h model ( @xmath48k , @xmath49 ) to the entire sed of bpm 4729 is excellent ( fig .
these values of @xmath50 and @xmath51 agree with the values of @xcite at the @xmath52 level .
the high quality of the fit of the uv spectrum demonstrates the validity of our model for the far red wing of the @xmath1 line under the conditions encountered in bpm 4729 . at the photosphere ,
the composition is @xmath53 and @xmath54 , where @xmath55 is the number density in @xmath56 . despite the lower abundance of @xmath6
, it is an important contributor to the @xmath1 line opacity in the far red wing ( fig .
2 ) . @xcite obtained a good fit of the uv and blue spectrum with a he - rich model with @xmath57 .
they point out that such a low h abundance fails to reproduce the balmer lines of bpm 4729 , however .
@xcite required a large amount of @xmath58 to match the uv spectrum because the broadening of ly @xmath59 by collisions with he is weaker than with h or h@xmath25 .
they also considered a pure hydrogen model but its ly @xmath59 line extends only up to @xmath60 . in view of the difference between the potential curves for the @xmath61 to @xmath62 states of the @xmath27 dimer ( fig .
1 ) this indicates that this particular transition was not included in the model of @xcite .
the main reason for our success with a pure h model is our inclusion of the broadening of @xmath1 line by collisions of hydrogen atoms with @xmath6 and using the b @xmath63 a transition for the h@xmath47h dimer .
we have also fitted bvrijhk photometry of most of the coolest da / dc white dwarfs in the @xcite sample .
two typical fits are shown in fig .
our models have no difficulty in reproducing the observed b flux and excellent fits are obtained for most stars with either pure h or pure he .
the importance of the @xmath1 line opacity in the models is revealed by comparing with fits of models that exclude this opacity source .
the improvement in fitting the b flux with our ly@xmath0 opacity model is clearly visible .
the star wd @xmath64 ( vb 11 , lower panel of fig .
3 ) was assigned a pure helium composition by @xcite .
we will return to this point in the next section .
the success of the pure h models indicates that the @xmath1 red wing opacity is the missing blue opacity source in models of cool h - rich white dwarf atmospheres .
color - color diagrams allow a broader comparison with data .
figure 4 shows two large samples of cool wds , that of @xcite , and a combination of two samples culled from the sloan digital sky survey @xcite .
the atmospheric composition of the stars in the first sample , as determined by @xcite , is indicated in the left panel of fig .
colors computed from our pure h model sequence , both with and without the @xmath1 opacity , are shown .
our sequence of new pure hydrogen atmospheres follows the observed sequence of stars very well in these two color - color diagrams .
equally good agreement is found in all other color - color diagrams involving bvrijhk and _ ugriz _ colors ( not shown here ) . on the other hand ,
the sequence of pure helium models of @xcite ( see also @xcite ) used in the analysis of @xcite and @xcite also reproduces the observed sequence in the bvi color - color diagram .
this potential ambiguity in the atmospheric composition of cool dc stars is resolved with a new sequence of pure he models .
we have computed pure he models based on much improved constitutive physics for dense helium , including a better calculation of the ionization equilibrium and of the he@xmath65 ff and rayleigh scattering absorptions @xcite
. a detailed description of these models will be the subject of a future publication .
qualitatively , the correlations in dense fluid he reduces the contribution of raleigh scattering by a factor of @xmath66 @xcite and the strong interactions in the fluid increase the ionization fraction ( and hence , the he@xmath65 ff opacity ) by 2 - 3 orders of magnitude @xcite . the combination of these effects make he@xmath65 ff the only important source of opacity in these models . on the other hand , because of a lower ionization fraction and the higher rayleigh scattering opacity in a dilute gas , the pure he model spectra of @xcite are affected by both opacities ( see their fig .
since he@xmath65 ff is a nearly gray opacity , our pure he atmosphere models are essentially gray and the emergent flux is close to that of a black body . the colors of this pure he sequence are shown in fig . 4 along with those of black bodies . compared to the @xcite pure he sequence , to our pure h sequence , and to the observed sequence of very cool wds , this new pure he sequence is much redder for @xmath67k . as the ionization fraction is sufficiently large for the he@xmath65 ff opacity to dominate , the colors of our new he sequence are insensitive to modest pollution by metals since increasing the fraction of free electrons will only increase the he@xmath65 ff opacity without any effect on the emergent spectrum . on the basis of the location of the pure h and pure he model sequences in the color - color diagrams ( fig .
4 ) and the excellent fit we obtain for wd2054@xmath47050 with a pure hydrogen model , it appears that the atmospheric composition of the coolest dc stars needs to be revisited .
the existence of an unidentified absorption mechanism in cool hydrogen white dwarfs atmospheres was reported a decade ago @xcite .
the interpretation of this missing opacity as the pseudo - continuum absorption from hydrogen atoms has been recently shown to be incorrect @xcite .
on the other hand , the red wing of the @xmath1 line opacity from hydrogen could provide the required absorption @xcite .
we present a new calculation of the extreme pressure - broadening of the @xmath68 line by both @xmath5 and @xmath6 .
when included in our new pure hydrogen atmosphere models , we obtained an excellent agreement with the uv / optical / near - ir spectral energy distribution ( sed ) of the cool da white dwarf bpm 4729 and we successfully fitted the b through k spectral energy distributions of stars with hydrogen - rich atmospheres , as determined by @xcite and @xcite .
the inclusion of broadening by collisions with h@xmath25 is essential to reproduce the sed of very cool hydrogen white dwarfs . in color - color diagrams ,
the new pure hydrogen models follow the observed sequences of cool white dwarfs very well , while an improved physical description of dense helium moves the cool end of the helium sequence to the red , near the locus of black body colors .
this suggest that detailed fitting of the sed of the coolest dc white dwarfs will result in a number of them , and perhaps most , having hydrogen - rich composition rather than helium - rich .
an example is provided by our fit to wd@xmath64 .
this has direct implications for the spectral evolution of very cool white dwarfs and for the physical mechanisms responsible for their atmospheric composition .
allard , n. f. & kielkopf , j. f. 1982 , rev .
phys . , 54 , 1103 allard , n. f. , drira , i. , gerbaldi , m. , kielkopf , j. & spielfiedel , 1998 , , 335 , 1124 allard , n. f. , kielkopf , j. f. , & loeillet , b. 2004 , , 424 , 347 allard , n. f. , koester , d. , feautrier , n. , & spielfiedel , a. 1994 , supp .
, 108 , 417 bergeron , p. 2001
, , 558 , 369 bergeron , p. , ruiz , m. t. , hamuy , m. , leggett , s. k. , currie , m. j. , lajoie , c. p. & dufour ,
, , 625 , 838 bergeron , p. , & leggett , s. k. 2002 , , 580 , 1070 bergeron , p. , leggett , s. k. , & ruiz , m. t. 2001 , , 133 , 433 bergeron , p. , ruiz , m. t. , & leggett , s. k. 1997 , , 108 , 339 bergeron , p. , saumon , d. & wesemael , f. 1995 , , 443 , 764 bergeron , p. , wesemael , f. & beauchamp , p. 1995
, , 107 , 1047 boothroyd , a. i. , keogh , w. j. , martin , p. g. , peterson , m. r. 1991 , j. chem .
phys . , 95
, 4343 boothroyd , a. i. , keogh , w. j. , martin , p. g. , peterson , m. r. 1996 , j. chem .
phys . , 104 , 7139 davydov , a. s. 1965 , quantum mechanics ( oxford : addison - wesley ) , chap .
123 iglesias , c. a. , rogers , f. j. & saumon , d. 2002 , , 569 , l111 .
harris , h. c. et al .
2006 , , 131 , 571 kilic , m. , munn , j. a. , harris , h. c. , liebert , j. , von hippel , t. , williams , k. a. , metcalfe , t. s. , winget , d. e. , levine , s. e. 2006 , , 131 , 582 koester , d. & wolff , b. 2000 , , 357 , 587 kolos , w. , & wolniewicz , l. 1965 , j. chem .
43 , 2429 kowalski , p. m. 2006a , , 641 , 488 kowalski , p. m. 2006b , , in press kowalski , p. m. , mazevet , s. , & saumon , d. 2006 , in proceedings of the 14th aps topical conference on shock compression of condensed matter 2005 , m. d. furnish , m. l. elert , t. p. russell and c. t. white .
, in press ( 2006 ) kowalski , p. m. , saumon , d. , & mazevet , s. 2005 , in 14th european workshop on white dwarfs , d. koester & s. moehler , eds .
, asp conference series , vol .
334 , 203 kowalski , p. m. , & saumon , d. 2004 , , 607 , 970 mccook , g. p. , & sion , e. m. 1999 , , 121 , 1 martynov , g. a. 1992 , fundamental theory of liquids ( bristol : adam hilger ) , chap . 5 oppenheimer , b. r. , saumon , d. , hodgkin , s. t. , jameson , r. f. , hambly , n. c. , i chabrier , g. , filippenko , a. v. & brown , m. e. 2001 , , 550 , 448 peng , z. , kristyan , s. & kuppermann , a. 1995 , phys .
rev . a , 52 , 1005 sando , k. , doyle , r. o. , & dalgarno , a. 1969 , , 157 , l143 staszewska , g. & wolniewicz , l. 1999 , j. mol .
spect . , 198 , 416 wolff , b. , koester , d. & liebert , j. 2002 , , 385 , 995 | we investigate the importance of the far red wing of the lyman @xmath0 line of hydrogen in the atmospheres of cool white dwarfs of pure hydrogen composition . we find that this absorption process dominates all important sources of opacity in the blue part of the optical spectrum of these stars .
our successful fits to the spectra of cool da / dc white dwarfs indicate that the far red wing of the @xmath1 line is the source of opacity that had been missing in the models .
the observed sequence of cool white dwarfs in color - color diagrams is very well reproduced by our new pure hydrogen atmosphere models , suggesting that the atmospheric composition of the coolest dc white dwarfs must be revisited . |
nanoscale devices can now be fabricated with channels where the smallest dimension is just a few nanometers @xcite , and the development of nanofluidic theory @xcite is more relevant than ever .
consider the following example .
et al . _
@xcite used a series of rectangular nanochannels with widths ranging from 14 to 300 nm to connect two micro - scale chambers . by means of capillary filling , fluid from one chamber
fills up the channels and thus connects the two chambers .
the filling rate can be measured for different channel widths and for both milli - q water ( filtrated de - ionized water ) and an electrolyte solution of sodium chloride .
the rate did not follow the washburn equation for channel widths smaller than 100 nm .
the washburn equation is based on the classical continuum picture @xcite using poiseuille law of fluid motion which includes the newtonian ( or macroscopic ) shear viscosity . for widths larger than 100
nm the washburn equation correctly predicts the filling rate .
this is in accordance with the common understanding that the discrete nature of the fluid at small scales destroys the continuum picture @xcite .
in fact , many researchers categorize continuum physics as physics on the macroscopic scale , see for example ref . .
several questions immediately arise : when exactly does the continuum picture fail ?
how is this breakdown manifested ?
does the length scale of the breakdown depend on the specific problem ?
can one improve the continuum description such that it applies on small scales ?
demanding sufficient smoothness of the macroscopic quantities with respect to time and position and using a simple statistical argument , lautrup @xcite estimates that the smallest volume accessible to the continuum description must contain at least 10@xmath0 molecules .
this corresponds to a length scale of 8 - 80 nm , depending on the density . for steady
flows the temporal fluctuations can be averaged out and the accessible volume is much smaller .
this is illustrated in fig .
[ fig : motivation ] , where we have performed an atomistic simulation ( data given by blue filled circles ) of a methane fluid confined between two graphene sheets undergoing a poiseuille flow .
the slit - pore has a width of approximately 3.3 nm and the flow is driven by an external force field . in this paper the simulations
are carried out using the ` seplib ` library @xcite .
the classical continuum prediction is plotted as two red lines illustrating the maximum and minimum profiles allowed within statistical uncertainty on the newtonian shear viscosity @xcite .
only the fluid slip velocity at the wall surface is used as a fitting parameter . for this system
the continuum theory gives a satisfactory description of the fluid average velocity a length scale of a few nanometers .
apparently , even on these small length scales the molecular structure and degrees of freedom can be coarsened into simple transport coefficients like the viscosity . for water undergoing a steady flow it has been shown by atomistic simulations that the continuum description holds for channel widths of just 6 - 10 nm @xcite .
these results contrast earlier assumptions about the validity of the continuum picture , and the statement that continuum physics is physics on the macroscopic scale @xcite .
interestingly , it was later argued by thamdrup _
et al . _
@xcite that the disagreement between the experiment by persson _
et al . _
@xcite and the washburn prediction is due to pinned micro - bubles resulting in an increase in hydraulic resistance . at some point
the classical continuum description will of course break down . to mention two examples , travis _ et al . _
@xcite showed that for atomic fluidic systems the velocity profile features modulations for confinements in the order of 5 atomic diameters .
decheverry and bocquet @xcite analyzed the effect of thermal fluctuations on mass transport of fluid through a nanotube .
interestingly , when the classical continuum theory fails , the dynamics is frequently quantified by different transport coefficients compared to those of the bulk system and effective transport coefficients are introduced into the continuum constitutive relations @xcite .
the main point of this paper is that the observation of a breakdown need not be a failure of the continuum picture itself , but a result of inadequate modeling wherein important dynamical processes are not accounted for by classical theories . a very well understood example is the effect of the debye screening layer in electrolyte micro - flows @xcite .
two other physical mechanisms that become important on the nanoscale are often ignored in the literature , and this paper will treat these in detail : \(i ) in classical hydrodynamics the fluid s local rotation is determined uniquely by the fluid streaming velocity .
one can quantify the rotation from the local angular velocity field which is one half the vorticity , that is , one half the curl of the streaming velocity itself @xcite .
however , if the couple force , that is , the force component producing pure rotation , is large , the rotation must be treated as an independent variable .
the extended description is known as cosserat ( or micropolar ) continuum mechanics @xcite , first formulated by the cosserat brothers @xcite in the late 19th century .
cosserat continuum theory is used in various areas such as liquid crystal studies @xcite and blood flows @xcite , and was studied intensively in the 1950s to 1970s , see refs . .
for some reason it is not , however adopted by the nanofluidic community . we show that cosserat theory must also be used for fluid flows in extremely small confinements where the molecular structure becomes important .
\(ii ) classical hydrodynamics is based on local constitutive models relating fluxes to thermodynamic forces .
for a shear flow the stress at some point depends on the strain - rate at that particular point .
if the stress depends linearly on the strain rate , this leads to the newtonian law of viscosity @xcite .
a more general constitutive relation is to let the stress be a function in the entire strain rate history and spatial distribution , _
i.e. _ , given by a spatial and temporal convolution integral of a viscosity kernel and the strain rate @xcite .
this is the approach of generalized linear response theory @xcite .
the viscosity kernel accounts for the characteristic length scale of the spatial correlations @xcite ; we show below that this must be taken into account in order to arrive at the correct fluid response on molecular length scales .
our presentation is based on comparisons of continuum predictions with atomistic molecular dynamics ( md ) simulation data .
these two descriptions are fundamentally different in two ways .
first , in md the system is characterized by discrete particles where the path of each individual particle constituting the fluid is traced out through classical mechanics @xcite , _
i.e. _ , the particle interactions must be known .
the discretization of matter is , of course , in strong contrast to the fundamental assumptions of continuum mechanics .
secondly , the continuum description applies constitutive relations to form mathematical closed problems .
no such models are enforced in the standard md simulations .
any discrepancy between md and the continuum description may therefore be a result of a breakdown of the constitutive relation rather than a break down of the continuum theory as such .
our basic conjecture is that md acts as an idealized numerical experiment , and if a given continuum theory agrees with the md data , the theory correctly accounts for the phenomena we study .
let us specify , more accurately , what is meant by continuum theory .
basically , one refers to deformable fluid volumes characterized by quantities which are continuous at any point @xmath1 over the entire volume and at any time @xmath2 @xcite .
this means that these quantities are described mathematically by field variables .
the basic continuum hypothesis is that one can associate a given fluid sub - volume ( or `` fluid particle '' ) with the same characteristic quantities of the entire deformable fluid volume , no matter how small the sub - volume @xcite .
lautrup @xcite suggests a lower limit of the order of 10@xmath0 molecules as stated above , but time averaging allows an arbitrarily small fluid particle volume as seen in fig .
[ fig : motivation ] .
one field variable is the streaming velocity , which is the mass - weighted average velocity of the individual molecules in the fluid particle around a given point @xcite .
the fluid s dynamics is governed by balance ( or conservation ) equations . in general
the balance equation for some quantity per unit mass , @xmath3 , reads in the eulerian differential form @xcite @xmath4 where @xmath5 is the mass density , @xmath6 a production term , @xmath7 the streaming velocity , and @xmath8 the flux of @xmath9 . here
@xmath9 can be a scalar or vector quantity . the right hand side of eq .
( [ eq : generalbalance ] ) is the sum of the body force and the surface force densities , that is , forces per unit volume . when @xmath9 represent the velocity field , @xmath10 , eq .
( [ eq : generalbalance ] ) is the momentum balance equation .
the body force density can be a gravitational - like force driving the flow as in fig .
[ fig : motivation ] , and the surface force density is the pressure tensor , @xmath11 @xcite .
a special case is the mass balance equation for which @xmath12 . since rotation is treated as an independent variable , a balance equation on the form of eq .
( [ eq : generalbalance ] ) must be formulated for rotation ; this is done in the _ supporting information_. ( si ) importantly , in the extended cosserat description the pressure tensor @xmath13 need not be symmetric @xcite as in the classical continuum theory .
comparison between the continuum description and md simulation data is carried out for molecular fluidic systems at equilibrium , as well as for steady flows in a slit - pore . here
we investigate four molecular fluids : a methane fluid , a generic di - atomic ( dumbbell ) fluid , liquid butane , and liquid water .
for methane 75 percent of the mass is centered in the carbon nucleus and methane is here considered as a simple spherical point - mass molecule , as it was done in fig .
[ fig : motivation ] .
water will , on the other hand , be treated differently using the flexible spc / fw water model @xcite that accounts for the molecular structure and hydrogen bonds and thus for the structure of liquid water .
the butane model is a coarse grained model where the methyl and methylene groups are represented by a united atomic unit , _
i.e. _ , a spherical point - mass . details about the butane model can be found in ref .
, however , here flexible bond are implemented with parameters from the generalized amber force field @xcite .
the simulations are done using the ` seplib ` library @xcite .
nanofluidic flows are often associated with fluid slippage at the wall boundary @xcite . just like the effect of the fluid - fluid interactions on the flow is lumped into a single parameter , _
e.g. _ , viscosity , one effect from the fluid - solid interaction can be modelled into a friction coefficient determining the boundary slip .
the slippage has a large effect on the flow rate in extreme confinement and is usually quantified by the slip length @xmath14 . for a hagen - poiseuille flow in a tube with radius @xmath15 the relative flow enhancement @xmath16 due to the slip is given as @xcite @xmath17 @xmath14 is typically in the order of a few nanometers .
thus , for a given non - zero slip length the flow enhancement increases hyperbolically as the tube radius decreases .
the slip is always present , but has insignificant effect on the flow rate for tube radii above microns .
@xmath14 is normally independent of system size , that is , is not an intrinsic nanofluidic phenomenon and is therefore not addressed in this paper .
slippage is here modelled in an _ ad - hoc _ fashion as it was done in fig .
[ fig : motivation ] . in si
we derive the cosserat extended continuum theory from the microscopic point of view using a microscopic hydrodynamic operator .
the derivation , which is based on the fundamental definition of the macroscopic field variables in terms of the corresponding molecular quantities , follows the idea of irving and kirkwood @xcite , and evans and morriss @xcite , see also ref . .
the derivation leads to a molecular interpretation of the fluxes entering eq .
( [ eq : generalbalance ] ) .
the final dynamical system of equations is sometimes referred to as the extended navier - stokes ( ens ) equations [ eq : sect1:ens ] @xmath18 where , @xmath19 is the material operator , @xmath20 the spin angular velocity field , and @xmath21 is the pressure gradient .
the transport coefficients @xmath22 , @xmath23 , and @xmath24 are the bulk , shear , and rotational viscosities , respectively , and @xmath25 , @xmath26 and @xmath27 the corresponding spin viscosities . finally , @xmath28 and @xmath29 represent production terms of linear and spin angular momentum , respectively .
we refer the reader to si for a derivation and discussion of eq .
( [ eq : sect1:ens ] ) .
the theory is only strictly true for isotropic systems , and we study such cases in sections [ sec : coupling ] and [ sect : nonlocal ] , comparing theoretical predictions with md simulation data .
flows in extreme confinement are characterized by strong density inhomogeneities and anisotropy .
we study such flows in section [ sect : flows ] , again comparing theory with md data .
finally , section [ sect : summary ] gives a brief summary .
the purpose of this section is to demonstrate the validity of the ens equations , eqs .
( [ eq : sect1:ens ] ) , by comparing the predictions of different thermally induced relaxation phenomena with md simulation data , see also refs . .
we start with this problem instead of the situation with confining walls as the latter is introduces density inhomogeneities and molecular alignment at the wall - fluid interface .
we return to this more complex situation in sect . [
sect : flows ] . rather than investigating the quantities directly ,
one typically studies the associated correlations @xcite . here
we will use the approach based on onsager s regression hypothesis @xcite , which states that thermal perturbations on average decay according to the deterministic hydrodynamic equations of motion .
specifically , we will compare mechanical spectra obtained from md simulations with predictions from theory . in equilibrium a fluctuating quantity @xmath30 can be written as @xmath31 , where @xmath32 is the average part and @xmath33 is the fluctuating part . in equilibrium
the average streaming velocity and spin angular velocity are both zero so @xmath34 and @xmath35 .
the fluctuations are modelled using the stochastic forcing approach @xcite . here
an uncorrelated zero mean stochastic force is added to the constitutive relations , see si . for example , for the antisymmetric pressure the constitutive relation with stochastic forcing reads @xmath36 , where @xmath37 is the fluctuating part of the flux . to a first order approximation in the fluctuation we have on the left - hand side of eq .
( [ eq : sect1:ens ] ) @xmath38 and @xmath39 in fourier space the stochastic ens equations read to first order in the fluctuations for wave vector @xmath40 @xmath41 due to the properties of the divergence operator .
see si eq .
( 2 ) for the definition of the fourier transform .
it is here convenient to introduce the following coefficients @xmath42 where subscripts @xmath2 indicates `` transverse '' and @xmath43 indicates `` longitudinal '' .
it has been shown @xcite that @xmath44 , and we write both coefficients @xmath45 .
we may then define the susceptibility @xmath46 where @xmath47 .
we will also drop the subscript @xmath48 from here on .
we take @xmath49 and write out the @xmath50-component of the velocity and @xmath51-component of the angular velocity [ eq : sect2:ens_u_k ] @xmath52 these two components are both transverse components to the wave vector and are coupled .
we also investigate the longitudinal angular velocity component @xmath53 which is given through @xmath54 note , that this longitudinal component is unaffected by the coupling between the linear and spin angular momenta .
we define the following three correlation functions [ eq : coupled - cf ] @xmath55 which we denote the transverse velocity auto - correlation function ( tvacf ) , the transverse cross - correlation function ( tccf ) , and longitudinal angular velocity auto - correlation function ( lavacf ) , respectively . by assumption
the fluctuating fluxes are uncorrelated with the velocity and angular velocity , _
e.g. _ , @xmath56 .
thus , multiplying eqs .
( [ ens_du_k.1 ] ) and eq .
( [ ens_dw_k.1 ] ) with @xmath57 and ensemble averaging we arrive at the differential equation system for the tvacf and the tccf [ eq : sect2:cdff ] @xmath58 similarly , multiplying eq .
( [ ens_dw_k.2 ] ) with @xmath59 one has for the lavacf @xmath60 upon ensemble averaging .
now , eqs .
( [ eq : sect2:cdff ] ) and ( [ eq : delavacf ] ) can be solved yielding to second order in wave vector [ eq : corrfun ] @xmath61 where the characteristic frequencies are @xmath62 the pre - factors in eqs .
( [ eq : corrfun_1 ] ) and ( [ eq : corrfun_2 ] ) are calculated from the first order approximation in the fluctuations @xmath63 as above and therefore @xmath64 from the definition of the linear momentum density , see si eq .
( 14 ) in si , and for one component systems with molecular mass @xmath65 .
likewise , to a first order approximation in density and moment of inertia @xmath66 , and from si eq .
( 29 ) @xmath67 as @xmath68 @xcite .
applying the equipartition theorem one arrives at the pre - factors .
equations ( [ eq : sigh1 ] ) and ( [ eq : sigh2 ] ) also provide a first order method to calculate the correlation functions in the md simulations ; this method is used here .
it is informative to work in the frequency domain , _
i.e. _ , to predict the peak frequencies in the corresponding spectra .
applying the fourier - laplace transform defined by @xmath69({\mathbf}{k } ,
\omega ) = \int_0^\infty f({\mathbf}{k},t ) \ , e^{-i\omega t } \ , \mathrm{d}t\ ] ] we get [ eq : sect2:corrfun ] @xmath70 from eq .
( [ eq : sect2:corrfun_3 ] ) we can make a very important conclusion , namely , @xmath71 this means that the coupling can be ignored on long length scales .
this is also expected as the classical navier - stokes theory holds for macroscopic systems and no coupling effect is observed . the relaxation of spin is still governed by the rotational viscosity , but this relaxation does not affect the relaxation of linear momentum controlled by the usual viscous dissipation processes . if we define @xmath72 as @xmath73 the lavacf and tvacf are , in the limit of zero wave vector , @xmath74 furthermore , for the fluids studied here the effect of the coupling on the tvacf is not large , that is , @xmath75 even for wave vectors in the sub - molecular diameter range , and the limit in eq .
( [ eq : sect2:limitcase ] ) need not to be taken as a strict limit .
it is also worth noticing that the rotational viscosity @xmath24 is a linear function of the moment of inertia @xmath76 for sufficiently large @xmath76 @xcite , so @xmath77 is independent of @xmath76 here .
we first compare the predictions from the continuum ens theory with md simulation data for the simple di - atomic molecule ( the dumbbell model ) in the super - critical fluid regime .
the transport coefficients , @xmath78 , and @xmath24 are listed in table i in si .
figure [ fig : spectradumbbell ] shows md data ( symbols connected with lines ) for imaginary parts of the spectra of the tvacf and lavacf ; normalization is carried out for clarity .
the prediction from the continuum theory is plotted as full blue lines .
+ + it is observed that for small wave vectors @xmath79 the continuum prediction is in excellent agreement with the md data , but it fails for larger wave vectors @xmath80 .
we emphasize that no fitting is performed , and all relevant parameters are taken from si table i found from independent simulations and methods . using typical values for the md units
@xmath81 and @xmath65 the results show that the continuum theory predicts the mechanical spectrum for wave lengths in order of 2 - 3 nm and above , and time scales in order of of 1 - 10 ps and above . for the tvacf , fig .
[ fig : spectradumbbell ] ( b ) , the result can be understood from the fluid stress relaxation at zero wave vector as suggested by bocquet and charlaix @xcite . from the last equation in eq .
( [ eq : eigen ] ) we can define a wave vector dependent relaxation time @xmath82 .
this relaxation time must be larger than the characteristic relaxation time @xmath83 at zero wave vector for the predictions to hold @xmath84 , _ i.e. _ , for the viscosity to be wave vector independent .
this means that @xmath85 we will denote this the bocquet - charlaix criterion .
estimates for the relaxation time @xmath83 is given through the shear pressure ( or equivalently stress ) auto - correlator @xmath86 where @xmath87 is the @xmath88 component of the symmetric part of the pressure tensor .
for the dumbbell model @xmath89 is fully decayed at @xmath90 which gives @xmath91 .
this is in perfect agreement with the results depicted in fig .
[ fig : spectradumbbell ] ( b ) .
alternatively , the relaxation time can be given through the maxwell relaxation time @xmath92 or the viscous relaxation time @xcite @xmath93 , where @xmath94 is the infinite shear modulus and @xmath95 is the first normal stress coefficient .
for the diatomic model studied here @xmath96 giving @xmath97 which is not what is observed .
therefore , the characteristic decay time that should be used for the bocquet - charlaix criterion is the time for the autocorrelation function @xmath89 to fully decay . in the small wave vector regime
the relaxation of spin angular momentum is dominated by the coupling mechanism between linear and angular momenta as the peak is located at @xmath98 .
the relaxation of linear momentum , fig .
[ fig : spectradumbbell ] ( b ) , is on the other hand due to usual viscous mechanisms seen by the peak frequency @xmath99 .
for large @xmath100 the continuum theory overestimates , by an order of magnitude , the peak frequency for the lavacf and tvacf due to over - estimation of the effect of the spin diffusion .
figure [ fig : spectradumbbell ] ( c ) depicts the tccf for the dumbbell model .
again , the theory performs surprisingly well for sufficiently small wave vectors ( @xmath101 ) , but fails for larger .
it is worth noticing that the amplitude of the tccf is a non - monotonic function with respect to wave vector , having a maximum around @xmath102 .
this behavior is also captured by the ens theory . to illustrate that the amplitude is a decreasing function of wave vector the tccf for @xmath103 and @xmath104
is plotted as predicted by the theory .
recall , in the limit @xmath105 the coupling vanishes .
next we apply the theory to liquid butane .
as discussed above the butane model is not uni - axial or rigid , however , from the principal moment of inertia we argued that the theory should be a good approximation . the result is shown in fig .
[ fig : spectrabutane ] . for the lavacf one observes a peak frequency at around 9 thz and
the relaxation process is extremely fast .
this fast mode is not precisely captured by the theory with @xmath106 thz .
interestingly , the peak frequency is almost independent of wave vector for the range studied here .
this indicates that for these fast modes the diffusion of spin is less important for the relaxation processes .
for the slower relaxation of linear momentum , we see that the peak frequency is predicted very well by the theory , the md result is @xmath107=1.0 thz and the predicted one is around @xmath108 thz for the lowest wave vector . again , for larger wave vectors the prediction fails as expected .
the classical linear constitutive relations are local in the sense that the flux only depends on the local and instantaneous thermodynamic force .
this is in general not the case , rather the response depends on the entire force distribution in the system , as well as its history .
one can model this phenomenologically by introducing frequency and wave vector dependent transport coefficients @xcite .
this renders the continuum description valid on arbitrary small length and time scales .
the generalized transport coefficients are referred to as kernels . in the homogeneous isotropic case , assuming space and time invariance , the linear non - local constitutive relation for the symmetric part of the pressure tensor reads @xcite @xmath109 @xmath110 is the symmetric part of the velocity gradient , _
i.e. _ , the strain rate .
fourier transforming with respect to space and fourier - laplace transforming with respect to time yields @xmath111 from the convolution theorem .
the shear viscosity kernel @xmath112 can be found from the tvacf as it is now shown . here
we focus on molecules with small moment of inertia and small wave vector regime , _
i.e. _ , small @xmath113 . in this limit
( [ eq : sect2:corrfun_2 ] ) can be rearranged giving @xmath114 in particular , we have at zero frequency @xmath115 this approximation holds even for large values of @xmath116 as discussed above , eq .
( [ eq : unequiv ] ) . for molecules that can be regarded as
point masses , say methane , the moment of inertia is zero and eq .
( [ eq : sect4:visckernel ] ) is exact . in fig . [
fig : kernel ] ( a ) the viscosity kernel at zero frequency is plotted for the methane , dumbbell , butane , and water systems using eq .
( [ eq : approxnonlocal ] ) .
one immediately notices that for @xmath117 the wave vector dependent viscosity approaches the zero wave vector limit .
this is in good agreement with the bocquet - charlaix criterion , eq .
( [ eq : ineqtvacf ] ) .
interestingly , this is independent of the specific fluid studied here and the local constitutive relations can be applied on length scales down to approximately @xmath118 2 - 2.5 nm . is this a general result that applies to all fluidic systems ?
the answer is no ! in fig .
[ fig : kernel ] ( b ) the zero - frequency viscosity kernel is plotted for the asymmetric dumbbell model for different temperatures .
the asymmetry arises due to the mass and lennard - jones parameter differences between the two constituent atoms .
the asymmetric dumbbell model allows one to probe the dynamics in the highly viscous regime without crystallization occurring @xcite .
the result shows that for relatively high temperatures the kernel has the same wave vector dependency , but approaching the viscous regime ( lower temperature ) the kernel only reaches the newtonian viscosity at longer length scales .
this indicates that the dynamical processes behind the viscous response take place on longer length scales in accordance with the cooperative motion in super - cooled liquids @xcite .
the non - local viscous response has also been studied for highly viscous two component lennard - jones system and polymer melts , see refs . .
the failure of the local constitutive relation , that is , of newton s law of viscosity , is very clearly illustrated by todd _
et al . _
@xcite for a point - mass weeks - chandler - andersen ( wca ) system @xcite . in real space the non - local description amounts to a convolution of the viscosity kernel and the strain rate distribution , eq .
( [ eq : nonlocalconst ] ) . in the homogeneous situation where the fluid undergoes a steady shear in the @xmath50-direction with varying amplitude in the @xmath51-direction we have one non - zero shear component in the pressure tensor , namely , the @xmath119 component . in this steady situation
( [ eq : nonlocalconst ] ) reduces to @xmath120 where @xmath121 .
if the shear is induced by an external force field @xmath122 , the fluid flow is @xmath123 , where @xmath124 is the excited fourier mode of the velocity field .
we assume that this is the only mode excited , _ i.e. _ , the force amplitude must be sufficiently low @xcite . also , this ensures a linear response as well as constant temperature and density .
the strain rate is then @xmath125 for simplicity we shall assume that the kernel is given by a gaussian function @xmath126 such that @xmath127 gives a characteristic decay length .
the kernel must fulfill @xcite ( i ) @xmath128 , and ( ii ) @xmath129 is an even function . substituting eqs .
( [ eq : straintodd ] ) and ( [ eq : kernelgaussian ] ) into eq .
( [ eq : nonlocalz ] ) we have upon integration @xmath130 if @xmath131 the model is local , corresponding to newton s law of viscosity , that is , for the local model @xmath132 the system can be simulated using the sinusoidal transverse force ( stf ) method @xcite , and it is possible to evaluate @xmath133 for different external force fields and wave vectors . the two different predictions can be compared to the actual shear pressure @xmath134 , which is found directly from the momentum balance equation that for the steady flow reads @xmath135 integrating we obtain @xmath136 the comparison is made in fig .
[ fig : kernelstresses ] . clearly , the local prediction fails for the larger wave vector , fig .
[ fig : kernelstresses ] ( b ) , whereas the non - local prediction agrees with the actual shear pressure . from the non - local model
we conclude that spatial correlations result in a reduced shear pressure . from eqs .
( [ eq : stressnonlocal ] ) and ( [ eq : stresslocal ] ) we can quantitatively evaluate the effect of spatial correlations on the stress .
specifically , we have the relative difference given by @xmath137 for the wca system studied here @xmath138 and the bocquet - chairlaix criterion gives @xmath139 , this corresponds to an error in the stress below 32 % according to eq .
( [ eq : aargh1 ] ) .
the gaussian function does not perfectly fit to the kernel data .
nevertheless , this simple functional form captures the non - local response well , due to the smoothing of the convolution .
other more complicated forms have been suggested , see refs . .
todd and hansen @xcite showed that the non - local response is only relevant for flows where the strain - rate is non - linear with respect to position .
couette and poiseuille flows are then not affected by non - locality . to illustrate this consider any functional form for the kernel which fulfills the criteria given above : its integral gives the zero wave vector viscosity and it is an even function with respect to @xmath51 .
first , making the change of variables @xmath140 eq .
( [ eq : nonlocalz ] ) reads @xmath141 assuming a strain rate on the form @xmath142 , we have @xmath143 as the integrand in the second integral is an odd function .
this result is the same as the local prediction . in general ,
if a taylor expansion of the strain rate @xmath144 exists , using the properties of odd and even functions one can verify that the non - zero non - local effects of the strain rate can be determined by the even moments of the kernel @xcite @xmath145 in the case of couette and poiseuille flows the taylor expansion terminates at zeroth and first order , respectively , and there are no non - local effects .
the spin and rotational viscosity kernels can be found by simply rearranging eq .
( [ eq : sect2:corrfun_1 ] ) giving the generalized susceptibility @xmath146 our group recently @xcite conjectured that the rotational viscosity @xmath24 governs the fast wave vector independent relaxation processes as indicated in figs .
[ fig : spectradumbbell ] ( a ) and [ fig : spectrabutane ] ( a ) .
this transport coefficient is therefore only frequency dependent .
we then have @xmath147 and therefore @xmath148 we called this the generalized extended navier - stokes ( gens ) theory . from md simulations
one can calculate the lavacf ( as shown above ) and from there find the kernels . for dense fluids @xmath149
is characterized by a sharp peak around zero wave vector @xcite since the diffusive contribution to the relaxation of the lavacf is very small for @xmath150 , see figs . [
fig : spectradumbbell ] ( a ) and [ fig : spectrabutane ] ( a ) .
the spin viscosity kernel has the same properties as the shear viscosity kernel and for this reason we do not expect any non - local effects for flows where the gradient of the angular velocity is constant or linear .
we first study a poiseuille flow , the geometry is shown in fig .
[ fig : motivation ] . in experiments this flow can be achieved by application of a constant pressure gradient . generating a pressure difference in simulations with , for example
, a piston and using molecular reservoirs can cause density variations in the direction of the flow and other inlet / outlet effects .
we therefore use a constant force field acting on each point mass in the fluid to drive the flow .
the wall particles are arranged on a simple cubic lattice and are allowed to vibrate around their initial lattice site using a simple restoring spring force . the viscous heating generated in the fluid
is removed by thermostating the wall particles .
this method resembles the real physical experiment and is therefore often referred to as direct non - equilibrium molecular dynamics .
the interested reader is referred to ref .
@xcite for further details . to get a satisfactory signal - to - noise ratio in the md simulations unrealistically large external forces
are typically applied to drive the system , and the resulting flow rates are very large , typically , in the order of 10 - 100 m s@xmath151 . despite these large flow rates
the reynolds number is usually less than unity due to the extremely small characteristic length scales involved .
finally , it is very important to ensure that the simulations are carried out in the linear regime which is discussed below . in the linear regime of low reynolds number and for the geometry shown in fig .
[ fig : motivation ] the ens equations form a two - point boundary value problem in the steady state @xmath152 where @xmath153 .
recall that @xmath154 and @xmath155 . introducing @xmath156 and applying no - slip boundary conditions , @xmath157 and @xmath158 ,
eringen @xcite solved this yielding @xmath159 with the following definitions of @xmath160 and @xmath161 @xmath162 the application of the no - slip boundary condition is not justified .
a correct treatment applies the neumann boundary condition for both the velocity and angular velocity fields , however , this is not straightforward in that the two are likely coupled and therefore dependent on each other . while the boundary condition for the velocity field has been studied in great detail ,
see _ e.g. _ refs .
@xcite , very little is known about the spin boundary condition .
just recently de luca _ et al . _
@xcite showed that the spin field does possess slippage and badur _ et al . _
@xcite used spin slip to account for flow enhancement .
as mentioned in the introduction , we will treat the problem in an _ ad hoc _ fashion and simply set the angular velocity slip in accordance with the md data . if one ignores the coupling , @xmath163 , the solution for the streaming velocity , eq .
( [ eq : sect3:solu ] ) , reduces to the classical poiseuille flow solution @xmath164 in this classical situation the angular velocity is found from the vorticity , @xmath165 , that is , @xmath166 in agreement with eq .
( [ eq : sect3:solw ] ) for @xmath167 . as the classical treatment
does not allow for specification of spin boundary condition , eqs .
( [ eq : sect3:solw ] ) and ( [ eq : sect3:solwclassical ] ) differ by a magnitude of @xmath168 at the walls . from eq .
( [ eq : sect3:solu ] ) one can see that the maximum velocity , located at @xmath169 , is lowered as a result of the coupling since from the last term we have @xmath170 and thus @xmath171 .
another way to quantify this effect is to evaluate the volumetric flow rate @xmath172 @xcite @xmath173 where @xmath174 is the characteristic half length in the @xmath175-direction .
this gives the relative volumetric flow rate reduction @xmath176 equation ( [ eq : sect3:flowreduction ] ) is plotted in fig .
[ fig : couplingeffect ] for the dumbbell fluid , liquid butane , and water .
the relevant coefficients can be found in si table i. it can be seen that for water flowing in a channel with a width of 9 nm the flow rate is reduced by about 10 % due to the coupling . as the channel width increases the flow rate approaches that of the classical predictions and the effect of the coupling can be ignored . from eq .
( [ eq : sect3:flowreduction ] ) the relative flow rate reduction increases as the product @xmath177 decreases . from this observation , one can define a characteristic fluid length scale @xmath178 @xcite below which the effect of the coupling becomes significant . to this end
we write the parameter @xmath161 as @xmath179 from si table i it is seen that @xmath180 and @xmath181 .
thus a significant flow - rate reduction occurs for fluids with large critical length scale @xmath178 . for water @xmath182 nm and for butane @xmath183 nm in agreement with the relative large flow rate reduction observed for water .
first we compare md data with continuum predictions for the dumbbell system . before the comparison ,
however , the linear newtonian response regime should be identified , at least , for the bulk fluid region . to this end
one can apply the synthetic sllod algorithm developed by evans and morris @xcite .
basically the method imposes a constant strain rate ( linear velocity profile ) on the system while ensuring a homogeneous density and iso - kinetic temperature . to achieve this the equations of motion
are reformulated according to the gaussian principle of least constraint , see also refs . .
performing a series of sllod simulations it is found that the newtonian regime occurs in the range @xmath184 for the dumbbell fluid .
the upper limit for the external force field can then be approximated by rewriting eq .
( [ eq : sect3:solwclassical ] ) to @xmath185 giving @xmath186 , with @xmath187 . note , in the wall - fluid region the velocity may feature rapid changes and here the linearity is not guaranteed .
based on the sllod approach delhommelle @xcite developed a synthetic method to calculate the rotational viscosity @xmath24 as a function of spin angular velocity .
see also edberg _ et al . _
@xcite . to our knowledge
no synthetic or controlled method exists to study the spin viscosity dependency of the gradient of the spin , or the rotational viscosity dependency on strain rate .
we will therefore here assume that the linear regime is identical to the newtonian regime , _
i.e. _ , in the regime where the viscosity is independent of the strain rate .
the classical description predicts that the poiseuille flow is local flow according to sec .
[ sect : nonlocal ] . also , from fig .
[ fig : couplingeffect ] we expect the flow - rate reduction due to the coupling to be very low for the dumbbell model .
thus , based on the theory we can expect the classical description to be a good approximation for this system .
the time - averaged velocity and spin angular velocity profiles are shown in fig .
[ fig : prof - dumbbell ] ( a ) for the dumbbell model where the pore width is approximately 14.8 atomic diameters .
the profiles are sampled after the system has reached the steady state .
the temperature profile ( not shown ) is constant and the temperature is @xmath188 throughout the channel .
( color online ) ( a ) velocity and angular velocity profiles for the dumbbell model undergoing a poiseuille flow .
symbols represent md data , and lines the classical predictions where slippage is included .
( b ) the corresponding shear pressures .
@xmath119 is calculated by integration of the momentum balance equation , using the density profile given in ( c ) .
the antisymmetric pressure is calculated from the constitutive relation si eq . (
40 ) using md data as input .
( c ) density and order parameter profiles .
the molecules in the lower left corner illustrate ( exaggerated ) the molecular ordering near the wall .
, title="fig : " ] + ( color online ) ( a ) velocity and angular velocity profiles for the dumbbell model undergoing a poiseuille flow .
symbols represent md data , and lines the classical predictions where slippage is included .
( b ) the corresponding shear pressures .
@xmath119 is calculated by integration of the momentum balance equation , using the density profile given in ( c ) .
the antisymmetric pressure is calculated from the constitutive relation si eq .
( 40 ) using md data as input .
( c ) density and order parameter profiles .
the molecules in the lower left corner illustrate ( exaggerated ) the molecular ordering near the wall .
, title="fig : " ] + ( color online ) ( a ) velocity and angular velocity profiles for the dumbbell model undergoing a poiseuille flow .
symbols represent md data , and lines the classical predictions where slippage is included .
( b ) the corresponding shear pressures .
@xmath119 is calculated by integration of the momentum balance equation , using the density profile given in ( c ) .
the antisymmetric pressure is calculated from the constitutive relation si eq .
( 40 ) using md data as input .
( c ) density and order parameter profiles .
the molecules in the lower left corner illustrate ( exaggerated ) the molecular ordering near the wall .
, title="fig : " ] the predictions from the classical navier - stokes theory , eqs .
( [ eq : pois ] ) and ( [ eq : sect3:solwclassical ] ) , are also plotted using the shear viscosity from si table i. velocity slippage at the wall - fluid interface is allowed , @xmath189 ; @xmath190 is then the only fitting parameter in the comparison .
the agreement between md data and classical continuum predictions is excellent , except at the wall - fluid interface .
this is highlighted by the shear pressures plotted in fig .
[ fig : prof - dumbbell ] ( b ) . according to the classical theory the shear pressure @xmath191 is linear , however , at the wall - fluid interface
this is not the case .
also , the classical theory assumes a zero anti - symmetric part of the shear stress @xmath192 .
this is clearly not fulfilled near the wall . to understand the disagreement at the boundary
, we analyze the fluid ordering .
it is well known that the wall induces a density variation in the fluid @xcite .
the density profile is shown in fig .
[ fig : prof - dumbbell ] ( c ) ( black dots ) .
it is seen that the density varies in a region approximately one atomic diameter away from the wall .
the transport properties are functions of density , and one should expect a variation in the viscosities here .
furthermore , one can evaluate the molecular alignment ordering through the parameter @xcite @xmath193 where @xmath194 is the angle between the molecular bond and @xmath88-plane .
for perfect parallel alignment @xmath195 , that is , the molecules closest to the wall are , on average , aligned with the wall . for distances around one atomic diameter
the molecules are slightly normal to the wall as @xmath196 .
the extremes are illustrated with the two molecules in the lower left corner in fig .
[ fig : prof - dumbbell](c ) . for zero order parameter ,
the molecules have random orientations which is the case in the interior of the channel .
this means that the system possesses a degree of anisotropy in the wall - fluid region . to fully account for the density variation and ordering one should therefore describe the transport properties through a position - dependent tensorial shear viscosity .
figure [ fig : prof - butane ] ( a ) shows the velocity and density profiles for a butane flow where the pore width is just 6 nm . for such extreme confinements the fluid layering stretches over the entire pore . the order parameter profile , fig .
[ fig : prof - butane ] ( b ) , shows that the molecular orientation is strongly anisotropic .
finally , the mean square molecular end - to - end distance @xmath197 also varies , showing that the butane molecule on average is elongated at the fluid - region by around 2 per cent .
such a complex system is not modeled appropriately by the classical or extended theories presented here .
this is not an indication of a breakdown of the continuum picture , but an incomplete modeling .
worth noting is that the fluid ordering and layering is constant over a large range of external forces including zero force , see fig .
[ fig : prof - butane ] , and is thus not flow induced .
( color online . ) ( a ) density and velocity profiles for butane in a slit - pore of width 6 nm .
the prediction from the theory breaks down and is not shown .
( b ) corresponding order parameter and square end - to - end distance profiles . ,
title="fig : " ] ( color online . )
( a ) density and velocity profiles for butane in a slit - pore of width 6 nm .
the prediction from the theory breaks down and is not shown .
( b ) corresponding order parameter and square end - to - end distance profiles .
, title="fig : " ] as pointed out by bitsanis _
et al . _
@xcite the velocity profile features surprisingly small modulations considering the density profile : one should expect the transport properties to vary significantly across the channel having large effects on the flow profile .
the authors suggested the local average density model ( ladm ) wherein the transport properties at a point @xmath51 is a function of the average density around that point . in the current geometry ,
where the density is constant in the plane parallel to the wall , the local average is @xmath198 where @xmath199 defines the region of averaging .
the agreement between the ladm and simulation data can be very good , especially if one introduces a non - uniform weighting function @xcite .
however , the ladm can not predict the shear pressure response in fig .
[ fig : kernelstresses ] as the density is constant . also , the ladm model is not capable of predicting the strain rate reversal observed by travis _
et al . _
@xcite to account for the observed velocity profile one can write the position - dependent ( in - homogeneous ) non - local constitutive model as @xmath200 the position dependency likely comes from the varying density at the wall - fluid region .
the application of this relation is not straightforward @xcite as it is unclear how the convolution should be performed at the wall where the support of the kernel goes beyond the boundary and is unknown @xcite .
recently , dalton _
et al._@xcite used a sinusoidal longitudinal force ( slf ) , also introduced by hoang and galliero @xcite , to control the density variation in a periodic system . due to the periodicity
, the boundary problem can be eliminated .
the density profile can be controlled to such an extent that it resembles that seen in confined systems .
the fluid can then be driven by an stf .
the authors showed that the non - local response is capable of predicting the strain rate reversal observed by travis _
et al . _
@xcite , as well as the relative small modulation on the velocity profile . a rigorous and general implementation of eq .
( [ eq : posnonlocal ] ) into the balance equations for confined systems is still lacking .
for the dumbbell and butane models the coupling between the linear and spin angular velocities has little effect on the flow . from fig .
[ fig : couplingeffect ] , however , the effect is significant for water flow in channel with widths below 5 nm . in fig .
[ fig : flow - water ] ( a ) md data for the velocity profile for water is plotted where the channel width is approximately 10 water diameters .
( color online . ) ( a ) velocity profiles of water undergoing a poiseuille flow .
( b ) relative profile curvature difference at @xmath201 . from ref . with modifications . , title="fig : " ] ( color online . ) ( a ) velocity profiles of water undergoing a poiseuille flow .
( b ) relative profile curvature difference at @xmath201 . from ref . with modifications . , title="fig : " ] also , shown are the predictions from the classical ns and ens theories . density and order parameter profiles ( not plotted here )
show little density variation and molecular alignment , except within 3 - 4 of the wall .
the slip velocities are estimated by fitting a second - order polynomial ( dashed lines ) to the velocity profile , excluding the wall - fluid region where the fluid is slightly anisotropic and inhomogeneous ; this then amounts to the apparent slip length @xcite and is the only fitting parameter used in the comparison .
it is seen that the classical prediction fails , while the ens theory captures the flow profile note , fit of the profile data to the classical description will result in a wrong viscosity no matter how many data points in the wall - fluid region are included . any shift of the profile will not change this either .
an extra source of dissipation must be present .
furthermore , note that a complete description involves a position dependent non - local anisotropic modelling of the wall - fluid region . to remove any effect in the boundary region one can evaluate the curvature in the channel midpoint , @xmath201 .
the predictions are simply found from the second - order derivative of eqs .
( [ eq : sect3:solu ] ) and ( [ eq : pois ] ) .
the relative difference is @xmath202 @xmath203 is plotted in fig .
[ fig : flow - water ] ( b ) together with the results from the md simulations . within statistical uncertainty
the ens theory and md simulation results agree .
as the channel width increases the relative curvature difference vanishes and the classical description is re - captured .
the particular model applied is parameterized with respect to the liquid state and the wall is a lennard - jones cubic lattice , see ref . .
the fluid structure near the wall and its effect on the dynamics will be affected by the different models , choice of model parameters , and wall details .
however , it is not the aim here to critically review the fluid structure at near the wall , but to see the effect of the coupling .
perhaps the most clear illustration of the translational - rotational coupling is seen by introducing an external torque into the system while having a zero production term for the linear momentum . in general , if the resulting torque density @xmath204 is sufficiently small then for the geometry in fig .
[ fig : motivation ] we have @xmath205 integrating eq .
( [ eq : torqueensu ] ) we get @xmath206 in terms of @xmath207 which is substituted into eq .
( [ eq : torqueensw ] ) resulting in a second order inhomogeneous differential equation for @xmath207 . from this and eq .
( [ eq : torqueensu ] ) and by application of dirichlet no - slip boundary conditions one has @xmath208 where @xmath209 and @xmath210 are integration constants .
one can show that @xmath210 goes rapidly to zero as @xmath211 increases . in this limit
the spin angular velocity is @xmath212 and velocity profile is linear with a slope given by the last term in eq .
( [ eq : torquesolu ] ) .
figure [ fig : exttorque ] depicts the two profiles for the butane liquid using @xmath213 m@xmath214s@xmath215 .
flow of butane as a result applying an external torque with @xmath216 .
the insert shows the corresponding angular velocity normalized with respect to the value in the limit of large @xmath211 , eq .
( [ eq : largeh ] ) , here denoted @xmath217 . ] from this one sees that the external torque produces a significant local flow ; the average flow is zero due to the system symmetry . in 2009
et al . _
@xcite showed that the coupling between the linear momentum and spin could be exploited in order to pump water through carbon nanotubes by application of a rotating field .
the was theory based on the ens equations at it was noted that in order to obtain a non - zero mean flow asymmetric boundary conditions must be employed which can be achieved by confining the fluid between two walls with different hydrophobicity in the case of water pumping .
recently de luca _ et al . _
@xcite performed extensive md simulations of the mechanism under experimentally feasible conditions , indicating that the mechanism is functional .
this could prove to be a way to overcome the large hydraulic resistance characterizing nanofluidic flows .
felderhof showed in 2011 that the coupling can also be utilized to perform plane - wave pumping @xcite and even propel microrobots @xcite .
we have derived the relevant dynamical equations for isotropic nanofluidic flows .
the formulation is based on the basic definition of a macroscopic field variable from the corresponding microscopic or molecular variable , and it includes the underlying molecular structure .
two intrinsic nanofluidic phenomena were discussed , namely , ( i ) the coupling between the spin angular momentum and ( ii ) the linear momentum and the non - local fluid response .
the important points are the following .
\(1 ) the effect of the coupling between the linear and spin angular velocities can be estimated through the characteristic length scale @xmath178 , eq .
( [ eq : lcdef ] ) . for large @xmath178 significant flow rate reduction
is observed , partly explaining the `` increased '' or `` effective '' viscosity reported in the literature .
other effects , like anisotropy , will also play a role in the change in effective transport properties . for polar molecular systems like water @xmath218 nm , and the coupling
must be considered on these length scales . for the non - polar fluids studied here @xmath178
is below 1 nm and the coupling effect is very small in most situations .
\(ii ) in general any fluid response can be described phenomenologically through a transport kernel that incorporates the spatial and temporal correlation effects . a method for calculating the shear viscosity kernel
was presented .
this showed that for non - highly viscous fluids the newtonian limit is reached on length scales of a few nanometers .
this is in agreement with the bocquet - charlaix criterion if the decay time for the stress autocorrelation function is applied as the relaxation parameter .
importantly , non - local effects are not present in simple flows where the strain - rate is linear with respect to spatial coordinate , which is the case for couette and poiseuille flows . for
non - linear flows the non - local response significantly affects the fluid stress for strain - rate variations on the atomic length scale .
\(iii ) for highly confined fluids molecular alignment phenomena and molecular deformation can occur along with the fluid layering , see fig .
[ fig : prof - butane ] . simple classical continuum theory does not include or account for such complex fluid structure
. it would be interesting to investigate this is more detail , for example , using the theory for liquid crystals @xcite . in conclusion ,
continuum theory is applicable even on the nanoscale if the relevant physical processes are modelled appropriately .
the authors wish to acknowledge lundbeckfonden for supporting this work as part grant no . r49-a5634 .
the centre for viscous liquid dynamics `` glass and time '' is sponsored by the danish national research foundation ( dnrf ) .
10 j.c.t .
eijkel and a. van den berg .
nanofluidics : what is it and what can we expect from it ? , 1:249267 , 2005 .
l. bocquet and e. charlaix .
nanofluidics , from bulk to interface . , 39:1073 , 2010 .
h. bruus . .
oxford university press , 2008 .
f. persson , l. h. thamdrup , m. b. l. mikkelsen , p. skafte - pedersen s. e. jaarlgard , h. bruus , and a. kristensen .
double thermal oxidation scheme for the fabrication of sio@xmath219 nanochannels .
, 18:245301 , 2007 .
landau and e.m .
lifshitz . .
elsevier , amsterdam , 1987 .
tritton . .
oxford science publications , 1988 .
b. lautrup . .
institute of physics publishing , bristol , 2005 .
r. l. rowley and m. m. painter .
diffusion and viscosity equation of state for a lennard - jones fluid obtained from molecular dynamics simulations .
, 18:1109 , 1997 .
hansen , j.c .
dyre , p.j .
daivis , b.d .
todd , and h. bruus .
nanoflow hydrodynamics .
, 84:036311 , 2011 .
l. h. thamdrup , f. persson , h. bruus , h. k. flyvbjerg , and a. kristensen . experimental investigation of bubble formation during capillary filling of sio@xmath219 nanoslits . , 91:163505 , 2007 .
k. p. travis , b. d. todd , and d. j. evans .
depature from navier - stokes hydrodynamics in confined liquids . , 55:42884295 , 1997 .
f. detcheverry and l. bocquet .
thermal fluctuations in nanofluidic transport . , 109:024501 , 2012 .
gupta , h.d .
cochran , and p.t . cummings .
nanorheology of liquid alkanes . , 150 - 151:125 , 1998 . g. karniadakis , a. beskok , and n. aluru .
springer , new york , 2005 .
g. lukaszewics . .
birkhauser , boston , 1999 .
eremeyev , l.p .
lebedev , and h. altenbach . .
springer , 2013 .
e. cosserat and f. cosserat .
sur la thorie de lesticit .
, 10:1 , 1896 .
e. cosserat and f. cosserat . .
hermann , 1909 .
a.m. sonnet , p.l .
maffettone , and e.g. virga .
continuum theory for nematic liquid crystals with tensorial order .
, 119:51 , 2004 .
p. muthu , b.v .
rathish kumar , and p. chandra .
a study of micropolar fluid in an annular tube with application to blood flow .
8:561 , 2008 . h. grad .
statistical mechanics , thermodynamics and fluid dynamics of systems with an arbitrary number of integrals . , 5:455 , 1952 .
dahler and l.e .
scriven . theory of structured continua i. general considarations of angular momentum and polarization .
, 27:504 , 1963 .
a. c. eringen .
contribution to mechanics .
oxford , 1969 .
pergamon , editor d. abir .
s. r. de groot and p. mazur . .
dover publications , 1984 . r. f. snider and k. s. lewchuk .
irreversible thermodynamics of a fluid system with spin . , 46:3163 , 1967 .
n. k. ailawadi , b. j. berne , and d. forster .
ydrodynamics and collective angular - momentum fluctuations in molecular fluids . , 3:14621472 , 1971 .
d. j. evans and w. b. streett .
transport properties of homonuclear diatomics ii . dense fluids . , 36:161176 , 1978 .
w. e. alley and b. j. alder .
generalized transport coefficients for hard spheres . , 27:3158 , 1983 .
j.p . boon and s. yip . .
dover publication , new york , 1991 .
d. j. evans and g. p. morriss .
. academic press , 1990 .
a. furukawa and h. tanaka .
nonlocal nature of the viscous transport in supercooled liquids : complex fluid approach to supercooled liquids .
, 103:135703 , 2009 .
puscasu , b.d .
todd , peter j. daivis , and j.s hansen .
nonlocal viscosity of polymer melts approaching their glassy state .
, 133:144907 , 2010 .
m. p. allen and d. j. tildesley . .
clarendon press , new york , 1989 .
y. wu , h. l. tepper , and g. a. voth .
flexible simple point - charge water model with improved liquid - state properties .
, 124:024503 , 2006 .
ryckaert and a. bellemans .
molecular dynamics of liquid alkanes .
, page 95 , 1978 .
j.wang , r.m .
wolf , j.w .
caldwell , p.a.kollman , and d.a . case .
development and testing of a general amber force field .
, 25:1157 , 2004 .
https://code.google.com / p / seplib/. j. a. thomas and j. h. mcgaughey .
reassessing fast water transport through carbon nanotubes .
, 8:27882793 , 2008 .
m. whitby , l. cagnon , m. thanou , and n. quirke .
enhanced fluid flow through nanoscale carbon pipes .
, 8:2632 , 2008 .
irving and j.g .
the statistical mechanical theory of transport processes .
iv . the equations of hydrodynamics . , 18:817829 , 1950 .
b. d. todd and p. j. daivis .
homogeneous non - equilibrium molecular dynamics simulations of viscous flow : techniques and applications . , 33:189 , 2007 .
j.s hansen , p.j .
daivis , j.c .
dyre , b.d .
todd , and h. bruus .
generalized extended navier - stokes theory : correlations in molecular fluids with intrinsic angular momentum .
, 138:034503 , 2013 .
j.s hansen .
generalized extended navier - stokes theory : multiscale spin relaxation in molecular fluids .
, 88:032101 , 2013 .
kadanoff and p.c .
hydrodynamic equations and correlation functions . , 281:800 , 2000 .
l. onsager .
reciprocal relations in irreversible processes .
i. , 37:405 , 1931 .
de zrate and j.v .
sengers . .
elsevier , amsterdam , 2006 .
s. sarman .
flow properties of liquid crystal phases of the gay - berne fluid .
, 108:7909 , 1998 .
r. j. d. moore , j. s. hansen , and b. d. todd . rotational viscosity of linear molecules : an equilibrium molecular dynamics study .
, 128:224507 , 2008 .
r. hartkamp , p.j .
daivis , and b.d .
density dependence of the stress relaxation function of a simple liquid .
, 87:0321551 , 2013 .
puscasu , b.d .
todd , peter j. daivis , and j.s hansen
. an extended analysis of the viscosity kernel for monatomic and diatomic fluids . , 22:195105 , 2010 .
j. s. hansen , p. j. daivis , k. p. travis , and b. d. todd .
parameterization of the nonlocal viscosity kernel for an atomic fluid . , 76:041121 , 2007 .
schrder , u.r .
pedersen , n.p .
bailey , s. toxvaerd , and j.c .
hidden scale invariance in molecular van der waals liquids : a simulation study . , 80:041502 , 2009 .
a. cavagna .
supercooled liquids for pedestrians . , 476:51 , 2009 .
b. d. todd , j. s. hansen , and p.j .
nonlocal shear stress for homogeneous fluids .
, 100:195901195904 , 2008 .
j. d. weeks , d. chandler , and h. c. andersen .
role of repulsive forces in determining the equilibrium structure of simple liquids .
, 54:52375247 , 1971 .
j. s. hansen , p. j. daivis , and b. d. todd .
local linear viscoelasticity of confined fluids . , 126:144706 , 2007 .
e. m. gosling , i. r. mcdonald , and k. singer . on the calculation by molecular dynamics of the shear viscosity of a simple fluid . , 26:1475 , 1973 .
travis , d.b .
searles , and d. evans . on the wavevector dependent shear viscosity of a simple fluid .
, 97:415 , 1999 .
b. d. todd and j. s. hansen .
nonlocal viscous transport and the effect on fluid stress .
, 78:051702 , 2008 . c. l. m. h. navier .
memoire sur les lois du mouvement des fluids .
, 6:389440 , 1823 .
l. bocquet and j .- l .
hydrodynamic boundary conditions , correlations functions , and kubo relations for confined fluids .
, 49:30793092 , 1994 .
hansen , b.d .
todd , and p.j .
prediction of fluid velocity slip at solid surfaces .
, 84:016313 , 2011 .
s. de luca , b.d .
todd , j.s .
hansen , and p.j .
daivis . molecular dynamics study of nanoconfined water flow driven by rotating electric fields under realistic experimental conditions .
, 30:3095 , 2012 .
j. badur , p. ziolkowski , and p. ziolkowski . on the angular velocity
slip in nano - flows . ,
2015 . j.s .
hansen , p. j. daivis , and b. d. todd .
viscous properties of isotropic fluids composed of linear molecules : departure from the classical navier - stokes theory in nano confined geometries . , 80:046322 , 2009 .
d. j. evans and g. p. morris .
nonlinear - response theory for steady planar couette flow .
, 30:1528 , 1984 .
j. delhommelle
. rotational viscosity of uniaxial molecules .
, 100:34793482 , 2002 .
r. edberg , g. p. morriss , and d. j. evans .
rheology of @xmath220-alkanes by nonequilibrium molecular dynamics . , 86:4555 , 1987 .
s. toxvaerd .
the structure and thermodynamics of a solid - fluid interface .
, 74:19982005 , 1981 .
gennes and j. prost . .
claredon press , 1993 .
i. bitsanis , t.k .
vanderlick , m. tirrell , and h.t .
tractable molecular theory of flow in strongly inhomogeneous fluids .
, 89:3152 , 1988 .
h. hoang and g. galliero .
shear viscosity of inhomogeneous fluids .
, 136:124902 , 2013 . k. p.
travis and k. e. gubbins .
poiseuille flow of lennard - jones fluids in narrow slit pores .
, 112:1984 , 2000 .
j. zhang , b.d .
todd , and k.p .
viscosity of confined inhomogeneous nonequilibrium fluids .
, 121:1077810786 , 2004 .
p. j. cadusch , b. d. todd , j. zhang , and p. j. daivis . a non - local hydrodynamic model for the shear viscosity of confined fluids : analysis of homogeneous kernel .
, 41:035501 , 2008 .
dalton , p.j .
daivis , j.s .
hansen , and b.d .
effects of nanoscale density inhomogeneities on shearing fluids .
, 88:052143 , 2013 .
e. lauga m. brenner and h. stone .
microfluidics : the no - slip boundary condition . in c. tropea , a.l .
yarin , and j.f .
foss , editors , _
springer handbook of experimental fluid mechanics_. springer , 2007 .
j. d. bonthuis , d. horinek , l. bocquet , and r. r. netz .
electrohydraulic power conversion in planar nanochannels .
, 103:144503 , 2009 .
b. u. felderhof .
efficiency of magnetic plane wave pumping of a ferrofluid through a planar duct .
, 23:092003 , 2011 . b. u. felderhof .
self - propulsion of a planar electric or magnetic microbot immersed in a polar viscous fluid .
, 83:056315 , 2011 .
s. sarman and d.j .
statistical mechanics of viscous flow in nematic fluids . , 99:90219036 , 1993 .
a macroscopic field variable @xmath221 is be given by the corresponding microscopic quantity @xmath222 associated with molecule @xmath223 through @xcite @xmath224 where @xmath225 is the dirac delta function and @xmath226 is the molecular center - of - mass position . in our treatment @xmath227
can be a scalar quantity ( _ e.g. _ the mass density ) or a vector quantity ( _ e.g. _ the momentum density ) . in fourier space we have for @xmath227 @xmath228 \equiv \widetilde{a}({\mathbf}{k } , t ) = \iiint_{-\infty}^\infty a({\mathbf}{r},t ) e^{-i{\mathbf}{k}\cdot { \mathbf}{r } } \ , \mathrm{d } { \mathbf}{r } = \sum_i a_i(t ) e^{-i{\mathbf}{k}\cdot { \mathbf}{r}_i(t ) } \ , .\ ] ] from now on we do not write the time and position dependencies of the free variables ( right hand side of the equations ) explicitly unless it provides important information . the rate of change in fourier space is given by where @xmath230 is the center - of - mass velocity .
the following identity has been applied , where @xmath231 is a scalar or a vector @xmath232 in the situation where @xmath231 is a vector the product @xmath233 is the dyadic between two vectors , @xmath234 and @xmath235 , and the resultant is the second rank tensor with components @xmath236 , @xmath237 .
the dyadic is not commutative , _
i.e. _ , @xmath238 , unless @xmath233 is symmetric .
the dyadic is also called the outer product and sometimes written as @xmath239 .
if @xmath234 and @xmath235 are parallel the dyadic is symmetric . to first order in wave vector eq .
( [ eq : sect0:rateexpanded ] ) becomes @xmath240 the particle velocity can be decomposed into a thermal ( or peculiar )
term @xmath241 and advective term @xmath242 by @xmath243 @xmath244 is the mass averaged fluid velocity in a region around @xmath245 .
this region is equivalent to the fluid particle volume encountered in the traditional continuum description @xcite as was briefly discussed in the introduction .
we assume that the thermal and advective velocities are uncorrelated and the thermal motion is , by definition , conserved with @xmath246 , where @xmath247 is the mass . from this eq .
( [ eq : sect : linear ] ) becomes @xmath248 \ \ \ ( \mbox{small } \ { \mathbf}{k}),\ ] ] where @xmath249 is a linear operator given by @xmath250 = \sum_i ( 1-i{\mathbf}{k}\cdot { \mathbf}{r}_i)\frac{{\mathrm{d}}a_i}{{\mathrm{d}}t } - i{\mathbf}{k}\cdot \sum_i { \mathbf}{c}_ia_i - i{\mathbf}{k}\cdot \sum_i { \mathbf}{u}({\mathbf}{r}_i , t)a_i \ .\ ] ] @xmath249 is henceforth refered to as the `` microscopic hydrodynamic operator '' ( mh - operator ) since it describes the microscopic interpretation of the fluid s dynamics in the hydrodynamic regime of small wave vectors .
the first term describes the rate of change of the microscopic quantity in question , the second and third terms are then the thermal and advective contributions to the dynamics , respectively .
we proceed and apply eqs .
( [ eq : ratechangea ] ) and ( [ eq : mho ] ) to express the balance equations of mass , linear momentum and angular momentum .
@xmath253({\mathbf}{k},t ) & = \sum_i ( 1-i{\mathbf}{k}\cdot { \mathbf}{r}_i)\frac{{\mathrm{d}}m_i}{{\mathrm{d}}t } - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{c}_i - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{u}({\mathbf}{r}_i , t ) \\ & = - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{u}({\mathbf}{r}_i , t ) \end{aligned}\ ] ] since the mass of each particle is constant and @xmath254 .
the dynamics in fourier space is thus simply @xmath255 relating to the general balance equation , eq.(1 ) in the manuscript , in the case of zero production term , one has @xmath256 , and no surface forces @xmath257 in fourier space this yields @xmath258 using that @xmath259({\mathbf}{k},t ) = -i { \mathbf}{k } \cdot \widetilde{\rho { \mathbf}{u}}$ ] , which follows from partial integration of eq.([eq : ftransform ] ) . in the limit of small wave vector we thus identify eq .
( [ eq : sect1:micromassbalance ] ) as the microscopic interpretation of the continuous mass balance equation .
in general , there is no microscopic definition of the streaming velocity @xmath7 in the form of eq .
( [ eq : sect1:micromacro ] ) @xcite .
rather , the linear fluid motion is given through the momentum density @xmath260 .
note , hansen and mcdonald @xcite define a _
current _ from the microscopic velocities and eq .
( [ eq : sect1:micromacro ] ) , which is then the correct mass weighted averaged streaming velocity for single species fluids .
if @xmath261 is the linear momentum of particle @xmath223 we have @xmath262 applying the mh - operator we have @xmath263({\mathbf}{k},t ) & = \sum_i ( 1-i{\mathbf}{k}\cdot { \mathbf}{r}_i)m_i\frac{{\mathrm{d}}{\mathbf}{v}_i}{{\mathrm{d}}t } - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{c}_i{\mathbf}{v}_i - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{u}({\mathbf}{r}_i , t){\mathbf}{v}_i \\ & = \sum_i ( 1-i{\mathbf}{k}\cdot{\mathbf}{r}_i ) { \mathbf}{f}_i^{tot } - i{\mathbf}{k}\cdot \sum_i m_i { \mathbf}{c}_i { \mathbf}{c}_i - i{\mathbf}{k } \cdot \sum_i m_i{\mathbf}{u}({\mathbf}{r}_i , t){\mathbf}{u}({\mathbf}{r}_i , t ) \label{eq : sect1:hydro_mom } \end{aligned}\ ] ] @xmath265 is the total force acting on molecule @xmath223 .
this force is decomposed into the force on @xmath223 due to interactions with all other molecules denoted henceforth by @xmath266 and external forces @xmath267 , _ i.e. _ , @xmath268 . the first term on the right hand side of eq .
( [ eq : sect1:hydro_mom ] ) is @xmath269 using newton s third law @xmath270 and the identity eq .
( [ eq : dyadicindentity ] ) . substituting this into eq .
( [ eq : sect1:hydro_mom ] ) and rearranging one arrives at @xmath271 where @xmath272 we recognize eq .
( [ eq : sect1:momentumcurrent4 ] ) as the balance equation for linear momentum in fourier space .
importantly , the last term on the right hand side gives _
one _ possible microscopic interpretation of the pressure tensor @xmath13 in the zero wave vector limit @xcite .
let @xmath273 be the system volume . then according to eq .
( 1 ) in the manuscript and ( [ eq : sect1:momentumcurrent4 ] ) we have for the zero wave vector pressure tensor @xmath274 this is the irving - kirkwood interpretation @xcite and will be discussed in the following .
the configurational part of the pressure tensor is the term @xmath275 and can be written in a different form assuming pairwise additive interactions only .
using newton s third law once again , @xmath276 , where @xmath277 is the force on @xmath223 due to interactions with @xmath278 , we get @xmath279 where @xmath280 .
the irving - kirkwood pressure tensor then reads @xmath281
at this point we emphasize the difference between the atomic and molecular formalisms .
if particles @xmath223 and @xmath278 represent point - masses ( including atoms in the molecules ) , rather than structured molecules , the vectors @xmath282 and @xmath277 are parallel , the dyadic @xmath283 is symmetric .
this implies that the pressure tensor is symmetric . in the case of structured molecules ,
however , the force on molecule @xmath223 due to @xmath278 is given by @xmath284 where indices @xmath285 and @xmath286 represent atom @xmath287 in molecule @xmath223 and atom @xmath288 in molecule @xmath278 , see fig .
[ fig : sect1:antisymmetric ] . the force @xmath277 needs not be parallel to the vector @xmath282 connecting the molecules center - of - masses and the configurational part of the pressure tensor is not generally symmetric .
there is no ambiguity between the two formalisms .
they are trivially equivalent for point - mass particles , and for structured molecules one can be expressed in terms of the other and the mass dispersion tensor @xcite .
the two formalisms provide different information and we use the molecular one here .
we use the term atom loosely : it represents a point - mass spherical particle , and can be a single atom or a united group of atoms like the methyl group . in the molecular formalism
the pressure tensor decomposes into a sum of the symmetric @xmath289 and anti - symmetric @xmath290 parts such that @xmath291 with @xmath292 following evans and morriss @xcite we here use the stack notation to indicate the symmetric and anti - symmetric parts . the symmetric part can be further decomposed into a sum of the equilibrium pressure @xmath293 , the viscous pressure @xmath294 , and the trace - less symmetric tensor @xmath295
@xcite @xmath296 where @xmath297 is the second rank identity tensor and @xmath298 .
the anti - symmetric part of the pressure tensor has zero diagonal components and three independent off - diagonal components .
it is therefore often written as a ( pseudo ) vector dual @xmath299 of the tensor . here
the subscripts indicate the tensor components . from the definition of @xmath290 , eq . ( [ eq : decompose ] ) , one has for @xmath300 @xmath301 here @xmath302 and @xmath40 are the unit vectors along the @xmath50- , @xmath175- and @xmath51-axes , respectively . from eq .
( [ eq : vdualtorque ] ) it is clear that the anti - symmetric pressure is associated with the torque on @xmath223 about @xmath278 and gives rise to a change of orbital angular momentum , _ i.e. _ , that of the center - of - mass motion of the molecules .
importantly , when the pressure tensor has a non - zero anti - symmetric part , the orbital angular momentum is a non - conserved quantity @xcite ; we address this in the following .
conservation of orbital angular momentum is sometimes used as an argument for a symmetric pressure , see for example aris @xcite , but note that later aris also allows for the possibility of asymmetry . finally , notice that the kinetic part of the pressure tensor @xmath303 is a symmetric dyadic and
does not enter the anti - symmetric part .
consider now the case when no external forces are present .
following the approach for the linear momentum we write the orbital angular momentum density @xmath304 as @xmath305 where @xmath306 is the orbital angular momentum per unit mass , and @xmath307 is the orbital angular momentum of molecule @xmath223 .
note , the angular momentum is defined with respect to some specific choice of coordinate system .
the mh - operator acting on @xmath308 gives [ eq : mho_l ] @xmath309({\mathbf}{k},t ) & = \sum_i ( 1-i{\mathbf}{k}\cdot { \mathbf}{r}_i ) \frac{{\mathrm{d}}{\mathbf}{l}_i } { { \mathrm{d}}t } - i{\mathbf}{k}\cdot\sum_i{\mathbf}{c}_i{\mathbf}{l}_i - i{\mathbf}{k}\cdot\sum_i{\mathbf}{u}({\mathbf}{r}_i , t){\mathbf}{l}_i \\ & = { \mathbf}{n } - i{\mathbf}{k } \cdot \sum_i { \mathbf}{r}_i{\mathbf}{n}_i - i{\mathbf}{k } \cdot \sum_i { \mathbf}{c}_i { \mathbf}{l}_i - i{\mathbf}{k}\cdot \sum_i { \mathbf}{u}({\mathbf}{r}_i , t){\mathbf}{l}_i , \end{aligned}\ ] ] where @xmath310 is the torque on molecule @xmath223 and @xmath311 is the sum of torques @xmath312 the torque @xmath311 does not sum to zero as the orbital angular momentum is a non - conserved quantity @xcite .
rather it is the total angular momentum , _ i.e. _ , the sum of orbital and spin ( or intrinsic ) angular momenta , which is conserved . for point - masses
spin angular momentum is not present ( or meaningful ) as there is no moment of inertia . in this case
the orbital angular momentum is conserved and the torque @xmath311 must be zero , that is , @xmath313 . the rate of change of orbital angular momentum is then given by @xmath314 this is the balance equation for the orbital angular momentum in fourier space .
the spin angular momentum density @xcite @xmath315 is given by @xmath316 @xmath317 is the spin angular momentum per unit mass , @xmath318 is the spin angular momentum of molecule @xmath223 , and @xmath319 is the vector from the center - of - mass to atom @xmath287 , see fig .
[ fig : sect1:antisymmetric ] . applying the mh - operator and following the procedure in eq .
( [ eq : mho_l ] ) above , the rate of change is to lowest order in wave vector @xmath320 where @xmath321 is the sum of torques on @xmath223 about the center - of - mass and @xmath322 .
equation ( [ eq : sect1:spinbalance ] ) is the balance equation for the spin angular momentum in fourier space .
the last term in eq . ( [ eq : sect1:spinbalance ] ) defines the zero wave vector irving - kirkwood couple tensor @xcite @xmath323 .
the flux of spin angular momentum associated with the couple tensor can be important for flows in extreme confinements .
after volume averaging @xmath324 can be written as @xmath325 where @xmath326 is the torque , specifically the couple force , on molecule @xmath223 due to interaction with all atoms @xmath288 in molecule @xmath278 , see fig .
[ fig : sect1:antisymmetric ] .
the couple tensor is not symmetric and can be decomposed into symmetric and anti - symmetric parts just as above for the pressure tensor , that is , @xmath327 also , one has a vector dual @xmath328 for the anti - symmetric part of the couple tensor . above we discussed the mass and momentum balance equations in the framework of the microscopic picture , which is easiest done in fourier space .
we will now return to the corresponding real space formulations , but write the balance equations in a slightly different form compared to eq .
( 1 ) in the main manuscript . in real space eq .
( [ eq : sect1:massbalance ] ) is the mass balance equation . using the product rule on the right hand side and re - arranging @xmath331 where , in general , the material derivative @xmath332 is defined by @xmath333 likewise , from eqs .
( [ eq : sect1:momentumcurrent4 ] ) and ( [ eq : sect1:spinbalance ] ) we get the relevant momentum and spin balance equations in real space after application of eq .
( [ eq : sect1:massbalancereal ] ) where @xmath29 represents the external production term for the spin angular momentum . in eqs .
( [ eq : sect1:finalbalance ] ) we have applied the identity @xmath335 ( @xmath336 or @xmath324 ) .
interestingly , the term @xmath337 in eq .
( [ eq : sect1:finalbalancespin ] ) can be regarded as a production term even in the absence of an external torque .
now , eqs .
( [ eq : sect1:massbalancereal ] ) , ( [ eq : sect1:finalbalancemom ] ) and ( [ eq : sect1:finalbalancespin ] ) give the relevant balance equations in the limit of small wave vectors or , equivalently , large length scales . also , from eqs .
( [ eq : sect1:finalbalance ] ) one immediately sees that by ignoring the anti - symmetric part of the pressure tensor we obtain the classical momentum balance equation for the linear momentum . as discussed above this
applies to point - mass structure - less fluids .
also , one observes that the dynamics of the spin angular momentum is then determined by the couple forces alone and the spin is a conserved quantity .
the treatment above is general .
we shall now focus on systems where the molecules can be well approximated as uni - axial rigid body molecules .
uni - axial molecules are defined as molecules where the principal moment of inertia tensor has two nonzero components , denoted @xmath338 , the third one being zero as the rotation around the main molecular axis is not associated with any inertia .
this includes di - atomic and linear molecules such as carbon dioxide , but other molecules are also well approximated as uni - axial and rigid , for example , butane as we will see below . the treatment in this section is not as detailed as above , and we refer the reader to refs .
@xcite .
it is convenient to describe the spin dynamics in the principal frame of reference where the moment of inertia tensor is constant . from the euler equation of motion for rigid bodies @xcite
one can show that in this frame and for uniaxial molecules the left hand side of eq .
( [ eq : sect1:finalbalancespin ] ) becomes @xmath339 where @xmath340 @xcite and @xmath341 is the angular velocity .
note , this does not hold in general where nonlinear coupling between the angular velocity components are present .
the general situation , however , can be included into the theory . here @xmath343 and @xmath24 are the shear , bulk and rotational viscosities , respectively , and @xmath344 and @xmath27 are the corresponding spin viscosities .
usually , a thermodynamic force is defined through the gradients of the field variables , however , in eq .
( [ eq : antistressconst ] ) we recognize the thermodynamic force @xmath345 as twice the difference between the classical prediction of the angular velocity , @xmath346 , and the actual one .
this force is denoted the sprain rate @xcite .
substituting eqs .
( [ eq : sect1:constrelations ] ) into eqs .
( [ eq : sect1:finalbalance ] ) and using eq .
( [ eq : sect1:angularvel ] ) we arrive at the extended navier - stokes ( ens ) equations @xcite we now present values for the transport coefficients entering eqs .
( [ eq : sect1:constrelations ] ) for the molecular fluids studied here .
methane is modelled as a point - mass molecule where the coupling is irrelevant , _
i.e. _ , we let @xmath349 .
butane and water are not uni - axial . for the butane model used the principal moment of inertia components are @xmath350 m@xmath214 , @xmath351 m@xmath214 and @xmath352 m@xmath214 , that is , the molecule has one major axis and two almost identical minor axes and we can expect the theory to hold reasonably well . in ref .
the water molecule was considered as a dipolar rotator with an effective moment of inertia of @xmath353 m@xmath214 .
this approach will be adopted here . in table
[ table : coeff ] the relevant state points and transport coefficients are listed . in the treatment below the dynamics is decomposed into transverse shear and longitudinal bulk modes . for linear momentum
the focus is on the shear mode , and the bulk viscosity is therefore not listed in the table .
the coefficients are evaluated from independent equilibrium md simulations as prescribed in refs . .
the values for the dumbbell model are given in reduced md units of length scale @xmath354 , energy scale @xmath355 , and mass @xmath65 , see ref .. .[table : coeff ] @xmath356 from ref . .
@xmath357 quantities given in reduce md units .
+ state points , transport coefficients and moments of inertia for the systems studied . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] | this paper introduces the fundamental continuum theory governing momentum transport in isotropic nanofluidic flows .
the theory is an extension to the classical navier - stokes equation , which includes coupling between translational and rotational degrees of freedom , as well as non - local response functions that incorporates spatial correlations .
the continuum theory is compared with molecular dynamics simulation data for both relaxation processes and fluid flows showing excellent agreement on the nanometer length scale .
we also present practical tools to estimate when the extended theory should be used .
it is shown that in the wall - fluid region the fluid molecules align with the wall and in this region the isotropic model may fail and a full anisotropic description is necessary in order to describe this region . |
recently , the study of unstable nuclei is one of the hot subjects . in the light unstable nuclei ,
the experimental informations near the drip lines increase rapidly and many interesting phenomena are found .
the exotic structures such as the halo , the skin structures , and the clustering structures are suggested in some light unstable nuclei .
are the features of the clusters in unstable nuclei similar to those in the stable nuclei ?
the shift of the magic number and the deformations of unstable nuclei are the concerned problems .
our aim is to research on the excited states of unstable and to study systematically the various structures of unstable nuclei .
we apply a theoretical method of antisymmetrized molecular dynamics ( amd ) .
the method of variation after projection(vap ) in the framework of amd has been already proved to be useful to investigate the excited states of light unstable nuclei @xcite as well as stable nuclei @xcite .
amd is a powerful approach which is applicable to the excited states of general nuclei .
the wave function of amd can represent various clustering structures as well as shell - model - like structures , because there is no inert core in the framework . in the previous study on the excited states of @xmath1be with amd @xcite
, we found that the molecule - like structures may coexist in the low - energy region with the ordinary shell - model - like states .
the data of the transitions such as @xmath2 and gamow - teller transitions are so helpful to classify the excited levels into rotational bands .
the reader is referred to the paper @xcite for the detailed discussions .
in this section we explain the formulation of amd for the study of the nuclear structure of the excited states . the wave function of a system is written by amd wave functions , @xmath3 an amd wavefunction of a nucleus with mass number
@xmath4 is a slater determinant of gaussian wave packets ; @xmath5,\\ \chi_{\xi_i}= \left(\begin{array}{l } { 1\over 2}+\xi_{i}\\ { 1\over 2}-\xi_{i } \end{array}\right ) , \end{array}\right . \end{aligned}\ ] ] where the @xmath6th single - particle wave function @xmath7 is a product of the spatial wave function , the intrinsic spin function and the iso - spin function .
the spatial part @xmath8 is presented by complex parameters @xmath9 ,
@xmath10 , @xmath11 , of the centers of gaussians , and @xmath12 is the intrinsic spin function parameterized by @xmath13 .
@xmath14 is the iso - spin function which is fixed to be up(proton ) or down(neutron ) in the present calculations .
thus an amd wave function is parameterized by a set of complex parameters @xmath15 which determine the centers of gaussians of the spatial parts and the directions of the intrinsic spins for all single - particle wave functions .
if we consider a spin - parity eigen state projected from a amd wave function , the total wave function is linear combinations of slater determinants , @xmath16 where @xmath17 is a parity projection operator .
the integrations for the euler angles are numerically calculated by a summation of mesh points on @xmath18 . in principal
the total wave function can be a superposition of independent amd wave functions , @xmath19 we make variational calculations to find the state which minimizes the energy of the system ; @xmath20 by the method of frictional cooling @xcite , one of the imaginary time methods .
the time development of the wave function @xmath21 is given by the equations , @xmath22 @xmath23 with arbitrary real numbers @xmath24 and @xmath25 .
it is easily proved that the energy of the system decreases with time . after sufficient time steps of cooling
, we obtain the optimum parameters for the minimum - energy state . in order to obtain the wave function for the lowest @xmath26 state , we perform the energy variation for the spin parity eigenstates projected from an amd wave function . in this case
the trial function is @xmath27 . for the preparation of vap calculations ,
first we make variational calculations before the spin projection in order to choose an initial wave function for each parity and an appropriate @xmath28 quantum for each spin - parity state . then we perform vap calculation for @xmath29 with the adopted @xmath28 quantum from the initial state . in the vap procedure , the @xmath30-axis of euler angle in the total - angular momentum projection
is not necessarily same as the principal @xmath31-axis .
the direction of the intrinsic deformation in the body - fixed frame is automatically determined in the energy variation depending on the adopted @xmath28 quantum . however , in many cases , the approximately principal axis @xmath31 obtained by a vap calculation is found to be almost equal to the @xmath30-axis . as mentioned above , with the vap calculation for @xmath32
, we obtain the set of parameters @xmath33 which presents the wave function for the first @xmath26 state . to search the parameters @xmath34 for the higher excited @xmath26 states
, we make the energy variation for the orthogonal component of a spin - parity projected amd wave function to the lower states by superposing the wave functions .
if the amd wave functions for the higher excited states have energetically local minimums , we do not need orthogonalization in vap calculations .
after vap calculations for various @xmath35 states , we obtain many intrinsic states @xmath36 , @xmath37,@xmath38 , @xmath39 .
finally we improve the wave functions by diagonalizing the hamiltonian matrix @xmath40 ) for all the intrinsic states and ( @xmath41 ) . in comparison with the experimental data ,
the theoretical values are calculated with the final states after the diagonalization .
we calculate the excited states of @xmath42be and @xmath43be .
the adopted interactions are the mv1 case3 ( the finite two - body and the zero - range three - body interactions @xcite ) for the central force , and g3rs for the spin - orbit force @xcite . the coulomb interaction is approximated by a sum of seven gaussians . in the present calculations ,
we adopt the majorana parameter @xmath44 and omit the bartlett and heisenberg terms in the two - body central force . as for the strength of the spin - orbit force , we try two cases of the parameter , @xmath45=3700 mev and @xmath45=2500 mev .
the resonance states are treated as a bound - state approximation in amd , because the restriction of the gaussian forms of the wave functions makes an artificial barrier for outgoing particles .
the energy levels of @xmath42be are displayed in figure [ fig : be12spe ] .
the theoretical results are calculated with the stronger spin - orbit force as @xmath46 mev which can reproduce the abnormal parity of the ground @xmath47 state in @xmath43be .
many excited states are seen in the low - energy region .
even though the @xmath42be has the neutron magic number 8 , it is surprising that the calculated ground state is not the ordinary state with the neutron closed @xmath48-shell , but a 2 particle 2 hole state with the developed clustering structure . as the evidence of the breaking of the closed @xmath48-shell ,
the calculations reproduce well the small value of the gamow - teller(gt ) transition strength from @xmath42be(@xmath49 ) to @xmath42b(@xmath50 ) .
the experimental @xmath51 is about one third as small as the value expected for the closed @xmath48-shell @xmath42be(@xmath49 ) . in other words ,
the weak gt transition is one of the proofs for the disappearance of the neutron magic number 8 in neutron - rich be isotopes .
it is consistent with the shell model analysis by t. suzuki et al .
we also find many excited states with the developed molecule - like structures which construct the rotational bands @xmath52 , @xmath53 , @xmath54 in the low - energy region .
since the ground state has the large deformation in the present calculations , the predicted @xmath55 in the ground band is 14 e@xmath56@xmath57 and is much larger than the value 7 e@xmath56@xmath57 of @xmath58 in the second @xmath59 band which consists of the closed @xmath48-shell states .
= 12 cm in this subsection , we discuss the excited states of @xmath43be . in fig
[ fig : be11spe ] , the calculated energy levels are shown with the experimental data .
we use the rather strong spin - orbit force in case ( 1 ) and the weaker ls force in case ( 2 ) .
the binding energies are underestimated as 58 mev in case ( 1 ) and as 54 mev in case ( 2 ) .
it is not difficult to fit to the experimental binding energy by choosing a smaller value for the majorana parameter . in the case
( 1 ) with the stronger ls force , the abnormal spin - parity @xmath47 state comes down lower than the normal negative parity states , which is consistent with the experimental data . in both cases , we found some rotational bands @xmath60 , @xmath61 and @xmath62 , although the order of the levels depends on the adopted interaction .
the @xmath62 band has a large moment of inertia and reaches to the high spin states .
the highest spin is @xmath63 in case(1 ) and @xmath64 in case ( 2 ) .
it is expected that the @xmath62 band may reach to the further high spin such as a @xmath65 state in the calculations with a smaller value of the majorana parameter .
next we analyze the intrinsic wave functions of the obtained rotational bands . in the results of @xmath43be ,
2@xmath0 clustering structures appear very often in many states .
the schematic figures are presented in fig .
[ fig : be11int ] .
two alpha cores are found in the states of the lowest negative parity band @xmath60 ( fig.[fig : be11int](a ) ) . in the positive parity states ,
one neutron is in the @xmath66 orbit and the 2@xmath0 clustering develops ( fig.[fig : be11int](b ) ) .
we think that one of the reason of the energy gain of the positive parity states is considered to be the clustering development .
the interesting point is that the @xmath62 band comes from the well - developed molecule - like structure with 2 neutrons in the @xmath66 orbit ( fig.[fig : be11int](c ) ) .
the large moment of inertia results from the largely deformed intrinsic structure . at about 10 mev
, we found another characteristic state where one of the 2 @xmath0-clusters completely breaks ( fig.[fig : be11int](d ) ) . in this state the intrinsic spins of 2 protons couple up to be totally a unit .
as mentioned above , various structures coexist in the low energy region in @xmath43be .
it is expected that the valence neutrons play important roles in these excited states .
we discuss the gamow - teller(gt ) transitions in order to analyze the level structures of @xmath43be .
first we consider @xmath67 decay from @xmath43b .
the experimental data of gt - type transitions from the ground state of @xmath43b are deduced from the @xmath68 cross sections of the charge exchange reactions @xcite . comparing the calculated log ft values with experimental data ( fig.[fig : be11beta ] ) , the observed three levels
well correspond to the excited states in the lowest @xmath60 band .
it is natural that the gt transitions from @xmath43b to these levels are reasonably strong because both the parent and the daughter states are ordinary @xmath48-shell states .
the beta decays from @xmath43li are shown in fig.[fig : be11beta ] .
we notice that the ideal 2@xmath0-cluster structures in the daughter @xmath43be states forbid any gt transitions from @xmath43li because of the pauli forbidden . in other words ,
the gt transitions from @xmath43li are allowed only if the daughter states of @xmath43be have the component with the breaking of the @xmath0 clusters .
the calculations well agree with the experimental log ft values . according to the calculated results ,
the lowest @xmath69 and @xmath70 states in the @xmath60 band have the more than 10 % breaking of the clusters .
one of the reason for the weak gt transition to the lowest @xmath71 state is the clustering structure of @xmath43be .
another reason is suggested by suzuki et.al @xcite as the halo structure of the parent @xmath43li state , which may not be described enough in the present amd calculations . in the calculations , transitions for the higher excited states in the @xmath62 band
are suppressed because of the @xmath0 clusters in the developed molecule - like structures . the strong transition to the excited state at 8 mev well fits to the calculated @xmath69 state ( @xmath7210 mev ) where one of the two @xmath0 clusters completely breaks . as for the @xmath73 transitions ,
the @xmath74 of the present calculations is much smaller than the experimental data .
we should describe the wave function more precisely to reproduce the strength @xmath75 .
we studied the structure of the excited states of @xmath42be and @xmath43be and analyzed the level structure with the help of the data of gamow - teller transitions . in @xmath42be
, it is found that the ground state has the developed clustering structure instead of the neutron @xmath48-shell closed state .
the main component of the ground state is the @xmath76-@xmath77 state in terms of a simple shell model .
that is to say the neutron magic number 8 disappears in @xmath42be .
it is consistent with the small gt strength to @xmath42be .
many molecule - like states make rotational bands in low energy regions . in the case of @xmath43be ,
various structures are seen in the excited states . in the results rotational bands with the the developed clustering states
are suggested .
we discuss the gamow - teller transitions comparing the amd calculations with the experimental data .
the experimental data of gt transitions from @xmath43b well agree with the theoretical results of the decay to the ordinary @xmath48-shell states of @xmath43be . with the help of the log ft values for the transitions from @xmath43li , we can estimate the dissociation of @xmath0 clusters in the daughter @xmath43be nucleus .
the observed strong gt transitions from @xmath43li to @xmath43be at 8 mev comes from the breaking of the one of the 2 @xmath0 clusters in the excited state of the daughter nucleus @xmath43be .
y. kanada - enyo , h. horiuchi and a. dot , j. phys .
part . phys . * 24 * 1499 ( 1998 ) . y. kanada - enyo , h. horiuchi and a. dot , phys .
c * 60 * , 064304 ( 1998 ) y. kanada - enyo , phys .
* 81 * , 5291 ( 1998 ) .
a. dot , h. horiuchi , and y. kanada - enyo , phys .
c * 56 * , 1844 ( 1997 ) .
y. kanada - enyo , a. ono , and h. horiuchi , phys .
c * 52 * , 628 ( 1995 ) ; y. kanada - enyo and h. horiuchi , phys .
c * 52 * , 647 ( 1995 ) .
y. kanada - enyo and h. horiuchi , prog .
. phys . * 93 * , 115 ( 1995 ) .
t. ando , k.ikeda , and a. tohsaki , prog .
. phys . * 64 * , 1608 ( 1980 ) . n. yamaguchi , t. kasahara , s. nagata , and y. akaishi , prog .
theor . phys . *
62 * , 1018 ( 1979 ) ; r. tamagaki , prog .
39 * , 91 ( 1968 ) .
t. suzuki and t. otsuka , phys .
c*50 * , r555(1994 ) ; t. suzuki and t. otsuka , phys .
c*56 * , 847 ( 1997 ) .
i. daito , et al .
b * 418 * 27 ( 1998 ) . n. aoi , et al .
phys . a*616 * 181c ( 1997 ) . | we study the structures of the unstable be isotopes with the theoretical method of antisymmetrized molecular dynamics .
it is found that various structures of the excited states appear in the low - energy region of neutron - rich be nuclei .
focusing on the 2@xmath0 clustering , we analyze the intrinsic structures with the help of the experimental data of gamow - teller transitions . |
for many galaxies , the cluster environment is considered hostile , often leading to interactions bound to modify their structure and perturbing the quiet deployment of their properties .
this is at least what is happening _ now _ , what we observe in the dense cluster environments at the present epoch in the form of perturbed morphologies .
however , the cluster environment has also had a great success in favoring the formation of massive galaxies , specifically the most luminous galaxies that are found .
how has this occurred , and when ? and what is the relation of these brightest cluster galaxies with the surrounding cluster members , the cluster morphology , and the density of the hot intracluster gas ?
+ the study of brightest cluster members and of their structural parameters has been traditionally pursued .
the better visibility of these objects has made them useful cosmological tools , similar to supernovae in certain aspects . in order to use them as standard candles ,
their properties have to be understood , and their possible differences studied in terms of induced potential biases due to similar yet somewhat different environments .
this line of research is still being followed , and would be further complemented by an investigation of the properties of these giant galaxies aimed at understanding their process of formation and evolutionary status .
+ we have recently completed a photometric survey of the cores of clusters of galaxies selected both through their x - ray emission properties and optically through the density contrast over the field ( garilli et al .
1996 , from now on gar96 ) . in that paper
it was shown that the same kind of passive evolution is shared by the whole early type galaxy population in a cluster , and no difference is evident between clusters selected in either way . in this paper
, we analyze the data set of gar96 in an attempt to infer the structural properties for a subsample of brightest cluster members ( bcm ) , and we try to demonstrate that there is a correlation between the size of the bcm and the _ local _ galaxy density within the cluster , as andreon et al .
( 1992 , 1995 ) had suggested from a study of a much smaller sample . + this paper
is organized in the following way : in section 2 , we describe the sample and the analysis tools ; in section 3 , we present the properties of the studied galaxies , specifically their isophotal shapes , their surface brightness profiles and color profiles ; in section 4 , we discuss the inferred bcm properties in relation to some other cluster observable parameters . + throughout this paper we adopt @xmath8 km s@xmath9 mpc@xmath9 and @xmath10 .
the number of bcms free from large superposed objects and for which our data allow a surface brightness analysis out to a reasonable radius is 17 , out of the 67 clusters for which @xmath0 , @xmath1 and @xmath2 photometry had been secured from the 2.1 m telescope of the _ observatorio astronomico nacional _ in san pedro martir , mexico .
details of the observations are given in gar96 .
table 1 lists the clusters in which the bcms are found , together with their redshifts , the bautz
morgan type ( bautz & morgan 1970 ) and the richness class .
clusters marked
blue or red are clusters in which the early type galaxy sequence is either significantly bluer or redder than what expected from a properly normalized virgo color magnitude relation ( see gar96 ) .
these clusters could have followed an evolution slightly different from the one of the bulk of the clusters we studied .
+ the bcm subsample spans almost the full redshift range of the cluster sample , @xmath11 ranging from 0.049 to 0.191 , it includes 8 x ray selected and 12 abell clusters ( abell , corwin & olowin 1989 ) ( 3 clusters being included both in the abell catalog and in the emss ( gioia et al .
1990 ) catalog ) , with b
m type i , i ii , ii and ii iii and richness class from 0 to 2 .
however , if we can claim that our sample spans other a vast range of cluster properties , we can not say that we have an unbiased sample .
our selection of objects favors galaxies with large apparent radii .
it is worth mentioning that the selection criteria of the main cluster sample did not consider in any way the presence of a bcm of certain characteristics , nor the observations were specifically tailored to study their surface brightness properties .
the 51 images containing the 17 bcms were reduced and flux calibrated as described in gar96 .
surface photometry , color profiles and structural parameters have been obtained following the procedures described in andreon et al .
+ briefly , the _ sky background _ has been determined in each image by averaging the count rates in independent areas of @xmath12 pixels free from detectable objects and as distant as possible from the bcm .
we further required that the growth curves of each bcms actually reached saturation once assumed a sky background level .
_ superposed or nearby objects _ have been removed and substituted with the interpolated value from nearby pixels prior to any further analysis . in some cases the areas occupied by these objects have been masked .
it must be said than in any case these removed objects are faint and small compared with the bcms .
it is worth recalling that the absence of bright superposed objects was one of the selection criteria for our subsample .
_ eccentricities and position angles _ were computed by means of the stsdas package in iraf . in order to keep the s / n ratio approximately constant as the radius increases ,
we ran the procedures several times over the same images rebinned by a proper factor . _ seeing effects _ were considered and the structural parameters have been determined only by taking into account data points outside an aperture 4 times the fwhm of the seeing disc .
the average seeing during our observations was @xmath13 .
we have obtained the _ eccentricity and position angle _ profiles outside the region affected by seeing and used intensity weighted averages and center coordinates determined by stsdas to compute the _ surface brightness profiles_. these profiles were obtained by integration of the counts in elliptical coronae of fixed eccentricity and position angle and of a step variable according to s / n criteria . from the surface brightness profiles in the three filters , _
color profiles _ have been extracted and analyzed with respect to possible color gradients following the method described by sparks & jrgensen ( 1993 ) and further developed by andreon et al .
ellipses are an excellent fit to the isophotes of every bcm studied here , as implied by their @xmath14 fourier coefficients that do not deviate from 0 in any significant manner .
however , only effects larger than 5% could be detected in our data .
+ for all galaxies in our sample , the eccentricity and position angle profiles do not significantly differ in the three filters .
figure 1 shows the eccentricity and position angle plots as a function of the logarithm of the major axis expressed in kpc . in table 2 , for each bcm ( column 1 ) , we give the minimum and maximum radius in kpc over which these parameters have been computed ( column 2 ) , an intensity weighted value of the eccentricity @xmath15 ( column 3 ) and its rate of variation along the radius expressed as @xmath16 kpc@xmath9 ( column 4 ) , an intensity weighted value of the position angle @xmath17 in degrees measured clockwise from the e w direction ( column 5 ) and its rate of variation along the radius , also expressed as @xmath18 kpc@xmath9 ( column 6 ) . +
as one can notice , the eccentricities are never larger than 0.3 , except in ms1558 + 3321 ( a2145 ) , for which @xmath19 . only in the cases of ms0102 + 3255 ,
a401 and ms1558 + 3321 ( a2145 ) the eccentricity increases rapidly with the radius .
the isophotes are concentric within @xmath20 , which is the precision limit that can be achieved with our data .
the rule is also to have a constant position angle of the isophotes , except in the case of ms0013 + 1558 for which we detect a change . in figure 2
we present the bcm rest
frame surface brightness profiles in the @xmath1 bandpass plotted as function of @xmath6 . at small radii
our data are affected by seeing effects and consequently for those central regions the following arguments do not apply . at large radii
we are limited by our flat fielding and s / n ratios .
the arrows in the plots indicate the galaxy rest frame surface brightness at a level 5 magnitudes fainter than the sky background . in the case of a401
our profile stops before this surface brightness limit as the edge of the ccd image was reached .
the structural parameters have been derived considering only those portions of the profiles , while , whenever meaningful , the profiles themselves have been extended to fainter surface brightnesses in order to investigate the possible presence of extended halos .
+ for each galaxy in figure 2 we also show as a continuous line the best fitting de vaucouleurs law ( de vaucouleurs 1948 ) . in the fitting procedure ,
only those data points that can be considered statistically independent given the image seeing have been taken into account .
for each bcm , we report in table 3 the results of the fits to the @xmath6 law . in column 2 and 3
we give the effective radius @xmath21 in kpc and the effective surface brightness @xmath22 in @xmath1 mag arcsec@xmath23 , respectively .
the quoted errors represent the 90% confidence levels given by the fitting procedure .
the surface brightness magnitudes were corrected for k dimming , galactic absorption and atmospheric extinction following schneider et al .
( 1983a ) and stark et al .
( 1992 ) , and also corrected for the @xmath24 cosmological dimming .
column 4 lists the obtained reduced @xmath25 .
column 5 and 6 give the rest
frame mean @xmath7 and @xmath26 colors , and column 7 the absolute magnitude @xmath27 computed within the 38 kpc radius .
generally , the de vaucouleurs law provides a satisfactory fit .
we do not think that our data justify any fit with a different , more generalized law like the @xmath28 proposed by sersic ( 1968 ) and recently adopted by caon et al .
( 1993 ) and graham et al .
( 1996 ) . from the radial light profiles presented in figure 2
, we can divide the sample bcms into three categories : ( a ) galaxies substantially well described by de vaucouleurs law over the whole range of data points ; ( b ) galaxies showing what can be termed a tidal cut off ; and , ( c ) galaxies showing a brightness excess over de vaucouleurs law .
we would place in category ( a ) ( which we will not further discuss ) ms0002 + 1556 , ms0013 + 1558 , ms0102 + 3255 , a180 , a401 , ms0301 + 1516 , a671 , a733 , and a1911 .
ms0037 + 2917(a77 ) , a279 , a399 , ms1201 + 2824 and ms1558 + 3321(a2145 ) fall into category ( b ) , and finally a84 , a150 , and ms0904 + 1651 ( a744 ) are good candidates for the presence of an extended halo ( category ( c ) ) .
+ the profile of _ ms0037 + 2917(a77 ) _ mimicks a tidal cutoff beyond @xmath29 kpc .
the apparent size of this galaxy is rather large with respect to the ccd field of view and another bright s0 galaxy is present south of the bcms .
therefore , there is the possibility that the curvature in the profile ( in any case , within 1 @xmath30 of the de vaucouleurs profile ) might be due to an overestimate of the sky background rather than to a real physical effect .
+ in _ a279 _ , a tidal cut
off could be present outward of @xmath31 kpc at a very low significance level . furthermore
, this occurs beyond the radius where systematic errors are important .
+ similarly to a77 , _
a399 _ shows a surface brightness deficit at large radii . again , because of the large apparent size of this galaxy with respect to the ccd size and of a removed object along the galaxy major axis , we can not confidently state that this deviation from de vaucouleurs law is evidence of a tidal cutoff .
+ in the case of _
ms1201 + 2824 _ , the impression of a surface brightness deficit at large radii is probably due to the somewhat anomalous behavior of the light profile with respect to the de vaucouleurs law .
+ as for _ ms1558 + 3321 ( a2145 ) _ , the situation is similar to that of a77 and a399 , yet with even less statistical significance .
+ of the three bcms showing some evidence of an extended halo , _ a84 _ and _ ms0904 + 1651(a744 ) _ show only scanty statistical significance ( the extrapolation of the best fit with de vaucouleurs law falls within the 1 @xmath30 error bars of the data points ) even if for a744 we were unable to trace the light profile in the rest frame to a magnitude fainter than @xmath32 mag arcsec@xmath23 . + the case of _
_ is different , since this is also the most peculiar bcm in our sample .
the light profile shows a statistically significant excess outward of @xmath33 kpc .
figure 3a shows the residuals in the bi dimensional image once a model galaxy with parameters equal to the best fitting de vaucouleurs is subtracted . as shown in the figure
, this excess does not correspond to that of a classical halo , since it appears essentially along the major axis .
other peculiarity of this galaxy is the presence of a double nucleus as can be seen in figure 3b .
+ the analysis of the surface brightness profiles of our bcm sample shows that de vaucouleurs law provides an excellent description of the surface brightness profiles at radii between 10 and 45 kpc , where deviations are never greater than 0.05 mag . at larger radii
, deviations generally remain of the order of the systematic errors , except in the case of a150 , which has a rather complex morphology .
+ @xmath7 and @xmath26 color profiles for all the bcms have been computed from the surface brightness profiles in the three filters and are shown in figure 4 .
the method we employed to assess the presence of color gradients takes into account the effect of incorrect estimates of the sky background in one or both of the images ( see andreon et al . ( 1995 ) for a more detailed and complete explanation of the method ) .
the solid lines in figure 4 represent the expected color profiles in the absence of color gradients .
vertical arrows indicate the position of superposed objects . except in the case of a150 ,
our plots are consistent with the absence of color gradients .
this fact implies that color gradients are generally not a characteristic feature of bcms , irrespective of the cluster selection criterion .
+ let us now consider the @xmath7 radial profile of a150 .
the delicate point of the adopted method is that a normalization procedure is required . whenever the data points scatter around the expected no color gradient line , the concept of a normalization radius becomes irrelevant since no color gradient would be meaningfull regardless of the radius chosen . in the case of a150 , if we normalize the expected value for the @xmath7 color profile at a radius smaller than that at which we found excess light with respect to the @xmath6 law , we would be led to state that the excess light is redder than that of the rest of the galaxy . on the other hand
, if the expected color profile is normalized at a radius where the excess is present , the interpretation would be that the excess and the central region of a150 share the same @xmath7 color while at intermediate radii the galaxy is bluer .
we have no reasons to prefer one of the two interpretations of this color profile ; our data do tell us , however , that a150 is the only bcm in the sample for which the @xmath7 color is not the same everywhere .
our sample includes seven galaxies which had been already studied by other authors : a150 and a401 were studied by malumuth & kirshner ( 1985 ) in the @xmath34 filter , a671 is also in the graham et al .
( 1996 ) sample observed in the kron
cousins @xmath35 filter and ms0037 + 2917 ( a77 ) , a279 , a399 , a401 , a671 and a733 are also in the sample studied by schneider et al .
( 1983b ) and hoessel & schneider ( 1985 ) in the same gunn @xmath1 filter .
+ figure 5 compares our profiles in the @xmath1 band with those obtained by malumuth & kirshner ( 1985 ) for a150 and a401 . in the case of a150 ,
the agreement in surface brightness levels with malumuth & kirshner s data is excellent once the differences in filter bandpasses are taken into account .
however , our fitting procedure yields a scale size @xmath36 smaller ( and a correspondingly brighter effective surface brightness ) , probably because of the different range in radius over which the two fits are carried out . in the case of
a401 we determined an effective radius @xmath37 larger than malumuth & kirshner .
porter et al .
( 1991 ) already compared the surface brightness profile they obtained for a401 with malumuth & kirshner s and showed that they derived a steeper profile and a higher surface brightness for the central regions .
our profile has a slope similar to that of malumuth & kirshner and also a value of the central surface brightness similar to that of porter et al.s .
we should mention that the profile we derived for a671 is in very good agreement with graham et al.s ( 1996 ) as well as are the parameters obtained with the @xmath6 law .
+ for the other commonly studied bcms , a comparison with hoessel & schneider s ( 1985 ) results shows differences in the effective radii of the order of @xmath38 to @xmath39 .
several causes could explain these differences : the way seeing effects are taken into account , the outermost radius to which the fit is applied , the way eccentricity is taken into account .
however , the relation between effective radius and surface brightness of bcms found by hoessel & schneider ( 1985 ) is also a very good fit to the whole of our data .
we conclude that a detailed , object to object comparison is rather meaningless because it depends on the data quality , and on the reduction and analysis techniques adopted
. however , the statistical properties of the samples are much more robust and reproducible regardless of the authors , therefore these are really representative of the samples . in table 3 , columns 5 and 6 give the rest
frame @xmath7 and @xmath26 colors , and column 7 the absolute magnitude @xmath27 , all computed within a radius @xmath40 kpc .
+ the average rest
frame @xmath26 color of the bcms in our sample is @xmath41 with a sample standard deviation of 0.05 .
this value agrees with the one given by schneider et al .
( 1983a ) and the dispersion around the mean is of the same order as the photometric accuracy obtained for this color ( gar96 ) . in this case
the mean and the median values do coincide .
+ the average rest
frame @xmath7 color is @xmath42 with a sample standard deviation of 0.09 .
once the shift in the @xmath0 filter of our photometric system with respect to that of schneider et al .
( 1983a ) ( gar96 ) is taken into account , also this value is compatible with the bcm rest frame colors of those authors .
the standard deviation is however about a factor of 2 higher than expected on the basis of the photometric errors .
if we use the k corrections given by frei & gunn ( 1994 ) for e galaxies , all galaxies have slightly redder rest frame colors with an average @xmath43 and a sample standard deviation of 0.07 . although the standard deviation remains higher than expected , it seems difficult to claim that bcms show an intrinsic dispersion in this color .
we recall that also in andreon et al.s ( 1995 ) bcm sample , no intrinsic dispersion in the @xmath44 color was found .
andreon et al .
( 1995 ) showed that a correlation existed between the bcm effective radius and the cluster richness . in that paper , abell s richness class was used as a measure of the environment galaxy density .
for the present sample , we can not use abell richness class as this is not available for the x ray selected clusters .
however , we can correlate bcm effective radii with the local galaxy density .
the cluster complete catalogs of gar96 allow us to count galaxies brighter than @xmath45 in circles of projected radii equal to 200 kpc at the cluster redshift and centered on the bcm itself .
plotted against the effective radius ( figure 6 ) , this measure of the local galaxy density confirms and quantifies the result of andreon et al .
shallower bcms are found in environments , which , at the present epoch , are locally denser .
+ sandage & perelmuter ( 1990 ) suggest that bcms can be used to test whether the universe is expanding or not .
to check the cosmological dimming , they looked for a redshift dependence of the galaxy surface brightness at a fixed metric radius .
+ for the galaxies studied by sandage & perelmuter ( 1990 ) , the observed scatter in the relation between @xmath46 and @xmath47 , i.e. the scatter in the cosmological dimming , once the malmquist bias ( malmquist 1920 ) is removed ( see sandage & perelmuter 1990 for details ) , is @xmath48 mag arcsec@xmath23 . for our bcm sample , in the range @xmath49 , we find a corresponding scatter of 0.22 mag arcsec@xmath23 .
the slope of the linear regression between @xmath46 and @xmath47 , before malmquist corrections , is 3 times larger than the tolman signal for sandage & perelmuter s ( 1990 ) samples , yet for our sample is of the same amplitude as the tolman signal .
this suggests that our sample is only slightly affected by the malmquist bias .
this fact is probably due to the different sample compositions , and , most probably , to the fact that in our sample about half of the bcms are there because they are members of clusters serendipitously detected in an x ray survey .
although the cluster x - ray luminosity is correlated with the bcm absolute magnitude ( valentijn & bijleveld 1983 ) , their surface brightness at a fixed metric radius is only weakly dependent on it , implying that through x ray selection the amount of bias introduced on the quantity to test as redshift increases is quite small .
we suggest that the tolman test could give significant results if carried out on a proper sized sample of bcms drawn from purely x - ray selected clusters .
unfortunately , the number of bcms in x ray selected clusters in our sample is still too small and extends to relatively low redshift values for the tolman test to be successfull .
the multicolor surface brightness photometry of this mixed sample of 17 bcms with redshifts extending to @xmath50 shows that : * the surface brightness profile of these bcms is generally well described by de vaucouleurs law out to radii of @xmath5 kpc .
structural peculiarities such as changes in isophotal position angle and/or ellipticity are largely absent in our sample . *
although normal and small size ellipticals do show color gradients ( sparks & jrgensen ( 1993 ) ) , these gradients are a rather rare feature in bcms .
the only color gradient detected in our sample corresponds to a150 .
this galaxy also shows a brightness excess with respect to the @xmath6 law along the major axis and a double nucleus .
andreon et al . (
1995 ) , who analyzed color profiles in the same way as we did , found that 2 bcms in a sample of 9 show gradients in the @xmath51 color ( and 2 out of 8 bcms in the @xmath44 color ) . we could thus infer that @xmath52 of bcms presents measurable color gradients in broad
bandpasses redwards of the 4000 break .
* bcm @xmath7 and @xmath26 colors ( integrated within 38 kpc radius ) do not show any intrinsic dispersion .
if an intrinsic dispersion is present in the bluer color , this is of the same order as the uncertainties in the k corrections .
this _ normality _ of bcms which tends to assimilate them to the more general class of elliptical galaxies does not remove their main peculiarity : the large scale size of many of them ( about half of the bcms we studied have effective radii @xmath53 kpc ) .
galaxies so large and bright ( @xmath54 ) are found only in clusters of galaxies .
furthermore , we have shown that their linear dimensions scale with the local galaxy density as determined at the present epoch .
the sizes do not seem to be related either to the cluster morphology , or to the way clusters are selected .
morgan type i clusters possess bcms among the less extended ( a180 , with an effective radius of only 22 kpc ) and among the most extended ones in our sample ( a401 , @xmath55 kpc ) .
it is true that only one of the five x ray selected clusters ( i.e. , clusters which have been missed in the abell catalog ) has a bcm with an effective radius larger than 40 kpc ( ms1201 + 2824 , @xmath56 kpc ) , but this is naturally explained by a selection effect : these clusters are not found in the abell catalog because they are not very rich , and therefore their bcms are not very extended . nor is the x ray luminosity of the clusters correlated with the size of the bcm galaxy .
+ merging is considered the most likely mechanism for the formation of these giant galaxies
. it also provides a way to dilute any color gradients ( white 1980 ) ( which we do not generally observe ) .
two different cd formation scenarios have been investigated : either cds are the product of repeated mergers after cluster virialization ( malumuth 1992 ; bode et al . 1994 ) , or the merging events leading to the cd formation occurred before or during cluster formation ( merritt 1988 ) .
there is evidence of merging events occuring in clusters at the present epoch , as shown by the peculiar e galaxy ic1182 in the hercules cluster which is the result of a recent interaction between an elliptical and a spiral galaxy ( tarenghi & maccagni , private communication ) .
generally , simulations consider the luminosity of the merger products and , if this is a few times @xmath57 , the resulting galaxy is identified as a cd , i.e. the most luminous galaxy at the center of the cluster .
dynamical arguments lead us to consider the cluster as a whole , that is to say , the entire cluster galaxy population comes into play , mainly since it is largely concentrated within the core radius .
furthermore , it is sometimes assumed that the large extent of the cd galaxy is mainly due to the presence of a halo .
the relation we found is between _ de vaucouleurs effective radius and the galaxy number density within a 200 kpc radius from the bcm itself_. it is therefore a relation between a quantity which measures the slope of the surface brightness profile , independently of the presence of an extended halo ( which does not seem to be a common feature in our sample ) , and a quantity which measures the density of the environment as it is now .
both quantities are presumably the end points of evolutionary processes , which might not necessarily occur with the same time scales
. the implications of malumuth s ( 1992 ) simulations for cds , provided the dynamical friction in the real universe is as efficient as in his code , are , among others , that cds are a relatively recent phenomenon or that they are not formed in their present environment .
our data would tend to exclude this last possibility , and , because of the generally dull appearance , do probably suggest that some time has elapsed since the last dramatic event .
this seems to be in contrast with the high peculiar velocities sometimes shown by cds ( hill et al .
1988 , sharples , ellis & gray 1988 , zabludoff , huchra & geller 1990 , malumuth et al .
however , substructure is a common feature of many clusters ( geller & beers 1982 , jones & forman 1984 , bird 1994 ) , and when it is taken into account in the dynamical analysis of the clusters , the peculiar velocities of the central galaxies tend to disappear ( bird 1994 ) . our data thus seem to suggest that the growth in size of the bcms is governed by the _ local _ environment and that , in many cases , the dominant galaxy and the concentration of galaxies surrounding it represent , at least for a non negligible amount of time , a dynamical entity with respect to the whole cluster
. it would be extremely interesting to probe the dynamical status of this sample with respect to the very local and to the overall cluster environment since this would be a powerful diagnostic tool for the formation scenarios .
the authors warmly acknowledge the support of the oan staff both in san pedro martir and ensenada .
the time allocation committee of the unam
oan is thanked for the generous allocation of telescope time . s.a .
acknowledges the award of a cnr fellowship .
we are grateful to the referee , e.m .
malumuth , for his useful suggestions which have helped us to improve the contents and presentation of the paper .
abell , g.o . , corwin , h.g .
, & olowin , r.p .
1989 , , 70 , 1 andreon , s. , garilli , b. , maccagni , d. , gregorini , l. , & vettolani , g. 1992 , , 266 , 139 andreon , s. , garilli , b. , & maccagni , d. 1995 , , 300 , 711 bautz , l.p . , & morgan , w.w . 1970 , , 162 , l149 jones , c. & forman , w. 1984 , , 276 , 38 bird , c.m .
1994 , , 107 , 1637 bode , p.w . , berrington , r.c . ,
cohn , h.n . , & lugger , p.m. 1994 , , 433 , 479 caon , n. , capaccioli , m. , & donofrio , m. 1993 , , 265 , 1013 de vaucouleurs , g. 1948 , ann .
dap . , 11 , 247 geller , m. & beers , t.c .
1982 , , 94 , 421 frei , z. , & gunn , j.e .
1994 , , 108 , 1476 garilli , b. , bottini , d. , maccagni , d. , carrasco , l. , & recillas , e. 1996 , , 105 , 191 [ gar96 ] gioia , i. , et al .
1990 , , 72 , 567 graham , a. , lauer , t.d . , colless , m. , & postman , m. 1996 , , 465 , 534 hill , j.m . ,
hintzen , p. , oegerle , w.r .
, romanishin , w. , lesser , m.p . ,
eisenhamer , j.d . , & batuski , d.j .
1988 , , 332 , l23 hoessel , j. , & schneider , d. 1985 , , 90 , 1648 malmquist , g. 1920 , lund medd , ser ii , no .
22 malumuth , e.m .
1992 , , 386 , 420 malumuth , e.m . ,
kriss , g.a .
, dixon , w.v.d . ,
ferguson , h.c .
& ritchie , c. 1992 , , 104 , 495 malumuth , e.m .
, & kirshner , r.p .
1985 , , 291 , 8 merritt , d. 1988 , the minnesota lectures on clusters of galaxies and large scale structure , ed .
dickey ( provo : brigham young univ .
press ) , 175 porter , a.c . ,
schneider , d.p . , & hoessel , j.g .
1991 , , 101 , 1561 sandage , a. , & perelmuter , j. 1990 , , 370 , 455 schneider , d.p . ,
gunn , j.e .
, & hoessel , j.g .
1983a , , 264 , 337 schneider , d.p . , gunn , j.e . , & hoessel , j.g .
1983b , , 268 , 476 sersic , j .- l .
1968 , _ atlas de galaxies australes _ ( cordoba : obs .
astronomico ) sharples , r. , ellis , r. , & gray , p. 1988
, , 231 , 479 sparks , w. , & jrgensen , i. 1993 , , 105 , 1753 stark , a.a . , et al . 1992 , , 79 , 77 tolman , r.c . 1930 , proc .
, 16 , 511 valentijn , e.a . , & bijleveld , w. 1983 , , 125 , 223 white , s. 1980 , , 191 , 1p zabludoff , a.i .
, huchra , j.p . , & geller , m.j .
1990 , , 74 , 1 ms0002 + 1556 & 0.116 & & & ms0013 + 1558 & 0.083 & & & ms0037 + 2917 ( a77 ) & 0.072 & i & 1 & a84 & 0.103 & ii & 1 & ms0102 + 3255 & 0.080 & & & a150 & 0.060 & i ii & 1 & a180 & 0.135 & i & 0 & blue a279 & 0.080 & i ii & 1 & a399 & 0.072 & i ii & 1 & a401 & 0.075 & i & 2 & ms0301 + 1516 & 0.083 & & & red a671 & 0.049 & ii iii & 0 & a733 & 0.116 & i & 1 & ms0904 + 1651 ( a744 ) & 0.073 & ii & 1 & blue ms1201 + 2824 & 0.167 & & & a1911 & 0.191 & ii iii & 2 & ms1558 + 3321 ( a2145 ) & 0.088 & & 0 & ms0002 + 1556 & 1765 & 0.30 & 0.003 & 65 & 0.6 ms0013 + 1558 & 1058 & 0.20 & 0.006 & 38 & 1.2 ms0037 + 2917 ( a77 ) & 949 & 0.20 & 0.008 & 26 & a84 & 1341 & 0.18 & & 165 & ms0102 + 3255 & 1047 & 0.23 & 0.012 & 28 & a150 & 949 & 0.18 & 0.008 & 105 & a180 & 1644 & 0.22 & 0.005 & 103 & a279 & 1230 & 0.13 & 0.007 & 46 & a399 & 751 & 0.27 & 0.006 & 128 & a401 & 855 & 0.29 & 0.010 & 59 & ms0301 + 1516 & 1237 & 0.26 & 0.003 & 14 & a671 & 436 & 0.19 & 0.005 & 157 & a733 & 641 & 0.11 & & 265 & 0.8 ms0904 + 1651 ( a744 ) & 943 & 0.06 & & 185 & ms1201 + 2824 & 2244 & 0.30 & & 95 & 0.4 a1911 & 2241 & 0.30 & 0.001 & 76 & 0.3 ms1558 + 3321 ( a2145 ) & 1161 & 0.50 & 0.013 & 82 & ms0002 + 1556 & @xmath58 & @xmath59 & 0.17 & 0.35 & 0.29 & 23.65 ms0013 + 1558 & @xmath60 & @xmath61 & 1.70 & 0.39 & 0.28 & 23.72 ms0037 + 2917 ( a77 ) & @xmath62 & @xmath63 & 1.04 & 0.37 & 0.27 & 23.58 a84 & @xmath64 & @xmath65 & 0.16 & 0.34 & 0.29 & 23.83 ms0102 + 3255 & @xmath66 & @xmath67 & 0.53 & 0.30 & 0.28 & 23.28 a150 & @xmath68 & @xmath69 & 7.58 & 0.26 & 0.33 & 23.69 a180 & @xmath70 & @xmath71 & 0.12 & 0.29 & 0.30 & 23.59 a279 & @xmath72 & @xmath73 & 0.16 & 0.20 & 0.35 & 23.90 a399 & @xmath74 & @xmath75 & 0.67 & 0.20 & 0.27 & 23.94 a401 & @xmath76 & @xmath77 & 0.04 & 0.42 & 0.24 & 24.25 ms0301 + 1516 & @xmath78 & @xmath79 & 0.29 & 0.40 & 0.30 & 22.99 a671 & @xmath80 & @xmath81 & 3.25 & 0.33 & 0.13 & 24.06 a733 & @xmath82 & @xmath83 & 0.32 & 0.21 & 0.18 & 24.36 ms0904 + 1651 ( a744 ) & @xmath84 & @xmath85 & 0.81 & 0.21 & 0.29 & 23.71 ms1201 + 2824 & @xmath86 & @xmath87 & 1.04 & 0.25 & 0.24 & 24.10 a1911 & @xmath88 & @xmath89 & 0.48 & 0.29 & 0.23 & 24.38 ms1558 + 3321 ( a2145 ) & @xmath90 & @xmath91 & 0.50 & 0.31 & 0.31 & 23.62 | the gunn @xmath0 , @xmath1 and @xmath2 ccd images of a representative sample of 17 brightest cluster galaxies ( bcm ) have been analyzed in order to derive surface brightness and color profiles , together with geometrical parameters like eccentricity and position angle .
the sample includes both x ray and optically selected clusters , ranging in redshift from @xmath3 to @xmath4 .
we find that bcms are substantially well described by de vaucouleurs law out to radii of @xmath5 kpc , and that color gradients are generally absent . only in two cases
we find a surface brightness excess with respect to the @xmath6 law , which for a150 is coupled with a change in the @xmath7 color .
the rest frame colors of bcms do not show any intrinsic dispersion . by parametrizing the environment with the local galaxy number density
, we find that it is correlated with the bcm extension , i.e. bcms with larger effective radii are found in denser environments . |
astrometry is an essential tool for characterizing directly imaged exoplanets and their physical relationship to other elements of the planetary system in which they reside .
the gemini planet imager ( gpi ; * ? ? ?
* ) was designed with a goal of achieving @xmath21.8 mas astrometric accuracy @xcite , necessary for characterizing the eccentricity distribution of exoplanet orbits from the gpi exoplanet survey @xcite .
to do so , the astrometric calibration of gpi has continually been benchmarked to well - calibrated astrometric fields @xcite .
this had led to some of the most precise astrometry on directly - imaged exoplanet systems to date @xcite , allowing us to constrain or fit the first ever orbit of some of these directly - imaged exoplanets .
however , limited by either the signal to noise ratio ( snr ) of these exoplanets or by biases in the various data analysis algorithms , so far no astrometric study with gpi has achieved the design goal of 1.8 mas precision .
the importance of understanding planetary orbits is highlighted by the @xmath0 pictoris system , a young ( @xmath3 myr ; * ? ? ?
* ; * ? ? ?
* ) and nearby system that has been extensively studied .
@xmath0 pic harbors a near edge - on debris disk that was first imaged by @xcite and which was subsequently observed to have a warp thought to be induced by a planet whose orbit is inclined relative to the debris disk @xcite .
additional indirect signatures of a planet were derived from variable spectral features modeled as infalling comets @xcite and a peculiar light curve anomaly detected in 1981 @xcite .
@xcite then discovered @xmath0 pic b , a planet at an appropriate mass ( @xmath4 @xmath5 ) and semi - major axis ( @xmath6 au ) to be responsible for the previously observed indirect signatures of planets .
a key focus of subsequent observations was determining the alignment of @xmath0 pic b relative to the main outer disk and the warp to determine if @xmath0 pic b is causing the warp . by observing the disk and planet simultaneously
, @xcite concluded that the planet is misaligned from the main outer disk and consistent with being responsible for the warp .
additionally , @xcite ruled out the possibility of having another giant planet in the system massive enough to cause the warp instead .
thus , @xmath0 pic b is responsible for the warp in the debris disk .
this was confirmed in astrometric monitoring campaigns @xcite which used homogeneous datasets to limit systematics and constrain the orbit of @xmath0 pic b. refining the orbital elements of @xmath0 pic b is not only crucial for investigating the dynamical link between the planet and the disk warp , but also because @xmath0 pic b may transit its host star once every @xmath720 years . to date , there are no other known systems where the physical properties of an exoplanet can be characterized by using both the direct imaging and transit techniques . currently , the tightest constraint on the probability of transit is @xmath70.06% , obtained with a dedicated astrometric monitoring campaign with gpi @xcite .
however , @xcite point out that gpi measurements from @xcite have a higher position angle than astrometry from previous measurements , which could arise from a possible systematic calibration offset between gpi and previous instruments instead of actually due to @xmath0 pic b s orbit being slightly inclined away from edge on .
we note there currently is no evidence of a position angle offset with the gpi astrometric calibration , and the gpi astrometry of hd 95086 b are consistent with previous astrometry from other instruments @xcite .
however , it is important to more accurately compute the orbit of @xmath0 pic b because in late 2017 to early 2018 @xcite , it will be at its closest projected separation from the star .
the transit of the planet and/or any circumplanetary material orbiting around it could therefore be detectable .
one of the obstacles in characterizing directly imaged exoplanets is that even with the newest instrumentation , the glare of the host star covers the signal of the planet . in order to subtract the point spread function ( psf ) of the star and
maximize the snr of the planet , observing techniques such as angular differential imaging ( adi ; * ? ? ?
* ) and spectral differential imaging ( sdi ; * ? ? ?
* ) and data reduction algorithms like karhunen - love image projection ( klip ; * ? ? ?
* ; * ? ? ?
* ) are used in combination to disentangle the psf of the star from potential astrophysical sources .
however , these techniques distort the planet signal and create data reduction artifacts , which usually are nuisance parameters that need to be calibrated out to obtain unbiased astrometry .
forward modelling effects of observing techniques and data reduction algorithms on the psf of the planet was first done in the context of adi and loci , showing significant improvements in astrometry and photometry for simulated planets @xcite . in similar contexts with adi and loci , @xcite and @xcite used forward modelling to correct for flux loss of exoplanets and the flux and morphology of disks respectively . for classical adi ( cadi )
, @xcite showed that forward modelling techniques can improve the sensitivity of cadi and mitigate biases .
the use of markov - chain monte carlo ( mcmc ) in conjunction with forward modelling was presented in @xcite for reference differential imaging .
recently , @xcite introduced a method , called klip - fm , to analytically forward model the degradation of a faint astrophysical signal that occurs when using least squares - based psf subtraction algorithms such as klip that is also generally applicable to any observing strategy .
additionally , the computation of the forward model with klip - fm is much quicker than negative fake planet injection methods @xcite , as the stellar psf subtraction algorithm , klip , needs only to be run once . in this paper
, we demonstrate the advantages of klip - fm for precise astrometry and constraining planetary orbits by applying it to gpi observations of @xmath0 pic b reduced using klip and adi+sdi . in section [ sec : obs ] , we describe our new astrometry technique , bayesian klip - fm astrometry ( bka ) , in which we forward model the psf of the planet with klip - fm and then use the forward model in a bayesian framework to measure the position of the planet while also modelling the correlated nature of the noise . in section [ sec : valid ] , we validate our technique by fitting an orbit to the data and showing that our astrometry and uncertainties are consistent with keplerian motion with no obvious systematics .
finally , in section [ sec : orbit ] , we apply our new astrometry to constrain the orbit of @xmath0 pic b and place the tightest constraints yet on the transit of the planet and its hill sphere .
to obtain a large temporal baseline of gpi astrometric points , we compiled gpi data of @xmath0 pic from commissioning ( gemini program gs - eng - gpi - com ) , an astrometric monitoring campaign of @xmath0 pic b ( gemini programs gs-2015a - q-21 and gs-2015b - q-9 ; pi : graham ) , the gpi exoplanet survey ( gs-2014b - q-500 ; pi : macintosh ) , and a gemini large and long program to characterize debris disks ( gs-2015b - lp-6 : pi : chen ) .
all the data used in the following analysis are listed in table [ table : obs ] .
most of these data were published and analyzed in a previous study characterizing the orbit of @xmath0 pic b by @xcite .
we have reprocessed those data with the new astrometry methods presented in this paper and combined them with five additional new epochs .
there were three datasets in @xcite that we did not use .
we did not use the polarimetry dataset as there was no instrumental psf obtained with the data to forward model .
the dataset on 2014 march 23 was taken during tests of the adaptive optics ( ao ) system , causing the instrumental psf to be highly varying and making it unsuitable for forward modelling .
the 2015 - 01 - 24 dataset contained artifacts in the construction of the forward model that we could not remove .
we chose not to include the measured astrometry from @xcite for the datasets we omitted in order to maintain homogeneity in astrometric datasets and reduce potential systematic errors .
we also note that in the last dataset taken on 2016 - 01 - 21 , there was saturation on the edge of the occulting mask due to bad seeing .
this affected the forward modelling of @xmath0 pic b , which was also near the occulting mask so a large portion of the frames taken could not be used in this analysis .
clcccc|cccc 2013 nov 16 & gs - eng - gpi - com & _ k1 _ & 1789 & 26 & 1.09 & 7 & 4 & no & 13 + 2013 nov 16 & gs - eng - gpi - com & _ k2 _ & 1253 & 18 & 0.93 & 7 & 4 & no & 13 + 2013 nov 18 & gs - eng - gpi - com & _ h _ & 2446 & 32 & 0.68 & 7 & 4 & no & 13 + 2013 dec 10 & gs - eng - gpi - com & _ h _ & 1312 & 39 & 0.77 & 7 & 4 & no & 13 + 2013 dec 10 & gs - eng - gpi - com & _ j _ & 1597 & 19 & 0.70 & 7 & 4 & no & 13 + 2013 dec 11 & gs - eng - gpi - com & _ h _ & 410 & 65 & 0.46 & 7 & 4 & no & 11 + 2014 nov 8 & gs - eng - gpi - com & _ h _ & 2147 & 25 & 0.77 & 7 & 4 & no & 11 + 2015 apr 2 & gs-2015a - q-21 & _ h _ & 1312 & 10 & 0.51 & 2 & 2 & no & 11 + 2015 nov 6 & gs-2014b - q-500 & _ h _ & 2207 & 28 & - & 7 & 4 & no & 11 + 2015 dec 5 & gs-2015b - lp-6 & _ j _ & 4948 & 66 & 0.92 & 7 & 4 & yes & 11 + 2015 dec 22 & gs-2015b - q-9 & _ h _ & 2088 & 19 & 0.76 & 7 & 4 & no & 11 + 2016 jan 21 & gs-2015b - q-9 & _ h _ & 954 & 17 & 1.18 & 10 & 2 & yes & 7 + the raw integral field spectrograph ( ifs ) data were reduced to form 3-d spectral data cubes using the gpi data reduction pipeline ( drp ) version 1.2.1 or 1.3 @xcite .
there were no significant changes between the two versions of the gpi drp that impacted the astrometry of @xmath0 pic b. we used the same data reduction process as in @xcite and will summarize them here .
first , dark subtraction and bad pixel correction were applied to each 2-d image . for earlier datasets
in which cryocooler vibration induced correlated noise on the detector , the frames were `` destriped '' to remove this noise @xcite .
then , we corrected for instrument flexure using an argon arc lamp taken before each sequence to align each individual spectrum for extraction @xcite .
then each 2-d frame was turned into a spectral data cube , corrected for any remaining bad pixels , and corrected for distortion @xcite . for the _
k_-band data , thermal background frames were taken along the sequence .
we constructed thermal background cubes in the same fashion and subtracted them from the _ k_-band data .
to spatially register and photometrically calibrate our data , we used the gpi drp to measure the flux and location of the `` satellite '' spots , which are centered on the occulted star and imprinted with its attenuated spectrum @xcite .
the satellite spot fluxes were used to derive a flux calibration in each spectral channel , which we used when constructing the forward models in section [ sec : fm ] .
the location of the occulted star at each wavelength in each datacube was found using a least squares fit to all of the satellite spots positions and the magnitude of the atmospheric differential refraction .
the occulted star center is used to align all the images together before psf subtraction , and is crucial for determining the astrometry of @xmath0 pic b relative to its host star .
the precision on the star center is 0.05 pixels ( 0.7 mas ) for satellite spots with signal - to - noise ratios ( snr ) @xmath8 20 @xcite , which is certainly the case for all of our data on bright stars like @xmath0 pic . to subtract the stellar psf from each individual datacube , we used ` pyklip ` @xcite , a python implementation of the klip algorithm .
we used both adi and sdi to decorrelate the stellar psf from the psf of @xmath0 pic b. as we were only concerned with @xmath0 pic b , we applied our psf subtraction only on the annulus that included the planet , rather than the full image .
we adjusted three main parameters for the psf subtraction , depending on the dataset .
the first was the number of modes used from the karhunen - love ( kl ) transform to model the stellar psf .
the second was an exclusion criteria for picking reference psfs .
the exclusion criteria is similar to the quantity @xmath9 in @xcite and is defined by the number of pixels that @xmath0 pic b would move azimuthally and radially in an observing sequence due to adi and sdi .
thirdly , we toggled an @xmath711 pixel wide spatial high - pass filter that was applied to some datasets before psf subtraction .
the high - pass filter was implemented using a gaussian filter in fourier space and the @xmath711 pixel size was chosen to remove low frequency background without significantly distorting any point sources in the image . in general , increasing the number of kl modes , decreasing the exclusion criteria , and applying a high - pass filter improves the subtraction of the stellar psf .
however , taken to the extreme , all three options attenuate signal from the planet .
additionally , for a planet as bright as @xmath0 pic b , the forward modelling described in section [ sec : fm ] may not be valid when the psf subtraction becomes too aggressive @xcite .
thus , to optimize the signal of the planet while maintaining the validity of the forward modelling , we varied these parameters for each dataset .
as @xmath0 pic b is bright , in most of the datasets we used an exclusion criteria of 4 pixels , which is slightly greater than 1 @xmath10 , where @xmath11 is the wavelength and @xmath12 is the diameter of the telescope . when there was little field rotation for adi , we decreased the exclusion criteria to not overly restrict our reference psfs , but also decreased the number of kl modes used to avoid being too aggressive
when the speckle noise was too bright due to observing at a shorter wavelength or when the planet moved closer to the star , we applied a high pass filter to remove some of the diffracted starlight .
we list the chosen parameters for each dataset in table [ table : obs ] .
after psf subtraction , all images were rotated so that north is up and east is left .
then , all the frames were mean combined in both the spectral and temporal directions , resulting in a single psf - subtracted frame for each dataset .
after stellar psf subtraction , the psf of the planet is distorted by over - subtraction , caused by the presence of the planet in the data we are subtracting , and self - subtraction , caused by the presence of the planet in the reference images . over and self - subtraction perturbs the astrometric and photometric properties of the planet s psf and prevents a straightforward measurement of its position .
typically , the biases and uncertainties in the astrometry caused by stellar psf subtraction are estimated by injecting fake planets into the data at other position angles and comparing the retrieved position to the injected position .
however , the over - subtraction and self - subtraction that distorts the planet psf is deterministically caused by the existence of a planet in the data and its apparent motion in the reference images induced by adi and sdi . in turn , these features , if modeled , can inform us about the location of the planet and improve our astrometric precision and accuracy .
recently , @xcite introduced klip - fm , an analytic framework to compute the effect of a planet on the kl modes and use these perturbations to reconstruct the over - subtraction and self - subtraction features . using this technique , we are able to generate the psf of the planet after psf subtraction ( see appendix [ appendix : fmpsf ] for a detailed description of the procedure ) .
briefly , we use the psf of the planet before psf subtraction , the apparent motion of the planet due to adi and sdi , and a model of the planet spectrum to compute the distorted psf of the planet after psf subtraction .
the result is a 2-d broadband planet psf , which we call @xmath13 , centered at ( @xmath14 , @xmath15 ) , our initial estimate for the location of the planet . for our gpi data ,
the forward models were generated using the implementation of klip - fm in ` pyklip ` .
for each dataset , we used the same parameters as our psf subtraction to construct the forward model . to construct a model of the instrumental psf at each wavelength
, we used the average psf of the satellite spots across all images for that wavelength .
we used a basic center of light centroiding routine to measuring an approximate @xmath16 and @xmath17 positions of the planet ( @xmath14 and @xmath15 ) in the psf subtracted image .
this initial estimate for the position is good to within a pixel for our gpi data . for the input planet spectrum of the forward model
, we used the normalized best - fit model spectrum from @xcite , which has an effective temperature of 1650 k and a @xmath18 ( cgs units ) .
we scaled the spectrum by eye to approximately match the contrast of @xmath0 pic b in our data and used the satellite spot fluxes to convert from contrast to digital numbers .
we found that varying the spectrum and photometry had negligible impacts on our measured astrometry .
even a spectrum differing by @xmath725% in shape changed the astrometry by @xmath19 mas , significantly smaller than the uncertainties we find in the following analysis . with this input information , we used ` pyklip ` to generate one forward modeled psf , @xmath13 , for each dataset . in figure
[ fig : data - fm - compare ] , we show an example forward modeled psf ( not optmized to fit the data ) and comparison to data for our 2014 november 18 dataset on @xmath0 pic b. qualitatively , the forward model matches all the features seen in data , including the prominent negative self - subtraction lobes on either side of the planet . as ` pyklip ` parallelizes the computation , the generation of the forward models is quick . on a 32-core machine with amd
opteron 6378 processors clocked at 2.3 ghz , forward models for all 37 channels of a representative 37-cube gpi dataset were generated in 15 minutes : 4 minutes of overheads for preprocessing and generating the instrumental psfs and 11 minutes to execute klip - fm and create the forward model .
we note that we chose to run the forward model on a large annulus to examine noise properties in the data and that the computation time for klip - fm decreases @xmath720 - 30% if a small sector around the planet was used instead .
to use the forward modeled psf , @xmath13 , to perform astrometry , we developed a bayesian framework to fit @xmath13 to the data , account for correlated noise , and estimate our fitting uncertainties .
first , ignoring the correlated noise , we can use three parameters to fit the forward model to the data : the location on the planet in @xmath16 ( @xmath20 ) , the location of the planet in @xmath17 ( @xmath21 ) , and a scale factor ( @xmath22 ) to match the flux of @xmath13 to the data .
we can then write the posterior probability for @xmath20 , @xmath21 , and @xmath22 given the data @xmath12 using bayes theorem as @xmath23 ignoring normalization constants .
the first term on the right - hand side is the likelihood and the second is the prior . to construct the likelihood
, we must first use our input parameters and @xmath13 to generate a model to compare to the data .
we scale @xmath13 by @xmath22 and recenter it from its guessed location ( @xmath14 , @xmath15 ) to ( @xmath20 , @xmath21 ) . the residual , @xmath24 , between the model and the data , @xmath12 , in fitting region @xmath25 is calculated by @xmath26 where fitting region @xmath25 is a fixed rectangular box centered at the approximate location of the planet in the data .
we pick the size of the fitting region to be a few @xmath10 to encompass the psf and the self - subtraction wings .
we varied the fitting box size for each of our datasets in order to keep the fit focused on the area where signal from the planet can be seen .
we list the size of the fitting box for each dataset in table [ table : obs ] .
note that we do not fit any background term to the image , as one of the first steps of klip is to subtract off the mean of the image , removing any spatially constant background . for data with uncorrelated errors ,
the log of the likelihood function of the data for a particular model is @xmath27 , \end{aligned}\ ] ] where @xmath28 is the uncertainty in pixel i and @xmath29 is the number of pixels in the fitting region .
however , the assumption of uncorrelated errors does not hold for images limited by speckle noise . except in the cases of very aggressive psf subtraction or at much greater separations from the star , the residual noise after psf subtraction in gpi data of a bright star like @xmath0 pic
is still dominated by correlated speckle noise , which has a correlation scale that depends on the aggressiveness of the psf subtraction . due to the bright nature of @xmath0 pic b , we could not use a very aggressive reduction to ensure that the planet remained a perturbation on our kl modes , thereby preserving the validity of the analytical forward modelling technique .
the conservative psf subtraction combined with the close separation of @xmath0 pic b from its bright host star required us to capture the correlated nature of the noise .
we thus write our likelihood function instead as @xmath30 where @xmath31 is the covariance matrix of size @xmath32 and @xmath24 is a @xmath33 matrix .
we applied a gaussian process framework to characterize the covariance in the noise ( see @xcite for an in - depth explanation of the application of gaussian processes to astronomical data ) .
we only aim to model the correlations within each individual speckle , which spans @xmath10 in spatial extent . while in a single unprocessed frame speckle noise
also has addition correlations on much larger scales , there are no significant correlations between speckles at larger spatial scales in our small fitting region due to stellar psf subtraction with klip and averaging uncorrelated speckles together when collapsing the frames in our adi+sdi sequence .
thus , within our fitting region , the dominant correlation in our noise is from pixels within a single speckle , and thus is the one correlation we modelled .
following the procedure in @xcite for fitting 1-d correlations in stellar spectra , we used the matrn covariance function parametrized with @xmath34 to model the correlated speckle noise .
we chose the matrn function with @xmath34 as it better fits the correlations at larger separations compared to a simple squared exponential relation .
however , we note that switching between covariance functions that have similar shapes does not significantly alter the error bars .
we also chose to assume symmetric noise as we did not find any difference in the correlation scale of our noise in the radial and azimuthal directions .
this is likely be due to the fact that we used both adi and sdi to model the stellar psf and thus are better at subtracting speckle noise , which before psf subtraction is more correlated radially than azimuthally due to the finite spectral bandwidth of the data . in instruments that are not able to utilize sdi , it might be necessary to separate the noise into radial and azimuthal components , each with its own characteristic correlation length . thus for our purposes , we chose the symmetric matrn covariance function with @xmath34 , which defines the covariance between two pixels @xmath35 and @xmath36 separated by a distance of @xmath37 as @xmath38 where @xmath28 and @xmath39 are the uncertainties for each pixel and @xmath40 is a characteristic correlation length scale that increases when noise is correlated at larger spatial scales .
we calculated the uncertainty for each pixel by computing the standard deviation of pixels in an annulus with a width of @xmath41 ( 6 pixels ) , centered on the star , and with a mean radius equal to the distance between that pixel and the star .
any pixels containing signal from the planet were masked and not used in estimating the noise in the annulus .
the correlation length @xmath40 is not fixed , but rather kept as a hyperparameter parameter in our bayesian framework that we will marginalize over in the end .
thus the final posterior we are trying to calculate is really @xmath42 where @xmath22 and @xmath40 are both hyperparameters as we are only interested in the astrometry . compared to our likelihood function ,
our prior is relatively straightforward .
we allow for uniform priors in @xmath16 and @xmath17 within 2 pixels from our initial guess location .
similarly , we allow for an uniform prior between 0 and 5 for @xmath22 to determine how much to scale @xmath13 , which was already scaled to an approximate contrast of @xmath0 pic b. the correlation length @xmath40 has a uniform prior between 0 and 10 pixels , which provides an ample range to explore the correlations within individual speckles of size @xmath43 pixels .
we used the @xcite affine invariant mcmc sampler implemented in the ` emcee ` python package @xcite to sample the posterior distribution and custom ` cython ` code to quickly generate the covariance matrix as we vary @xmath40 .
the mcmc sampler was run for 800 steps using 100 walkers , with a `` burn - in '' of 200 steps beforehand . in figure
[ fig : astrom - tri ] , we show the posterior distributions from the mcmc fit for the 2014 november 18 _ h_-band dataset as a representative posterior distribution .
the value of @xmath40 is close to @xmath44 pixels , indicating we are accurately fitting the correlated speckle noise .
we use the resulting posterior distributions to calculate the most - likely values and uncertainties for the location of @xmath0 pic b in our image at each epoch . to convert our results to more useful physical values , we need to convert our measured location of @xmath0 pic b to its positional offset from its host star in right ascension ( ra ) and declination ( dec ) . as our images
are already rotated so that @xmath45 is positive ra and @xmath46 is positive dec , it is straightforward to convert from image fates to sky coordinates .
we use the satellite spots to measure the location of the star behind the coronagraph , which has a precision of 0.7 mas @xcite .
this allows us to derive the separation of the planet from the star in pixels . to convert from pixel separations to physical separations in ra and dec
, we use the most - recent astrometric calibration numbers from @xcite : a plate scale of @xmath47 mas lenslet@xmath48 and a residual north angle offset of @xmath49 from the north angle value used in the gpi drp ( versions 1.2.1 to 1.3 , the current version ) .
these astrometric calibration numbers show no significant change over time , so we apply them to all our epochs of data .
then , we assume all of these error terms are uncorrelated and add them in quadrature with our measurement errors from our mcmc fit to determine our full astrometric precision .
the combination of the forward modeled psf and the bayesian framework makes up the bayesian klip - fm astrometry technique we introduce in this work .
we apply bka for all twelve gpi datasets and report the measured astrometry and error budgets in table [ table : astrometry ] .
the best fit forward models and residuals to the fit are shown in appendix [ appendix : res ] .
on most of our datasets , we are not limited just by the uncertainty in the location of the planet , which was as low as 0.3 mas .
the uncertainties in the location of the star and north angle also make significant contributions to the error budget .
typically , we achieved @xmath71 mas precision on the relative astrometry between @xmath0 pic b and its host star .
this is a factor of @xmath724 improvement over previous techniques @xcite using the same data , indicating this technique can be useful in reanalysis of archival data to obtain better astrometry in cases where the limiting factor is the uncertainty on the planet position . in two of the later datasets where
the planet is observed closer in , we were limited by the snr of the planet and unable to achieve 1 mas precision . in the 2015 december 5 dataset ,
the noise was higher due to the planet being fainter relative to the star in _ j_-band . in the 2016 january 21 dataset , a combination of poor seeing and a small amount of usable data limited our astrometric precision .
overall though , this gpi @xmath0 pic b data is an excellent demonstration for bayesian klip - fm astrometry as the planet is bright enough that the extended psf features , such as the negative self - subtraction lobes , are clearly seen and provide significant information to constrain the position of the planet . for fainter planets ,
the extended features are harder to distinguish from the noise .
as one of the main advantages of bka over techniques that do not forward model the psf is being able to forward model the extended self - subtraction lobes , the astrometric improvement would not be as large for lower signal - to - noise ratio planets .
there still should be some improvement though due to accurately modelling the over - subtraction on the core of the psf and small contributions from the extended features even if they are hard to distinguish from noise .
regardless , in addition to the improved precision , bka should also more accurately estimate the uncertainties as it fits for the correlation scale of the noise at the location of the planet .
lcccc|cccc 2013 nov 16 @xmath50 & 0.6/0.7 & 0.7/0.7 & 0.3 & 0.13 & -228.5 @xmath51 1.3 & -366.2 @xmath51 1.1 & 431.6 @xmath51 1.0 & 212.0 @xmath51 0.2 + 2013 nov 16 @xmath52 & 0.5/0.4 & 0.7/0.7 & 0.3 & 0.13 & -229.2 @xmath51 1.2 & -364.5 @xmath51 1.0 & 430.6 @xmath51 0.9 & 212.2 @xmath51 0.2 + 2013 nov 18 @xmath53 & 0.3/0.3 & 0.7/0.7 & 0.3 & 0.13 & -229.1 @xmath51 1.1 & -364.7 @xmath51 1.0 & 430.6 @xmath51 0.8 & 212.1 @xmath51 0.2 + 2013 dec 10 @xmath53 & 0.4/0.4 & 0.7/0.7 & 0.3 & 0.13 & -227.9 @xmath51 1.2 & -359.9 @xmath51 1.0 & 426.0 @xmath51 0.9 & 212.3 @xmath51 0.2 + 2013 dec 10 @xmath54 & 0.6/0.7 & 0.7/0.7 & 0.3 & 0.13 & -227.2 @xmath51 1.3 & -361.1 @xmath51 1.2 & 426.6 @xmath51 1.1 & 212.2 @xmath51 0.2 + 2013 dec 11 @xmath53 & 0.5/0.4 & 0.7/0.7 & 0.3 & 0.13 & -227.8 @xmath51 1.2 & -359.2 @xmath51 1.0 & 425.4 @xmath51 0.9 & 212.4 @xmath51 0.2 + 2014 nov 8 @xmath53 & 0.5/0.5 & 0.7/0.7 & 0.2 & 0.13 & -194.0 @xmath51 1.1 & -299.1 @xmath51 1.0 & 356.5 @xmath51 0.9 & 213.0 @xmath51 0.2 + 2015 apr 2 @xmath53 & 0.4/0.5 & 0.7/0.7 & 0.2 & 0.13 & -172.1 @xmath51 1.0 & -266.5 @xmath51 1.0 & 317.2 @xmath51 0.9 & 212.9 @xmath51 0.2 + 2015 nov 6 @xmath53 & 0.7/0.7 & 0.7/0.7 & 0.2 & 0.13 & -137.8 @xmath51 1.1 & -207.2 @xmath51 1.0 & 248.8 @xmath51 1.0 & 213.6 @xmath51 0.3 + 2015 dec 5 @xmath54 & 1.2/1.3 & 0.7/0.7 & 0.2 & 0.13 & -133.9 @xmath51 1.5 & -200.5 @xmath51 1.5 & 241.1 @xmath51 1.4 & 213.7 @xmath51 0.4 + 2015 dec 22 @xmath53 & 0.5/0.5 & 0.7/0.7 & 0.2 & 0.13 & -130.0 @xmath51 1.0 & -194.7 @xmath51 0.9 & 234.1 @xmath51 0.9 & 213.7 @xmath51 0.2 + 2016 jan 21 @xmath53 & 1.8/1.6 & 0.7/0.7 & 0.2 & 0.13 & -126.8 @xmath51 2.0 & -185.8 @xmath51 1.8 & 225.0 @xmath51 1.8 & 214.3 @xmath51 0.5 + ccccccc semi - major axis ( @xmath55 ) & au & @xmath56 & uniform in @xmath57 & 9.02 & 9.66 & 10.78 + epoch of periastron ( @xmath58 ) & - & @xmath59 & uniform in @xmath58 & 0.32 & 0.73 & 0.87 + argument of periastron ( @xmath60 ) & & @xmath61 & uniform in @xmath60 & 192.8 & 205.8 & 258.4 + position angle of the ascending node ( @xmath62 ) & & @xmath63 & uniform in @xmath62 & 31.67 & 31.76 & 31.84 + inclination ( @xmath35 ) & & @xmath64 & uniform in @xmath65 & 88.70 & 88.81 & 88.93 + eccentricity ( @xmath66 ) & - & @xmath67 & uniform in @xmath66 & 0.027 & 0.080 & 0.171 + total mass ( @xmath68 ) & @xmath69 & @xmath70 & uniform in @xmath69 & 1.76 & 1.80 & 1.83 + + period ( @xmath71 ) & years & - & - & 20.21 & 22.47 & 26.24 + hill sphere ingress & mjd & - & - & 57,840 & 57,846 & 57,854 + @xmath72 hill sphere ingress & mjd & - & - & 57,916 & 57,924 & 57,934 + closest approach date & mjd & - & - & 57,986 & 57,996 & 58,008 + @xmath72 hill sphere egress & mjd & - & - & 58,056 & 58,069 & 58,082 + hill sphere egress & mjd & - & - & 58,132 & 58,147 & 58,162 +
to explore the validity of our new astrometry technique , we fit a keplerian orbit to our twelve epochs of astrometry . since each epoch of astrometry is fit independently , and
since the keplerian orbit is agnostic towards the exact data reduction methods , having all twelve epochs of astrometry fit the keperlain orbit would only be possible if all of the astrometry is accurate and precise .
if there are errors in estimating the magnitude of the uncertainties or any remaining biases in our measurements , we expect them to become evident in the residuals of the keplerian fit either as systematic trends or fit outliers .
we followed the same analysis as in @xcite to obtain the orbital elements of @xmath0 pic b using a mcmc fit with the parallel - tempered sampler implemented in ` emcee ` .
we combined the twelve gpi astrometric points presented here along with the datasets presented in @xcite and @xcite . unlike @xcite , we did not explicitly include the radial velocity measurement of @xmath0 pic b from @xcite in order to limit potential systematics between instruments , but we do use it to constrain the prior on the position angle of the ascending node and thus the direction of the orbit ( i.e. , we know that that @xmath0 pic b has been moving towards us since @xmath72007 ) .
our model of the orbit fits the same seven parameters as @xcite . for convenience ,
the parameters are listed in table [ tab : orb_params ] .
the mcmc sampler was run for 30,000 steps using 1024 walkers at each of the 20 temperatures after 30,000 steps of `` burn - in '' to allow the walkers to converge .
we thinned the chains to remove any remaining correlations , keeping every 75th step to result in effectively 400 samples per walker for a total of 409,600 samples to construct our posterior .
the posteriors on the seven parameters in our model are shown in figure [ fig : orbit_triangle ] and the 16% , 50% , and 84% percentiles of the marginalized posterior distribution for each parameter is listed in table [ tab : orb_params ] . for comparison ,
we plot 500 randomly - chosen possible orbits along with the measured astrometry in figure [ fig : orbit - vs - time ] .
the fit residuals of our measured gpi points are shown in figure [ fig : residuals ] .
the residuals in ra and dec offset ( which are the parameters we use in our mcmc ) are consistent with zero and do not show any systematics .
the residuals in radial separation and position angle ( pa ) , neither of which are used in our fit , are slightly further away from zero , but do not indicate any obvious errors in either our astrometry or our error estimates .
to quantitatively assess the validity of our measurements , we use the reduced chi - squared ( @xmath73 ) statistic , which measures how consistent our astrometry is with a keplerian orbit .
ideally , @xmath73 would be unity if all measurements and uncertainties were accurate .
however , biases and improperly estimated errors would cause it to deviate from unity . due to systematic astrometric calibration offsets between instruments that have not been characterized
, we expect @xmath73 to be slightly above unity .
for example , @xcite reported a @xmath73 of @xmath74 when combining gpi astrometry , measured using standard techniques in the field , with previous astrometric measurements even with @xmath724 times smaller error bars on the gpi points , the @xmath73 of our accepted orbits is @xmath75 , unchanged from the @xmath74 reported in @xcite . with bka contributing twelve out of the thirty astrometry measurements ,
if the bka technique contained biases larger than one milliarcsecond , they would have caused a significant increase in @xmath73 .
likewise , if we had been too optimistic with our error estimates , @xmath73 should have also increased as the reported uncertainties would not have matched the scatter in the measurements .
the lack of increase in @xmath73 indicates that the more precise astrometry from bka is not biased and has accurate uncertainties .
in addition to the keplerian orbit fit , we can examine the accuracy of our estimated one milliarcsecond uncertainties by looking at the astrometry measured on the same or consecutive days .
as we do not expect the planet s position to change significantly in the span of a single day , the three measurements in november of 2013 and the three measurements in december of 2013 ought to be consistent with each other . indeed
, our measurements in table [ table : astrometry ] indicate that in both sets of astrometry , the measurements agree at the milliarcsecond level , confirming that our estimated uncertainties are accurate .
thus , the well - behaved residuals of our gpi measurements and the lack of change in @xmath73 from @xcite even with significantly smaller error bars lead us to conclude the measured astrometry using bka are accurate and free from biases .
the fact the residuals are consistent with zero , the lack of change in @xmath73 , and the consistency of repeated measurements taken around the same time all indicate that the one milliarcsecond uncertainties estimated from bka are also accurate . together , these multiple assessments of the validity of bka all indicate that this technique is both accurate and precise .
having demonstrated the accuracy and precision of this new technique , we now analyze the new constraints on the orbit of @xmath0 pic b. not surprisingly , the estimates for @xmath55 and @xmath66 have not changed significantly from @xcite since the gpi points reanalyzed in this paper do not have a sufficiently long time baseline to constrain these parameters .
as seen in figure [ fig : orbit - vs - time ] , all but one astrometric measurement is on one half of the orbit curve . the other half of the orbit is not as well constrained , leaving a degeneracy in @xmath55 and @xmath66 .
this degeneracy can be broken with more measurements obtained when the planet appears on the other side of the star .
better constraints on @xmath55 and @xmath66 will provide better insight on how @xmath0 pic b interacts with the debris disk and potential unseen planets in the system .
the new total mass of the system , @xmath76 @xmath69 , is significantly higher than the @xmath77 @xmath69 from @xcite .
this new total system mass , which effectively measures the mass of the star at this precision , is in better agreement with the stellar mass of @xmath78 @xmath69 derived from stellar photometry @xcite .
our measurement of the position angle of the ascending node , @xmath62 , slightly improves upon the value obtained by @xcite , but the overall value remains consistent .
thus , @xmath0 pic b is still consistent with being the planet responsible for the known warp in the debris disk ( e.g. * ? ? ?
compared to @xcite , the argument of the periastron , @xmath60 in our fit has increased from @xmath79 to @xmath80 .
this new value is consistent with @xmath81 that is required for the falling evaporating bodies ( febs ) scenario proposed by @xcite to explain redshifted absorption features in @xmath0 pic s spectrum .
note that under previous definitions of the orbital parameters , this has been expressed as @xmath82 from the line of sight .
this scenario also requires a slightly eccentric orbit , which is consistent with our derived orbital parameters . for a more in - depth discussion of @xmath0 pic b s relationship to the debris disk and the feb scenario , we direct the reader to @xcite . the biggest improvement in our understanding of the orbit of @xmath0 pic b is the improved constraint on the inclination of the orbit .
we find the inclination to be @xmath83 which allows us to place the tightest constraints on the probability that @xmath0 pic b will transit its host star . assuming an angular diameter of the star of 0.736 mas @xcite and considering the range of @xmath55 from our orbit fit , we find that in most cases , we need @xmath84 in order for the planet to transit . with our current constraints on the inclination ,
we have ruled out the possibility that @xmath0 pic b will transit at 10-@xmath1 significance .
this tight constraint on the inclination and transit probability is due to the slightly longer time baseline and the improved precision in the measured position angle of the gpi astrometry compared to @xcite . for an edge - on orbit that transits the star
, we should see no significant change in pa over time .
however , figure [ fig : orbit - vs - time ] shows that the gpi points alone reveal a significant increase in pa over time .
thus , regardless of systematic astrometric calibration errors between instruments , we conclude that @xmath0 pic b will not transit its star . unlike the planet , @xmath0 pic b s hill sphere , the region around the planet that could contain gravitationally bound circumplanetary material , will transit the star .
we define the radius of the hill sphere as @xmath85{\left(\frac{m}{3m}\right)}$ ] using the approximate form proposed in @xcite where @xmath86 is the mass of the planet and @xmath87 is the mass of the star . assuming @xmath0 pic b follows ahot - start " evolutionary track with a mass of @xmath88 @xmath5 @xcite and using the range of semi - major axes , eccentricities , and stellar masses from our mcmc orbit fit , @xmath89 au ( @xmath90 mas ) . given that our prediction for the closest approach of @xmath0 pic b will be @xmath91 mas ( @xmath92 projected au ) from the star , the hill sphere of @xmath0 pic b will transit the star , as shown in figure [ fig : hillsphere_transit ] . from our mcmc orbit
fit , we can compute when the transit of the hill sphere will occur .
we pick five notable events to focus on : two are the ingress and egress of the hill sphere which are the extrema in time between which any circumplanetary material could transit ; another is the date of closest approach , which gives the opportunity to probe material closest in to the planet ; and the last two are the transit of the sphere that is @xmath72 @xmath93 in radial extent .
almost all stable prograde circumplanetary orbits reside within @xmath72 @xmath93 @xcite , so it is more likely to find material within half a hill sphere . in table
[ tab : orb_params ] , we list our constraints on the date of these events , and in figure [ fig : orbit_triangle ] , we plot the posterior distribution for these events
. the duration of the hill - sphere transit will be long : ingress is @xmath72017 april 3 and egress is @xmath72018 january 29 .
the 1/2 hill sphere begins transit @xmath72017 june 20 and ends transit on @xmath72017
november 12 , with closest approach on @xmath72017 august 31 . given the ra of @xmath0 pic ( @xmath94 )
, the star will not be visible from most ground - based observatories during almost the entire time between ingress of the hill sphere and closest approach .
ground - based telescopes in antarctica , airborne observatories capable of travelling to antarctica , and space - based observatories would provide the only opportunities to observe @xmath0 pic during this time period .
the second half of the hill - sphere transit will be visible from most southern - hemisphere ground - based observatories , making @xmath0 pic a well - suited candidate for photometric monitoring in late 2017 . as it is a rare opportunity to probe circumplanetary material ,
it is not certain what will be seen when the hill sphere of @xmath0 pic b transits the star .
one possibility is that satellites could reside in the hill sphere . to approximate the photometric transit depths , the 0.736 mas angular diameter of the star corresponds to 1.53 @xmath95 or 1.065@xmath9610@xmath97 km .
moons as large as ganymede ( @xmath98 = 2630 km ) or io ( @xmath98 = 1820 km ) would give transit depths of 2.473@xmath9610@xmath99 ( 2.7 mmag ) and 1.711@xmath9610@xmath99 ( 1.9 mmag ) , respectively . detecting these photometric signatures will require a high cadence , as any single satellite orbiting @xmath0 pic b will have a transit duration of @xmath72 days . additionally , @xmath0 pic is a variable star with pulsation timescales of @xmath70.5 hours and variability amplitudes in _
b_-band of @xmath100 mmags @xcite , so careful modelling of stellar activity is necessary to be sensitive to these transit depths .
another possibility is that , as @xmath0 pic b is still young and evolving , it may harbor a circumplanetary disk or ring system comprised of leftover material from planet formation .
such a hypothesis is not unprecedented as @xcite found evidence for a large circumplanetary disk around an unseen planet in the 1swasp j140747.93 - 394542.6 ( hereafter j1407 ) system , which has a similar age ( @xmath716 myr ; * ? ? ? * ) and thus likely at a similar stage in its evolution .
@xcite interpreted the series of complex and deep eclipses in the j1407 lightcurve as from a circumplanetary disk 0.6 au in radial extent in the process of forming rings due to newly formed satellites .
it is plausible that @xmath0 pic b can harbor a similar disk as a 0.6 au disk would be @xmath72 @xmath93 in extent and consistent with where we would expect stable orbits to reside around @xmath0 pic b. additionally , since @xmath0 pic b is young , it is plausible that there is a large amount of circumplanetary material which has yet to be cleared out dynamically .
such a large disk would transit the star and be suitable for detection through photometric monitoring of the host star if the disk is inclined by @xmath101 with respect to the orbital plane of @xmath0 pic b. to maintain such an inclination , the disk needs to be massive enough to prevent stellar tidal forces from aligning the disk to the orbital plane @xcite .
so far , there have not been many observational constraints of circumplanetary material around @xmath0 pic b. @xcite reported a photometric event in 1981 and hypothesized it could be due to the transit of a planet that cleared out a hole in the debris disk around its hill sphere .
however , @xcite show that the planet is not embedded in the debris disk and we definitely show the planet itself will not transit .
still , our orbit models give a @xmath102% and @xmath103% chance the photometric event coincided with the transit of the hill sphere and 1/2 hill sphere respectively , during which time material around the planet could have passed in front of the star .
additionally , a circumplanetary disk or ring may also be detectable through the planet s spectral energy distribution ( sed ) .
the planet s near - infrared spectrum would experience extinction if the dust resides between us and the planet , with the magnitude of extinction depending on the amount of dust . in the near infrared , the extinction would be greater at shorter wavelengths due to the increased scattering and absorption by dust and would produce a spectral slope in the planet s near - infrared sed .
dust around the planet would also scatter starlight , causing the planet s sed to appear brighter in the optical , as has been postulated for fomalhaut b @xcite .
furthermore , the dust will produce millimeter emission .
however , detecting circumplanetary material in the planet s sed will require being able to distinguish it from the circumstellar disk with precise spectral data .
in the first part of this work , we have presented a new technique for more precise and accurate astrometry of directly imaged exoplanets using a new analytical forward modelling approach in a robust statistical framework . * using the klip - fm framework presented in @xcite , we are able to analytically forward model the psf of the planet through the data reduction process , giving us better information on the location of the planet .
we apply klip - fm to gpi data on @xmath0 pic and forward model the psf of @xmath0 pic b using the open - source ` pyklip ` package . * for a close - in planet orbiting a bright star like in the case of @xmath0 pic b , we are limited by correlated speckle noise in our data .
we developed a bayesian framework utilizing gaussian processes and mcmc to account for the correlated noise and to find the position of the planet simultaneously .
* with this technique , we have achieved the most precise astrometry on @xmath0 pic b to date .
on most of our gpi datasets , we achieve @xmath71 mas precision on the relative separation between @xmath0 pic b and its host star , a @xmath72 - 4 fold improvement over previous techniques using the same data @xcite . * in datasets where the astrometry is limited by noise and not by astrometric calibration uncertainty
, bayesian klip - fm astrometry approach should improve astrometric precision .
in the second part of this work , we apply our bayesian klip - fm astrometry technique to the orbit of @xmath0 pic b. * to validate this new astrometric technique , we used it to measure the position of @xmath0 pic b in twelve epochs of gpi data .
we combined these twelve astrometric points with two previous astrometric monitoring campaigns and fit a keplerian orbit using mcmc methods .
we find the residuals to the fit are consistent with zero and show no apparent systematic trends , indicating that our fit is accurate and the uncertainties we estimate are reliable .
* due to the improved pa measurements from our technique , we have the tightest constraints on the inclination of the orbit and can exclude a possible transit of @xmath0 pic b at 10-@xmath1 significance . *
while the planet will not transit , we are confident the hill sphere around @xmath0 pic b will transit .
the hill sphere will begin transit at the start of april in 2017 and finish transiting at the end of january in 2018 with closest approach in the end of august in 2017 .
the transit of @xmath0 pic b s hill sphere should be our best chance in the near future to investigate young circumplanetary material . in the future , this mcmc forward modelling technique can be applied to photometry and spectral extraction alongside of astrometry of directly imaged exoplanets , allowing for improved characterization of their atmospheres . for
@xmath0 pic b , continued monitoring of its orbit will yield more insight into the dynamics of the star system , although the planet will soon be too close to its star to be seen with current direct imaging instrumentation .
however , once the planet appears on the other side , continued astrometric monitoring should be able to constrain the semi - major axis and eccentricity of the orbit much better , which will improve our understanding of how @xmath0 pic b perturbs the disk and if there are other planets perturbing @xmath0 pic b. the gpi project has been supported by gemini observatory , which is operated by aura , inc .
, under a cooperative agreement with the nsf on behalf of the gemini partnership : the nsf ( usa ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , mcti ( brazil ) and mincyt ( argentina ) .
we thank the national science foundation ( nsf ast-1518332 ) and nasa ( nnx15ac89 g and nexss program nnx15ad95 g ) for contributing to support of this research .
portions of this work were performed under the auspices of the u.s .
department of energy by lawrence livermore national laboratory under contract de - ac52 - 07na27344 .
this research has made use of the simbad database , operated at cds , strasbourg , france .
the posterior distribution plots were made with ` corner ` @xcite .
using the notation of @xcite , we can analytically forward model the psf of the planet through the process of stellar psf subtraction .
let us denote the target image as @xmath104 , which is taken at wavelength @xmath105 and time @xmath106 and contains the pixels @xmath107 from which we want to subtract the stellar psf . to forward model the psf of the planet , we need a model of the psf of the planet prior to stellar psf subtraction . for our gpi data ,
we use the satellite spots to generate a realization of an unocculted and unprocessed point source as a function of wavelength . using this model
, we can generate @xmath108 , a frame that consists solely of the unprocessed psf of the planet in the target frame .
the initial estimate for the planet s position in the frame is determined using an input separation and pa of the planet and the location of the star .
similarly , we can also generate an array of frames , each containing the unprocessed psf of the planet for each image in the reference library , accounting for the fact that the planet position changes due to adi field rotation and due to sdi rescaling to align speckles . in these frames ,
the psf shape only depends on the wavelength of the frame and the position of the psf depends on the apparent movement of the planet in the data due to adi and sdi .
this array of unprocessed psfs of the planet will be used to calculate the perturbation of the kl modes due to the existence of the planet . assuming a spectrum , @xmath109 , of the planet in the case of ifs data ( for imaging data
, @xmath109 is just a scalar ) , we can compute @xmath110 , the perturbation on the @xmath111th kl mode , @xmath112 , due to the existence of a planet , using equation e18 or e20 of @xcite .
we will not write out explicitly the exact formula to compute @xmath113 here since we have not made any modifications to it .
we note that we use equation e18 from @xcite to compute @xmath113 as it is computationally faster than equation e20 for a fixed input spectrum .
we calculate the perturbed @xmath113 for the first @xmath114 kl modes , where @xmath114 is the number of kl modes we choose to use in our psf subtraction . from here ,
the forward - modeled psf of the planet for this frame is computed as @xmath115 \\ - \left [ \sum_{k=1}^{k_{klip } } \left ( < t_{\lambda_{p},t_p}(\mathbf{x } ) , z_k ( \mathbf{x } ) > \delta z_k ( \mathbf{x } ) \right .
\\ + \left . \left .
< t_{\lambda_{p},t_p}(\mathbf{x } ) , \delta z_k ( \mathbf{x } ) > z_k ( \mathbf{x } ) \right ) \vphantom{\sum_{k=1}^{k_{klip } } } \right],\end{gathered}\ ] ] where @xmath116 is the inner product . equation [ fmmath ] is very similar to eq .
f7 in @xcite , but is focused on generating the forward model with a fixed input spectrum and not concerned with keeping the planet s spectrum as a free parameter for spectral extraction . as mentioned in @xcite , the term in the first square bracket is the over - subtraction term that is due to the projection of the kl modes on data with a planet in it , and the terms in the second square bracket are the self - subtraction terms due to the presence of a planet in the reference images influencing the kl modes . after generating forward models for each frame ,
we take the mean of all the forward models in time to give us a single forward modeled psf cube , @xmath117 that still has spectral information in the third dimension . in principle
, this psf cube can be used to retrieve spectral information from the data , but it is outside of the scope of this paper . for our astrometry purposes , we also take the mean in wavelength , generating a single 2-d forward modeled psf , which we call @xmath13 , for each dataset .
in figure [ fig : astrometry_residuals ] , we plot the residuals to the best fit model for each epoch using bka .
as @xmath0 pic b moves closer to its host star , the magnitude of speckle noise increases relative to the signal of @xmath0 pic b. this is especially true in the last epoch when @xmath0 pic b was closest to its star .
however , the fit was still accurate as the bright positive core of the psf of the planet was successfully modeled and not seen in the residuals , which are consistent with noise .
apai , d. , schneider , g. , grady , c. a. , et al .
2015 , , 800 , 136 beust , h. , & morbidelli , a. 2000 , , 143 , 170 binks , a. s. , & jeffries , r. d. 2016 , , 455 , 3345 bottom , m. , kuhn , j. , mennesson , b. , et al .
2015 , , 809 , 11 brandt , t. d. , mcelwain , m. w. , turner , e. l. , et al .
2013 , , 764 , 183 burrows , c. j. , krist , j. e. , stapelfeldt , k. r. , & wfpc2 investigation definition team 1995 , bulletin of the american astronomical society , 27 , 32.05 cantalloube , f. , mouillet , d. , mugnier , l. m. , et al .
2015 , , 582 , a89 chauvin , g. , lagrange , a .- m . , beust , h. , et al .
2012 , , 542 , a41 chilcote , j. , barman , t. , fitzgerald , m. p. , et al .
2015 , , 798 , l3 crifo , f. , vidal - madjar , a. , lallement , r. , ferlet , r. , & gerbaldi , m. 1997 , , 320 , l29 czekala , i. , andrews , s. m. , mandel , k. s. , et al .
2015 , , 812 , 128 dawson , r. i. , murray - clay , r. a. , & fabrycky , d. c. 2011 , , 743 , l17 de rosa , r. j. , nielsen , e. l. , blunt , s. c. , et al .
2015 , , 814 , l3 dent , w. r. f. , wyatt , m. c. , roberge , a. , et al .
2014 , science , 343 , 1490 defrre , d. , lebreton , j. , le bouquin , j .- b .
, et al .
2012 , , 546 , l9 esposito , t. m. , fitzgerald , m. p. , graham , j. r. , & kalas , p. 2014
, , 780 , 25 foreman - mackey , d. , hogg , d. w. , lang , d. , goodman , j. 2013 , 125 , 306 foreman - mackey d. , vousden w. , price - whelan a. , et al .
2016 , corner.py , v1.0.2 , zenodo , doi:10.5281/zenodo.45906 galicher , r. , & marois , c. 2011 , in second international conference on adaptive optics for extremely large telescopes ( victoria , bc : victoria conference centre ) , p25 , http://ao4elt2.lesia.obspm.fr graham , j. r. 2009 , gemini planet imager operational concept definition document v4.1 , ( hilo , hi : gemini observatory ) goodman , j. & weare , j. , 2010 , comm .
math . comp .
, 5 , 65 hamilton , d. p. , & burns , j. a. 1992 , , 96 , 43 heap , s. r. , lindler , d. j. , lanz , t. m. , et al . 2000 , , 539 , 435 ingraham , p. , perrin , m. d. , sadakuni , n. , et al . 2014 , , 9147 , 91477o kalas , p. , graham , j. r. , chiang , e. , et al .
2008 , science , 322 , 1345 kenworthy , m. a. , & mamajek , e. e. 2015 , , 800 , 126 koen , c. , balona , l. a. , khadaroo , k. , et al .
2003 , , 344 , 1250 konopacky , q. m. , thomas , s. j. , macintosh , b. a. , et al .
2014 , , 9147 , 914784 lafrenire d. , marois c. , doyon , r. , nadeau d. , & artigau .
2007 , , 660 , 770 lagrange , a .- m . , gratadour , d. , chauvin , g. , et al .
2009 , , 493 , l21 lagrange , a .- m . , bonnefoy , m. , chauvin , g. , et al .
2010 , science , 329 , 57 lagrange , a .- m . , boccaletti , a. , milli , j. , et al .
2012 , , 542 , a40 lecavelier des etangs , a. , vidal - madjar , a. , burki , g. , et al .
1997 , , 328 , 311 lecavelier des etangs , a. , & vidal - madjar , a. 2016 , , 588 , a60 macintosh , b. a. , graham , j. r. , ingraham , p. , et al . 2014 , pnas , 111 , 35 males , j. r. , close , l. m. , morzinski , k. , m. , et al .
2014 , , 786 , 32 mamajek , e. e. , bell , c. p. m. 2014 , , 445 , 2169 mamajek , e. e. , quillen , a. c. , pecaut , m. j. , et al .
2012 , , 143 , 72 marois , c. , doyon , r. , racine , r. , nadeau , d. 2000 , , 112 , 91 marois , c. , lafreniere , d. , doyon , r. , macintosh , b. , & nadeau , d. 2006a , , 641 , 556 marois , c. , lafrenire , d. , macintosh , b. , & doyon , r. 2006b , , 647 , 612 marois , c. , macintosh , b. , & vran , j .-
2010 , , 7736 , 77361j mawet , d. , milli , j. , wahhaj , z. , et al .
2014 , , 792 , 97 millar - blanchaer , m. a. , graham , j. r. , pueyo , l. , et al .
2015 , , 811 , 18 morzinski , k. m. , males , j. r. , skemer , a. j. , et al .
2015 , , 815 , 108 mouillet , d. , larwood , j. d. , papaloizou , j. c. b. , & lagrange , a. m. 1997 , , 292 , 896 nielsen , e. l. , liu , m. c. , wahhaj , z. , et al .
2014 , , 794 , 158 perrin , m. d. , maire , j. , ingraham , p. , et al .
2014 , , 9147 , 91473j pueyo , l. , soummer , r. , hoffman , j. , et al .
2015 , , 803 , 31 pueyo , l. 2016 , , 824 , 117 rameau , j. , nielsen , e. l. , de rosa , r. j. , et al .
2016 , , 822 , l29 shen , y. , & tremaine , s. 2008 , , 136 , 2453 sivaramakrishnan , a. , & oppenheimer , b. r. 2006 , , 647 , 620 smith , b. a. , & terrile , r. j. 1984 , science , 226 , 1421 snellen , i. a. g. , brandl , b. r. , de kok , r. j. , et al .
2014 , , 509 , 63 soummer , r. , pueyo , l. , & larkin , j. 2012 , , 755 , l28 thbault , p. , & beust , h. 2001 , , 376 , 621 van leeuwen , f. 2007 , , 474 , 653 wang , j. j. , rajan , a. , graham , j. r. , et al .
2014 , , 9147 , 914755 wang , j. j. , ruffio , j .- b . ,
de rosa , r. j. , et al . 2015 , astrophysics source code library , ascl:1506.001 wolff , s. g. , perrin , m. d. , maire , j. , et al . 2014 , , 9147 , 91477h zanazzi , j. j. , & lai , d. 2016 , arxiv:1605.02365 | a principal scientific goal of the gemini planet imager ( gpi ) is obtaining milliarcsecond astrometry to constrain exoplanet orbits .
however , astrometry of directly imaged exoplanets is subject to biases , systematic errors , and speckle noise .
here we describe an analytical procedure to forward model the signal of an exoplanet that accounts for both the observing strategy ( angular and spectral differential imaging ) and the data reduction method ( karhunen - love image projection algorithm ) .
we use this forward model to measure the position of an exoplanet in a bayesian framework employing gaussian processes and markov chain monte carlo ( mcmc ) to account for correlated noise . in the case of gpi data on @xmath0 pic b ,
this technique , which we call bayesian klip - fm astrometry ( bka ) , outperforms previous techniques and yields 1@xmath1-errors at or below the one milliarcsecond level .
we validate bka by fitting a keplerian orbit to twelve gpi observations along with previous astrometry from other instruments .
the statistical properties of the residuals confirm that bka is accurate and correctly estimates astrometric errors .
our constraints on the orbit of @xmath0 pic b firmly rule out the possibility of a transit of the planet at 10-@xmath1 significance .
however , we confirm that the hill sphere of @xmath0 pic b will transit , giving us a rare chance to probe the circumplanetary environment of a young , evolving exoplanet .
we provide an ephemeris for photometric monitoring of the hill sphere transit event , which will begin at the start of april in 2017 and finish at the end of january in 2018 . |
grb counterparts are multiwavelength emitters , unlike supernovae , which emit most of their power at the optical frequencies .
however , observations in the optical band have had the biggest impact in the study of the grb phenomenon , by sampling the time histories of the grb counterparts from a few seconds up to years after the explosion , by establishing the nature of the afterglow emission ( synchrotron radiation ) , and by unambiguously assessing the extragalactic origin of grbs .
grb optical counterparts can be initially extremely bright .
however , the delayed emission , even though it is long lived with respect to the prompt event , still fades quite rapidly , and the host galaxies of grbs are very small and faint , because of their cosmological distances .
therefore , telescopes of all aperture sizes have been involved in the investigation of the optical afterglows .
the smaller telescopes are more flexible for an efficient search and the larger ones ( including hst ) play a leading role in early spectroscopy and polarimetry , as well as in the photometric monitoring at late epochs , and in the study of the host galaxies .
the most important observations of optical afterglows and their host galaxies , which have led to the current understanding of the physics of these sources , are discussed .
previous reviews on this subject include @xcite .
although the present review focuses on the optical observations , information provided by the scanty infrared and ultraviolet data is also included , considering the proximity of these bands to the optical .
the chapter is organized as follows : in section 2 the steps which led to the first detection of optical afterglow emission are summarized and the observational problems related to the afterglow investigation are described ; in section 3 the temporal behavior of the optical prompt and afterglow emission is reviewed ; in section 4 the observed spectral continuum at infrared , optical and ultraviolet wavelengths is compared to the standard afterglow model based on the propagation of an external shock , and the characteristics of the afterglow absorption spectra are illustrated .
section 5 is devoted to a discussion of those aspects of grb host galaxies which seem to be relevant to the afterglow investigation .
future perspectives in the exploration of grb optical counterparts are sketched in section 6 .
the search for optical counterparts of grbs started soon after grb discovery in the belief that they held the key toward the physical origin of the grb phenomenon .
the first searches in the 1970 s were based on inspection , on archival plates or on photographic images of the grb field , of the large error boxes yielded by the early high energy missions ( see @xcite for a review of these pioneering attempts ) . the reasons of their failure , as we now understand , reside in the inadequacy of the methods . since grb afterglows fade quickly , only timely and sensitive surveys of accurate , rapidly disseminated grb error boxes can be effective in detecting optical counterparts . the first identification of an optical afterglow was made possible on 28 february 1997 by the bepposax satellite @xcite , whose wide field cameras ( wfc , @xcite ) determined the position of a grb with a @xmath0 uncertainty radius and disseminated it to the community within a few hours @xcite .
optical observations made 21 hours and @xmath11 week after the grb allowed the detection of a variable source in the grb error area , which had been meanwhile refined and reduced to @xmath11 arcmin@xmath2 .
variability and positional coincidence suggested association of the optical transient with the grb @xcite .
five years later , out of about 100 grbs accurately and rapidly localized by bepposax or by other spacecrafts , and timely followed by optical telescopes ( delays of a few hours to a few days of the grb explosion ) , @xmath130 have a detected optical afterglow .
these are reported in table 1 , together with those grbs for which , despite the lack of optical afterglow detection , the precise afterglow positioning at other wavelengths has allowed the identification of a host galaxy ( in these cases the lower limit on the magnitude of the transient counterpart is reported , as determined in the earliest search ) .
2 of table 1 reports the instrument , or suite of instruments , which have localized the grb : bepposax " means that the event has triggered the gamma - ray burst monitor ( grbm ) of the bepposax satellite @xcite , and has been localized with arcminute precision by the onboard wfc . in two cases the bepposax wfc detected in the x - rays and localized an event which triggered batse , but not the grbm . for almost all grbs localized by the rossi xte satellite ,
the accurate positioning was obtained with its all sky monitor , the exception being grb990506 , which was localized precisely only after the rxte proportional counter array ( pca ) detected its x - ray afterglow @xcite .
many grbs have been localized by the interplanetary network ( ipn ) of spacecrafts , whose synergy allows triangulation of the grb position , yielding in many cases arcminute - sized error areas , although with delays seldom shorter than 12 hours , and often larger than 24 hours @xcite . for most cases ,
a refinement of the localization error box has been possible after detection of the x - ray afterglow by the bepposax narrow field instruments , by the rxte pca , or by chandra ( see also chapter by f. frontera ) .
optical afterglows are generally identified for being previously unknown sources and for their variability , either by comparing deep images of the grb field , acquired soon after the grb detection , with the digitized sky survey ( dss ) , or , for transients fainter than the dss limit , with images obtained at later epochs .
an alternative technique for the selection of grb counterpart candidates is based on the characteristic power - law spectral shape of afterglows , and can be successfully employed when sufficient color information is available @xcite .
the method may suffer from redshift - dependent biases and possible contamination by other classes of sources ; however , it can be advantageously applied to a single set of images , instead of two ( or more ) sets of images taken at different epochs .
it should be specified that so far optical counterparts have been detected only for long duration grbs , which represent one of the two populations into which grbs are subdivided ( see , however , @xcite for the possible existence of a third class ) . according to their duration ,
the grbs of the batse sample are divided into long ( 75% of the total ) and short events ( 25% ) , with average durations of @xmath120 s and @xmath10.2 s , respectively .
this bimodality is reflected also in the spectral hardness , with long grbs tending to be softer than sub - second grbs @xcite , and may be due to a different origin of the two classes of sources .
sub - second grbs could not be accurately localized by the bepposax wfc @xcite ; four of them have been localized with arcminute precision by the ipn , and followed up in the optical with delays no shorter than @xmath120 hours .
no afterglow has been detected ; the upper limits show that , within the limited statistics , the optical afterglow behavior of sub - second grbs may not differ from that of long ones @xcite .
the grb000301c , which was detected by the ipn with a 2 s duration and exhibited a bright , variable counterpart ( see 3.2 ) , can not be unambiguously classified as a long , sub - second hard , or intermediate grb @xcite .
considering only grbs for which the angular localization is better than @xmath130 arcmin@xmath2 and was disseminated within 24 hours , the statistics of detected optical afterglows is @xmath140% of the total .
therefore , many grbs are optically dark " , though nearly all of these have x - ray and/or radio afterglows .
many causes can concur to make optical searches unsuccessful ( see also @xcite ) .
optical afterglows can be intrinsically faint , rapidly decaying , or dim because of the large redshift , causing the ly@xmath3 break to affect the optical spectrum .
therefore , in some cases , the lack of an optical detection may be due to the insufficient sensitivity of the search @xcite .
the different decay rates of the optical afterglows on the one hand and the rather wide range of magnitudes measured , at equal intervals after the grb , for optical counterparts with comparable decay rates on the other ( see table 1 ) , indicate that the chance of detection may be very different from case to case for similarly prompt and deep exposures .
this , together with the lack of a straightforward correlation between optical emission and gamma - ray brightness of the prompt event , makes it difficult to predict the detection level and to devise an optimal observing strategy .
in addition , optical afterglows may be affected by extinction in the plane of our galaxy ( which is instead transparent to gamma- and hard x - rays and to radio wavelengths ) , and by absorption in their host galaxies , which makes their detection all the more challenging . by simulating the absorption experienced at the center of a dense dust clump , similar to those found in star forming regions in our galaxy and in external ones , in a number of directions randomly distributed , lamb @xcite determined that in only 35% of the lines of sight the optical depth is @xmath4 , while in the remainder it is @xmath5 .
this statistic is consistent with the percentage of dark grbs , supporting the idea that local dust absorption may hamper or completely prevent optical detection of the grb afterglow and strengthening the importance of infrared observations ( see 4.1 ) .
the presence of substantial quantities of dust at the burst explosion site favors , in turn , the association of grbs with star forming regions ( see 5 ) .
although the result of this test would relate all dark grbs to the effect of dust extinction , perhaps this is only one of the possible causes for a failed optical detection .
an alternative view @xcite is that dark grbs can be heavily extincted only if dust sublimation by the strong uv / optical @xcite and x - ray radiation @xcite following the explosion does not play a significant role .
if dust destruction around the burst site is important @xcite , then dark grbs should belong to a distinct population with respect to grbs with detected optical afterglows .
prompt emission at optical and near - infrared wavelengths simultaneous with a grb , or delayed by a few seconds , is expected to take place as a consequence of a reverse shock propagating into the explosion ejecta , and is therefore distinct from the afterglow , which is produced by the interaction of the forward shock with the interstellar medium @xcite .
this low energy early emission can be very bright , in principle up to the @xmath15th magnitude in the visual band , for an intense grb at @xmath6 , but is expected to last only tens of seconds .
since current instrumentation disseminates accurate grb localizations with delays of a few hours , grb fields are usually imaged in optical starting no earlier than some hours after the explosion .
even if dissemination occurred in real time ( as was the case for the large batse error boxes ) , the observer reaction at a traditional optical telescope would still take at least several minutes .
small robotic telescopes , which can slew automatically and rapidly to the grb position in response to a localization delivered in real time , are therefore the most efficient instruments for grb follow - up at the earliest epochs . on 23 january 1999
, an optical flare was detected by the rotse - i robotic telescope ( consisting of a two - by - two array of 35 mm lenses ) , starting 22 seconds after the onset of a grb which triggered both bepposax and batse . at maximum ,
the optical transient reached @xmath7 @xcite , implying a power output in the optical of about @xmath11% of that at the gamma - ray energies @xcite , in agreement with the reverse shock interpretation , and allowing an estimate of the plasma initial lorentz factor @xcite . in fig .
[ lc990123 ] the rotse - i data are reported ( from immediately after the light maximum , to @xmath110 minutes after the grb , before the decreasing flux level becomes undetectable ) , along with the measurements of the successive afterglow taken at bigger telescopes .
the rotse - i points are fitted by a steeper temporal power - law than the afterglow points , indicating two different radiation mechanisms .
owing possibly to the exceptional brightness of the event , to the rapidity of the rotse - i slew , to the precise bepposax localization ( which allowed identification of the transient in the @xmath8 rotse - i ccd image ) , and to the limited sensitivity of state - of - the - art robotic telescopes , grb990123 still remains the unique case of detection of an optical flash " simultaneous with the grb event itself .
prompt searches of other grb error boxes both with rotse - i and with other robotic systems have yielded no detection to limiting magnitudes spanning from @xmath14 to @xmath115 at epochs comprised between 10 seconds and 30 minutes after the grb @xcite
. -band light curve of the grb990123 afterglow .
all points , except for the hst point ( rightmost filled square ) , represent measurements taken from the ground ( see @xcite for references ) and are reduced to a common flux standard with the galaxy flux subtracted . error bars ( 1 @xmath9 ) are shown where available , and arrows indicate 95% confidence upper limits ( from @xcite).,scaledwidth=60.0% ] at epochs between a few hours and @xmath11 day after the grb the afterglow decays following approximately a temporal power - law @xmath10 with an index @xmath3 ranging from @xmath10.7 to @xmath12 ( see table 1 , col . 4 , and fig.[lc990123 ] ) .
this behavior was predicted before the detection of the first afterglow as a consequence of the simplest version of the fireball model @xcite . according to this scenario ( see chapter by e. waxman ) , the grb explosion triggers a relativistic forward shock which develops into the interstellar medium ( or in a wind pre - ejected by the grb stellar progenitor @xcite ) and accelerates the particles .
these interact with the magnetic field and radiate at all frequencies through the synchrotron mechanism ( significant contribution from the inverse compton scattering emission is also predicted and observed in some cases @xcite ) .
linear polarization , measured in the optical afterglows of grb990510 and grb990712 at the level of a few percent @xcite , represents a good test of the synchrotron mechanism @xcite and of collimated emission @xcite ( see below ) .
near - infrared polarimetry of grb000301c yielded an upper limit of 30% .
although not very constraining , this is consistent with a synchrotron origin of the continuum in a relativistic jet @xcite . on average , the initial luminosities of grb optical afterglows are two orders of magnitude larger than maximum supernova luminosities , and obviously outshine their parent galaxies .
at longer intervals after the grb , when the flux of the transient subsides under the brightness of the host galaxy , it is possible to measure the magnitude of the latter . in some cases ( notably grb970228 , grb980326 and grb011121 ) ,
the light curve of the optical afterglow exhibits a rebrightening with respect to a power - law behavior at few weeks after the grb .
following claims for the possible association of grb980425 with the close - by ( @xmath11 ) sn1998bw @xcite ( see also chapter by t. galama ) , the light curves of those optical transients have been decomposed into a non - thermal , pure afterglow contribution and a supernova profile , using sn1998bw as a template , appropriately redshifted . in some cases
the results are convincing @xcite , while in other cases they are not decisive @xcite .
systematic decomposition into non - thermal and supernova emission components has been attempted also by dado et al .
@xcite for the afterglows of grbs with known redshift , in the framework of the cannonball model , and good results have been obtained in most cases .
the afterglow magnitudes reported in table 1 ( col .
6 ) have been obtained from a fit with a single or double power - law after subtraction of the host galaxy flux , and in some cases after decomposition of a possibly underlying supernova ( see references for individual cases in col .
9 ) . in those cases where the few optical measurements did not allow a proper fit , the optical magnitude measured closest to 1 day after explosion was used ( e.g. , grb980613 ) . in a large fraction of the best monitored
optical and/or near - infrared afterglows the initial power - law decline steepens at times ranging from @xmath10.5 to @xmath15 days after the grb explosion .
the effect is clearly seen as a smooth increase of the flux decay rate @xcite , and is suggested also by the x - ray data in a few cases @xcite . in table 1 ( cols . 4 and 5 ) the early and late temporal indices are reported as determined via empirical fits to the optical light curves with double power - laws ( see also fig.[lc990123 ] ) .
the change in the temporal slope is thought to witness the presence of a decelerating jet .
collimation of the radiation in a jet structure would reduce the huge energy outputs ( @xmath12 erg ) derived from the observed gamma - ray brightnesses and the measured distances of grbs , in the assumption of isotropy , and thus help resolving the paradox of energy conversion efficiency @xcite . when the aperture of the radiation cone ( beaming angle ) , which progressively increases as the relativistic plasma decelerates , becomes larger than the jet opening angle
, the observer perceives a faster light dimming , independent of wavelength , due to the jet edge becoming visible and/or to jet sideways expansion @xcite .
the change in fading rate is however smooth , due to light travel time effects at the ending surface of the jet @xcite .
the steepening of the afterglow light curve would then be a probe of the grb and afterglow emitting geometry .
stanek et al .
@xcite note an anti - correlation between the slope change @xmath13 and the isotropic gamma - ray energy of the burst , suggesting that the different jet opening angle may be responsible for it .
specific jet models for individual cases have been proposed @xcite , and afterglow emission from jets has been modeled in the firecone scenario as a function of the viewing angle @xcite , as also suggested by dado et al .
@xcite .
if jets are unavoidable to relax the energy crisis in grbs and a light curve steepening is their signature , one may wonder why all observed optical afterglows do not exhibit a detectable steepening in their light curve .
this may simply be due to undersampling : when not detected , the steepening may have occurred at early , not well - sampled epochs ( many afterglows are described by power - laws with temporal indices steeper than 2 , see table 1 ) , or at late epochs , when the afterglow behavior is significantly contaminated by the emerging host galaxy or possible supernova , so that discerning a decay rate variation is more difficult ( see e.g. , @xcite ) .
on the other hand , light curve steepening can not be univocally ascribed to a decelerating jet , but may be caused instead , or in addition , by the transition of a spectral break through the observing frequency band @xcite ( see 4.1 ) or by the propagation of the external shock in a non homogeneous medium @xcite ( although in these cases the steepening would be frequency dependent ; see however @xcite for detectability of a jet in a stratified medium ) , or by the transition of the plasma kinematic conditions from relativistic to newtonian in a dense medium @xcite . in some cases the interpretation is not unique @xcite , although simultaneous multiwavelength observations may resolve the ambiguity @xcite .
we finally note that some months after the grb a flattening of the afterglow light curve may be expected instead @xcite .
optical interday or intraday variations superimposed to the overall afterglow decline are rarely detected , because of the limited photometric precision of the measurements .
two cases where significant deviations from a steady decay have been observed during the optical monitoring are grb970508 , which exhibited an initial shallow decline , followed by a 2-day re - bursting of a factor @xmath15 amplitude @xcite correlated with an x - ray flare @xcite , and grb000301c , which showed intraday achromatic variability of 20 - 30% amplitude @xcite . for both events , an interpretation based on microlensing has been proposed @xcite .
rapid variability can otherwise be produced by small scale inhomogeneities of the plasma flow , or irregularities of the external medium in which the blast wave propagates @xcite , or local re - energization episodes @xcite .
the classical fireball model has specific predictions for the temporal evolution of the broad - band spectral shape @xcite .
this has been modified to include the detectable effects of the presence of a jet @xcite .
both the spectral slope and the temporal decay rate depend on the index @xmath14 of the electron energy distribution @xmath15 ( where @xmath16 denotes the electron energy above a certain cutoff ) .
the radio - to - x - ray spectral shape is characterized by smooth breaks at typical frequencies ( self - absorption , peak and cooling frequencies ) , which evolve with time in a predicted way @xcite , so that simultaneous multiwavelength observations at various epochs during the evolution of the afterglow allow the measurement of the spectral slopes and breaks and the estimate of the relevant physical parameters of the afterglow ( see e.g. , @xcite , and chapter by f. frontera ) .
when the optical photometric observations are accurate and sufficiently extended in time to make a good signal - to - noise ratio measurement of the spectral and temporal slopes possible , they show that the spectra of some afterglows , corrected for galactic extinction , are steeper ( i.e. , redder ) than expected from the fireball theory based on comparison with the temporal decay rate .
this has been commonly attributed to absorption intrinsic to the source or , especially for grbs at very high redshift , intervening along the line of sight @xcite .
a small amount of reddening by dust in the grb host galaxy has been invoked in many cases to reconcile the observations with the theoretical scenarios , using extinction curves typical of our own galaxy , of star - forming galaxies , or of the lmc and smc @xcite .
while even a moderate quantity of dust at the grb source redshift may significantly attenuate the observed optical spectrum ( which corresponds , at the average @xmath17 , to rest - frame ultraviolet wavelengths ) , or even completely obscure it ( see 2 ) , near - infrared data are less affected and may be more effective in determining the overall afterglow spectrum , when combined with data at other frequencies @xcite .
observations in the near - infrared range are therefore critical for the study of afterglows .
low and medium resolution spectra of bright optical afterglows have allowed , in a number of cases , the detection of absorption lines of metallic species caused by intervening absorbers , and the measurement of lower limits to the grb redshift ( see table 1 and fig.[010222spec ] ) .
frequently , the evidence of a low - ionization , high density medium ( e.g. , related to the detection of mg i in absorption ) suggests that the absorbing system is actually the host galaxy @xcite .
spectroscopy of the likely host galaxy generally shows that the highest absorption redshift coincides with the redshift of the galaxy emission lines , confirming the association of the grb with the proposed host .
no variability of the absorption line equivalent widths is detected at the 20% level ( which represents the 1-@xmath9 uncertainty on the measurements ) in time scales of some hours to few days @xcite .
( from @xcite).,scaledwidth=80.0% ] for grb000301c and grb000131 the redshift has been determined through identification of absorption edges with the lyman limit and intervening ly@xmath3 forest , respectively .
the afterglow of the former grb is the only one which has been observed by hst at ultraviolet wavelengths : in the low - resolution spectrum taken 5 days after explosion with the stis instrument equipped with the nuv mama prism a discontinuity at @xmath12800 is clearly detected , which has been identified with the hydrogen ionization edge @xcite , implying @xmath18 ( see fig.[smette ] ) .
the measurement was then confirmed and refined from optical spectroscopy at the keck and vlt telescopes @xcite .
the redshift of grb000131 , @xmath19 , was determined photometrically from simultaneous near - infrared and optical observations ( see fig.[000131spec ] ) , and supported by optical spectrophotometry @xcite . .
this matches with the ground - based measurement of the redshift ( from @xcite).,scaledwidth=80.0% ] the distance of grb980329 is controversial : the afterglow photometry suggests a redshift as large as @xmath20 or lower , @xmath21 , according to whether the observed continuum suppression shortward of @xmath16500 is identified with ly@xmath3 intervening absorption @xcite , or with molecular hydrogen dissociation by the strong initial ultraviolet flash @xcite .
a redshift @xmath22 would be suggested by the absence of the ly@xmath3 break in the host galaxy spectrum @xcite .
the redshifts measured so far either with spectroscopy or broad - band photometry span the range @xmath10.4 to @xmath14 ( table 1 ) , excluding the peculiar case of grb980425 ( see chapters by t. galama and k. iwamoto ) , and prove that long duration grbs have an extragalactic , cosmological origin , which makes their early bright optical afterglows excellent probes of the high redshift universe . - and @xmath23-band fluxes include the formal error from the extrapolation of the light curves back to the epoch of the optical measurements , @xmath24 days . a fit with a power - law spectrum with ly@xmath3 forest absorption and smc reddening is shown as a dashed curve .
this yields @xmath25 , when an intrinsic spectral slope @xmath26 and a redshift of 4.5 is assumed .
the solid curve shows the corresponding spectrum with its lyman absorption edges ( from @xcite).,scaledwidth=70.0% ]
for almost every well studied optical afterglow , deep late epoch optical or near - infrared observations from the ground or with hst have detected a galaxy close to the point - like optical transient or at its position after it has faded away ( see table 1 ) .
host galaxies have been detected also for some dark grbs with arcsecond afterglow localizations from radio telescopes or from chandra .
few observations of host galaxies have been made at longer wavelengths @xcite ) .
the hosts optical magnitudes ( and upper limits ) are consistent with those expected for a reasonable redshift distribution and galaxy population @xcite .
this solves the no - host galaxy " problem @xcite , which , a posteriori , turns out to be clearly related to the very faint flux of the host galaxies , which are usually detected only with long exposures at telescopes larger than 2 m .
hst observations taken at early stages of the afterglow evolution , when the transient is still bright ( see fig.[990123host ] ) , show that this lies always within the stellar field of its host galaxy .
if only late epoch hst images are available , their comparison with accurate astrometry of the bright transient on early epoch ground - based images still yields projected angular offsets of a fraction of an arcsecond between the transient and the galaxy center .
when normalized to the individual hosts half - light radii , the median offset is 0.98 @xcite .
this has led to the conclusion that grbs are associated with star forming regions , which would support their origin as hypernovae @xcite ( or collapsars @xcite ) or as _ supranovae _ @xcite , as opposed to progenitor scenarios which envisage the explosion as taking place at many kiloparsecs from the parent galaxy , in its halo ( like binary neutron star systems , see chapter by e. waxman ) .
there is specific evidence that host galaxies of grbs undergo strong star formation : 1 ) their integrated colors are remarkably blue @xcite ; 2 ) they are usually underluminous ( luminosities of a fraction of the characteristic luminosity @xmath27 of the schechter @xcite luminosity function ) , small , and have often a compact morphology @xcite , which are characteristics common to galaxies hosting star formation at the typical grb redshifts , @xmath17 @xcite ; 3 ) their spectra exhibit star formation emission lines like [ o ii ] , [ ne iii ] ( fig.[970508host ] ) , [ o iii ] , ly@xmath3 and balmer series @xcite ; 4 ) their star formation rates , derived either from the emission line intensities or from the ultraviolet rest frame galaxy continuum are high , compared with the galaxy size @xcite , although sometimes substantially obscured @xcite and possibly measurable only at rest frame far - infrared wavelengths @xcite ; 5 ) they sometimes have morphologies consistent with being mergers or interacting systems @xcite , where star formation is enhanced . furthermore , the imprints of local extinction on the afterglow spectra ( see 4.1 ) point to dusty and likely star forming environments as the favored grb explosion sites .
finally , the observed host galaxy magnitudes and redshifts are consistent with a model in which the comoving rate density of grbs is proportional to the cosmic star formation rate density @xcite .
these suggestions are consistent with the fact that most measured grb redshifts are around @xmath17 , where the cosmic star formation rate is one order of magnitude larger than locally @xcite .
since this is predicted to increase monotonically back to @xmath28 , one may expect that , given the opportunity of detecting grbs up to that redshift , it would be possible to select the youngest star forming galaxies in the universe .
the enormous progress achieved over the last few years through the optical afterglow follow - up study has also indicated some fundamental , unclear aspects of grb and afterglow physics : 1 ) the conversion of energy into radiation ; 2 ) the structure and geometry of the emitting regions ; 3 ) the nature , density and composition of the circumburst medium ; 4 ) the cosmological evolution of the grb population . the solution of these problems would ultimately lead to unveiling the major unknown , i.e. the identity of grb progenitors . while at the present stage these issues remain a matter of investigation , current knowledge suggests the observational approach to tackle them most effectively .
the most serious limitation of present day observations is the substantial temporal delay between grb trigger and follow - up of its field at lower frequencies ( a few hours ) .
this causes a lack of sampling of the initial portions of the light curves . bridging this gap will be possible when real - time disseminated localizations , made available by the coming generation of high energy satellites , can be rapidly followed by small , fast reacting telescopes with suitably large fields of view .
thanks to the uvot camera onboard the grb mission swift ( to be launched in 2003 ) and to the advent of ground - based robotic telescopes @xcite , the monitoring will start as early as few tens of seconds after grb detection , allowing astronomers to catch the transient counterpart in its maximum emission state , and to follow the temporal evolution of its optical - to - infrared continuum .
it is at these early epochs that the models differ most in their predictions @xcite , and the strength of the signal can discriminate them with the highest confidence .
early , simultaneous optical / near - infrared searches will either detect more counterparts , and reduce the number of dark grbs , or put stronger constraints on the darkness " .
with respect to robotic systems , all - sky optical monitoring cameras @xcite present the advantage that they would be independent from spacecraft triggers , and therefore they could detect possible precursors of grbs and burstless " afterglows , which would be obviously missed by robotic telescopes .
weeks after the grb explosion , the light curve will result from the sum of different components of comparable brightness : the fading afterglow , the possible supernova rising to maximum light , and the host galaxy . to disentangle these contributions
we will need sensitive and densely sampled photometric observations at late epochs in the optical and infrared .
these will also possibly discern the presence of a light echo , due to dust scattering @xcite or sublimation @xcite .
this task is a prerogative of the sensitive , high resolution optical and near - infrared cameras in space , such as the hst wfpc2 , stis , nicmos , and the newly - deployed advanced camera for surveys .
the future ngst will collect the legacy of these instruments and push the research on grb late afterglows and hosts toward even larger redshifts .
the measurement of a large number of redshifts , through early absorption spectra of optical afterglows as well as emission line spectra of the host galaxies , is necessary to construct a luminosity function of grbs to be compared with models of star forming rate evolution .
this will allow us to test the link between grbs and star formation history up to very high redshifts , many hints of which have been so far collected @xcite . a primary role in early and late sensitive optical and infrared spectroscopy
will be played by the ground - based 4 to 8 m class telescopes in both hemispheres .
early bright counterparts will be excellent targets for spectroscopic monitoring : variations in the equivalent widths of the absorption lines will be measured with good signal - to - noise ratio , thus making it possible to place constraints on the density and distribution of the circumburst medium @xcite , which is a critical diagnostic of the grb progenitor .
furthermore , the intense initial optical flares associated with grbs will also act as background light bulbs " to probe the ionization state and metallicity of the intergalactic medium through high resolution spectroscopy .
more generally , multiwavelength observations of grbs allow a series of cosmological tests on a wide range of redshifts @xcite ( see also chapter by a. loeb ) .
prompt , long and intensive polarimetric monitoring of afterglows will detect possible changes of polarized light percentage and position angle and thereby set constraints on the most important open issue of afterglow physics , the generation of magnetic fields @xcite .
moreover , knowledge of the magnetic field geometry and of the circumburst density profile will be instrumental to defining the structure of the jet and its interaction with the ambient medium .
finally , real time dissemination of accurate localizations of sub - second grbs and the prompt follow - up of these fields in the optical will hopefully afford detection of their elusive counterparts , and will allow us to get clear insights into their genesis and physics .
i am grateful to the members of the grace consortium led by e. van den heuvel for a longstanding , fruitful collaboration , and in particular to f. frontera , a. fruchter , n. masetti and e. palazzi , who have been my closest collaborators in grb optical counterparts research since the beginning of the afterglow era " .
i thank s. recchi for valuable inputs , and a. de rjula , a. fruchter , p. mazzali and e. waxman for a critical reading of the manuscript .
i would like to dedicate this review to the memory of jan van paradijs .
ccccccccc grb & instr.@xmath29 & @xmath30 & @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & refs .
+ & & & & & ( mag ) & ( mag ) & ( mag ) & + + 970228 & bepposax & 0.695 & 1.7 & & @xmath36 & @xmath37 & 0.63 & @xcite + 970508 & bepposax & 0.835 & 1.3 & & @xmath38 & @xmath39 & 0.13 & @xcite + 970828 & batse / rxte & 0.9578 & ... & ... & @xmath40 ( 4 h ) & @xmath41 & 0.10 & @xcite + 971214 & bepposax & 3.418 & 1.4 & & @xmath42 & @xmath43 & 0.04 & @xcite + 980326 & bepposax & @xmath44 ? & 2.0 & & @xmath45 & @xmath46 & 0.21 & @xcite + 980329 & bepposax & 3 - 5 & 1.2 & & @xmath47 & @xmath48 & 0.19 & @xcite + 980425 & bepposax & 0.0085 & ... & ... & @xmath49 & @xmath50 & 0.17 & @xcite + 980519 & batse / wfc & ... & 1.73 & 2.22 & @xmath51 & @xmath52 & 0.69 & @xcite + 980613 & bepposax & 1.097 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & 0.23 & @xcite + 980703 & batse / rxte & 0.966 & 1.4 & & @xmath57 & @xmath58 & 0.15 & @xcite + 981226 & bepposax & ... & ... & ... & @xmath59 ( 10 h ) & @xmath60 & 0.06 & @xcite + 990123 & bepposax & @xmath611.6004 & 1.13 & 1.8 & @xmath62 & @xmath63 & 0.04 & @xcite + 990308 & batse / rxte & ... & @xmath11.2 & & @xmath64 & @xmath6128.4 & 0.07 & @xcite + 990506 & batse / rxte & 1.3 & ... & ... & @xmath65 ( 1 h ) & @xmath66 & 0.18 & @xcite + 990510 & batse / wfc & @xmath611.619 & 0.82 & 2.18 & @xmath67 & @xmath68 & 0.53 & @xcite + 990705 & bepposax & @xmath10.86 & 1.7 & @xmath612.6 & @xmath69 & @xmath70 & 0.20 & @xcite + 990712 & bepposax & 0.4331 & 0.97 & & @xmath71 & @xmath72 & 0.08 & @xcite + 991208 & ipn & 0.706 & 2.3@xmath73 & 3.2 & @xmath74 ( 2 d ) & @xmath75 & 0.04 & @xcite + 991216 & batse / rxte & 1.02 & 1.0 & 1.8 & @xmath76 & @xmath77 & 1.64 & @xcite + 000131 & ipn & 4.5 & 2.3 & & @xmath42 ( 3 d ) & @xmath6125.6 & 0.14 & @xcite + 000210 & bepposax / cxo & 0.846 & ... & ... & @xmath78 ( 12.4 hr ) & @xmath79 & 0.05 & @xcite + 000301c & ipn & 2.04 & 1.1 & 2.9 & @xmath80 ( 2 d ) & @xmath81 & 0.13 & @xcite + 000418 & ipn & 1.118 & 1.22 & & @xmath82 ( 3 d ) & @xmath83 & 0.08 & @xcite + 000630 & ipn & ... & 1.0 & & @xmath84 & @xmath85 & 0.03 & @xcite + 000911 & ipn & 1.058 & 1.4 & & @xmath86 ( 1.4 d ) & @xmath125 & 0.31 & @xcite + 000926 & ipn & 2.037 & 1.5 & 2.3 & @xmath87 & @xmath125 & 0.06 & @xcite + 001007 & ipn & ... & 2.05 & & 20.2 ( 3.5 d ) & @xmath88 & 0.11 & @xcite + 001011 & bepposax & ... & 1.33 & & @xmath58 & @xmath41 & 0.26 & @xcite + 001018 & ipn & ... & ... & ... & @xmath6122.6 ( @xmath612 d ) & @xmath89 & 0.06 & @xcite + 010222 & bepposax & 1.476 & 0.65 & 1.7 & @xmath90 & @xmath91 & 0.06 & @xcite + ccccccccc + grb & instr.@xmath29 & @xmath30 & @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & refs .
+ & & & & & ( mag ) & ( mag ) & ( mag ) & + + 010921 & hete-2 & 0.45 & 1.59 & & @xmath92 & @xmath93 & 0.38 & @xcite + 011030@xmath94 & bepposax & ... & ... & ... & @xmath6121 ( 8 hr ) & @xmath95 & 1.00 & @xcite + 011121 & bepposax & 0.36 & 1.7 & & @xmath96 & @xmath97 & 1.26 & @xcite + 011211 & bepposax & 2.14 & 0.83 & 1.7 & @xmath98 & @xmath99 & 0.11 & @xcite + + + + + + + + + + + + + + + + + | optical , infrared and ultraviolet observations of grb fields have allowed detection of counterparts and host galaxies of the high energy transients , thus crucially contributing to our present knowledge of the grb phenomenon .
measurements of afterglow variable emission , polarized light , redshifted absorption and emission line spectra , as well as host galaxies brightnesses and colors have clarified many fundamental issues related to the radiation mechanisms and environments of grbs , setting the background towards disclosing the nature of their progenitors . |
time - reversal techniques have attracted great attention recently due to the large variety of interesting potential applications .
the basic idea of time - reversal ( often also referred to as phase - conjugation in the frequency domain ) can be roughly described as follows : a localized source emits a short pulse of acoustic , electromagnetic or elastic energy which propagates through a richly scattering environment .
a receiver array records the arriving waves , typically as a long and complicated signal due to the complexity of the environment , and stores the time - series of measured signals in memory . after a short while
, the receiver array sends the recorded signals time - reversed ( i.e. first in - last out ) back into the same medium . due to the time - reversibility of the wave fields , the emitted energy backpropagates through the same environment , practically retracing the paths of the original signal , and refocuses , again as a short pulse , on the location where the source emitted the original pulse .
this , certainly , is a slightly oversimplified description of the rather complex physics which is involved in the real time - reversal experiment . in practice ,
the quality of the refocused pulse depends on many parameters , as for example the randomness of the background medium , the size and location of the receiver array , temporal fluctuations of the environment , etc .
surprisingly , the quality of the refocused signal increases with increasing complexity of the background environment .
this was observed experimentally by m. fink and his group at the laboratoire ondes et acoustique at universit paris vii in paris by a series of laboratory experiments ( see for example @xcite ) , and by w. a. kuperman and co - workers at the scripps institution of oceanography at university of california , san diego , by a series of experiments performed between 1996 and 2000 in a shallow ocean environment ( see for example @xcite ) .
the list of possible applications of this time - reversal technique is long .
in an iterated fashion , resembling the power method for finding maximal eigenvalues of a square matrix , the time - reversal method can be applied for focusing energy created by an ultrasound transducer array on strongly scattering objects in the region of interest .
this can be used for example in lithotripsy for localizing and destroying gall - stones in an automatic way , or more generally in the application of medical imaging problems .
detection of cracks in the aeronautic industry , or of submarines in the ocean are other examples .
see for example @xcite , and for related work @xcite . the general idea of time - reversibility , and its use in imaging and detection , is certainly not that new . looking into the literature for example of seismic imaging , the application of this basic idea can be found in a classical and very successful imaging strategy for detecting scattering interfaces in the earth , the so - called migration technique @xcite . however , the systematic use and investigation of the time - reversal phenomenon and its experimental realizations started more recently , and has been carried out during the last 1015 years or so by different research groups .
see for example @xcite for experimental demonstrations , and @xcite for theoretical and numerical approaches .
one very young and promising application of time - reversal is * communication*. in this paper we will mainly concentrate on that application , although the general results should carry over also to other applications as mentioned above .
the paper is organized as follows . in section [ tro.sec ]
we give a very short introduction into time - reversal in the ocean , with a special emphasis on underwater sound communication .
wireless communication in a mimo setup , our second main application in this paper , is briefly presented in section [ mimo.sec ] . in section [ sym.sec ]
we discuss symmetric hyperbolic systems in the form needed here , and examples of such systems are given in section [ exa.sec ] .
in section [ spa.sec ] , the basic spaces and operators necessary for our mathematical treatment are introduced . the inverse problem in communication , which we are focusing on in this paper ,
is then defined in section [ opt.sec ] . in section [ mat.sec ]
, we derive the basic iterative scheme for solving this inverse problem .
section [ adj.sec ] gives practical expressions for calculating the adjoint communication operator , which plays a key role in the iterative time - reversal schemes presented in this paper . in section [ atrm.sec ] the acoustic time - reversal mirror
is defined , which will provide the link between the acoustic time - reversal experiment and the adjoint communication operator .
the analogous results for the electromagnetic time - reversal mirror are discussed in section [ etrm.sec ] .
section [ tra.sec ] combines the results of these two sections , and explicitly provides the link between time - reversal and the adjoint imaging method .
sections [ gmc.sec ] , [ mns.sec ] , and [ rls.sec ] propose then several different iterative time - reversal schemes for solving the inverse problem of communication , using this key relationship between time - reversal and the adjoint communication operator .
the practically important issue of partial measurements ( and generalized measurements ) is treated in section [ pm.sec ] .
finally , section [ sfr.sec ] summarizes the results of this paper , and points out some interesting future research directions .
the ocean is a complex wave - guide for sound @xcite . in addition to scattering effects at the top and the bottom of the ocean , also its temperature profile and the corresponding refractive effects
contribute to this wave - guiding property and allow acoustic energy to travel large distances .
typically , this propagation has a very complicated behaviour .
for example , multipathing occurs if source and receiver of sound waves are far away from each other , since due to scattering and refraction there are many possible paths on which acoustic energy can travel between them .
surface waves and air bubbles at the top of the ocean , sound propagation through the rocks and sedimentary layers at the bottom of the ocean , and other effects further contribute to the complexity of sound propagation in the ocean .
when a source ( e.g. the base station of a communication system in the ocean ) emits a short signal at a given location , the receiver ( e.g. a user of this communication system ) some distance away from the source typically receives a long and complicated signal due to the various influences of the ocean to this signal along the different connecting paths . if the base station wants to communicate with the user by sending a series of short signals , this complex response of the ocean to the signal needs to be resolved and taken into account . in a classical communication system , the base station which wants to communicate with a user broadcasts a series of short signals ( e.g. a series of zeros and ones ) into the environment
the hope is that the user will receive this message as a similar series of temporally well - resolved short signals which can be easily identified and decoded .
however , this almost never occurs in a complex environment , due to the multipathing and the resulting delay - spread of the emitted signals .
typically , when the base station broadcasts a series of short signals into such an environment , intersymbol interference occurs at the user position due to the temporal overlap of the multipath contributions of these signals . in order to recover the individual signals ,
a significant amount of signal processing is necessary at the user side , and , most importantly , the user needs to have some knowledge of the propagation behaviour of the signals in this environment ( i.e. he needs to know the channel ) .
intersymbol interference can in principle be avoided by adding a sufficient delay between individual signals emitted at the base station which takes into account the delay - spread in the medium .
that , however , slows down the communication , and reduces the capacity of the environment as a communication system .
an additional drawback of simply broadcasting communication signals from the base station is the obvious lack of interception security .
a different user who also knows the propagation behaviour of signals in the environment , can equally well resolve the series of signals arriving at his location and decode them .
several approaches have been suggested to circumvent the above mentioned drawbacks of communication in multiple - scattering environments .
some very promising techniques are based on the time - reversibility property of propagating wave - fields @xcite .
the basic idea is as follows .
the user who wants to communicate with the base station , starts the communication process by sending a short pilot signal through the environment .
the base station receives this signal as a long and complex signal due to multipathing .
it time - reverses the received signal and sends it back into the environment .
the backpropagating waves will produce a complicated wave - field everywhere due to the many interfering parts of the emitted signal .
however , due to the time - reversibility of the wave - fields , we expect that the interference will be constructive at the position of the user who sent the original pilot signal , and mainly destructive at all other positions .
therefore , the user will receive at his location a short signal very similar to ( ideally , a time - reversed replica of ) the pilot signal which he sent for starting the communication .
all other users who might be in the environment at the same time will only receive noise speckle due to incoherently interfering contributions of the backpropagating field .
if the base station sends the individual elements ( ones and zeros ) of the intended message in a phase - encoded form as a long overlapping string of signals , the superposition principle will ensure that , at the user position , this string of signals will appear as a series of short well - separated signals , each resembling some phase - shifted ( and time - reversed ) form of the pilot signal . in order to find out whether this theoretically predicted
scenario actually takes place in a real multiple - scattering environment like the ocean , kuperman have performed a series of four experiments in a shallow ocean environment between 1996 and 2000 , essentially following the above described scenario .
the experiments have been performed at a mediterranean location close to the italian coast .
( a similar setup was also used in an experiment performed in 2001 off the coast of new england which has been reported in yang @xcite . ) a schematic view of these experiments is shown in figure [ figure1 ] .
the single user is replaced here by a probe source , and the source - receive array ( sra ) plays the role of the base station. an additional vertical receive array ( vra ) was deployed at the position of the probe source in order to measure the temporal and spatial spread of the backpropagating fields in the neighbourhood of the probe source location . in this shallow environment ( depths of about @xmath0 m , and distances between @xmath1 and @xmath2 km )
the multipathing of the waves is mostly caused by multiple reflections at the surface and the bottom of the ocean .
the results of the experiments have been reported in @xcite .
they show that in fact a strong spatial and temporal focusing of the backpropagating waves occurs at the source position . a theoretical explanation of the temporal and spatial refocusing of time - reversed waves in random environments has been given in blomgren @xcite .
the underwater acoustics scenario described above directly carries over to situations which might be more familiar to most of us , namely to the more and more popular wireless communication networks using mainly electromagnetic waves in the microwave regime . starting from the everyday use of cell - phones , ranging to small wireless - operating local area networks ( lan ) for computer systems or for private enterprise communication systems ,
exactly the same problems arise as in underwater communication .
the typically employed microwaves of a wavelength at about 1030 cm are heavily scattered by environmental objects like cars , fences , trees , doors , furniture etc .
this causes a very complicated multipath structure of the signals received by users of such a communication system .
since bandwidths are limited and increasingly expensive , a need for more and more efficient communication systems is imminent .
recently , the idea of a so - called multiple - input multiple - output ( mimo ) communication system has been introduced with the potential to increase the capacity and efficiency of wireless communication systems @xcite .
the idea is to replace a single antenna at the base station which is responsible for multiple users , or even a system where one user communicates only with his own dedicated base antenna , by a more general system where an array of multiple antennas at the base station is interacting simultaneously and in a complex way with multiple users .
a schematic description of such a mimo system ( with seven base antennas and seven users ) is given in figure [ figure2 ] .
see for example @xcite for recent overviews on mimo technology .
time - reversal techniques are likely to play also here a key role in improving communication procedures and for optimizing the use of the limited resources ( especially bandwidth ) which are available for this technology @xcite .
one big advantage of time - reversal techniques is that they are automatically adapted to the complex environment and that they can be very fast since they do not require heavy signal processing at the receiver or user side . in @xcite , an * iterated time - reversal scheme * for the optimal refocusing of signals in such a mimo communication system
was proposed , which we will describe in more details in section [ gmc.sec ] . in the present paper
we will establish a direct link between the time - reversal technique and solution strategies for inverse problems . as an application of this relationship
, we will derive iterative time - reversal schemes for the optimization of wireless or underwater acoustic mimo communication systems .
the derivation is performed completely in time - domain , for very general first order symmetric hyperbolic systems describing wave propagation phenomena in a complex environment .
one of the schemes which we derive , in a certain sense the basic one , will turn out to be practically equivalent to the scheme introduced in @xcite , although the derivation uses different tools .
therefore , it provides a new interpretation of that scheme . the other schemes which we introduce
are new in this application , and can be considered as either generalizations of the basic scheme , or as independent alternatives to that scheme .
each of them addresses slightly different objectives and has its own very specific characteristics .
we treat wave propagation in communication systems in the general framework of symmetric hyperbolic systems of the form @xmath3 @xmath4 here , @xmath5 and @xmath6 are real - valued time - dependent @xmath7-vectors , @xmath8 , and @xmath9 $ ] .
@xmath10 is a real , symmetric , uniformly positive definite @xmath11-matrix , i.e. , @xmath12 for some @xmath13 .
moreover , @xmath14 is a symmetric positive semi - definite @xmath11 matrix , i.e. , @xmath15 .
it models possible energy loss through dissipation in the medium .
the @xmath16 are real , symmetric and independent of @xmath17 .
we will also use the short notation @xmath18 in addition to the above mentioned assumptions on the coefficients @xmath10 and @xmath14 , we will assume throughout this paper that all quantities @xmath10 , @xmath14 , @xmath5 and @xmath6 are sufficiently regular in order to safely apply for example integration by parts and green s formulas .
for details see for example @xcite . throughout this paper
we will assume that no energy reaches the boundaries @xmath19 during the time @xmath20 $ ] , such that we will always have @xmath21.\ ] ] the _ energy density _
@xmath22 is defined by @xmath23 the _ total energy _
@xmath24 in @xmath25 at a given time @xmath26 is therefore @xmath27 the _ flux _ @xmath28 is given by @xmath29
in the following , we want to give some examples for symmetric hyperbolic systems as defined above .
of special interest for communication are the system of the linearized acoustic equations and the system of maxwell s equations .
we will discuss these two examples in detail in this paper .
another important example for symmetric hyperbolic systems is the system of elastic waves equations , which we will however leave out in our discussion for the sake of shortness .
we only mention here that all the results derived here apply without restrictions also to linear elastic waves .
elastic wave propagation becomes important for example in ocean acoustic communication models which incorporate wave propagation through the sedimentary and rock layers at the bottom of the ocean . as a model for * underwater sound propagation * , we consider the following linearized form of the acoustic equations in an isotropic medium @xmath30 @xmath31 here , @xmath32 is the velocity , @xmath33 the pressure , @xmath34 the density , and @xmath35 the compressibility .
we have @xmath36 , @xmath37 and @xmath38 ( where @xmath39 as a superscript always means transpose ) .
moreover , we have @xmath40 with the notation @xmath41 , we can write @xmath42 as @xmath43 the operators @xmath16 , @xmath44 , can be recovered here from @xmath42 by putting @xmath45 the energy density @xmath22 is given by @xmath46 and the energy flux @xmath28 is @xmath47 we mention that the dissipative case @xmath48 can be treated as well in our framework , and yields analogous results to those presented here . as a second example
, we will consider maxwell s equations for an anisotropic medium with some energy loss due to the inherent conductivity .
this can for example model * wireless communication * in a complex environment .
@xmath50 we have @xmath51 , and @xmath52 , @xmath53 .
moreover , @xmath54 here , @xmath55 and @xmath56 are symmetric positive definite @xmath57 matrices modelling the anisotropic permittivity and permeability distribution in the medium , and @xmath58 is a symmetric positive semi - definite @xmath57 matrix which models the anisotropic conductivity distribution . in wireless communication , this form can model for example dissipation by conductive trees , conductive wires , rainfall , pipes , etc .
the operator @xmath42 can be written in block form as @xmath59 with @xmath60 the operators @xmath16 , @xmath44 , can be recovered here from @xmath42 by putting @xmath61 the energy density @xmath22 is given by @xmath62 the energy flux @xmath28 is described by the _ poynting vector _
@xmath63 as already mentioned , also elastic waves can be treated in the general framework of symmetric hyperbolic systems . for more details
we refer to @xcite .
for our mathematical treatment of time - reversal we introduce the following spaces and inner products .
we assume that we have @xmath64 users @xmath65 , @xmath66 in our system , and in addition a base station which consists of @xmath67 antennas @xmath68 , @xmath69 .
each user and each antenna at the base station can receive , process and emit signals which we denote by @xmath70 for a given user @xmath65 , @xmath66 , and by @xmath71 for a given base antenna @xmath68 , @xmath69 .
each of these signals consists of a time - dependent @xmath7-vector indicating measured or processed signals of time - length @xmath39 . in our analysis
we will often have to consider functions defined on the time interval @xmath72 $ ] instead of @xmath20 $ ] .
for simplicity , we will use the same notation for the function spaces defined on @xmath72 $ ] as we use for those defined on @xmath20 $ ] .
it will always be obvious which space we refer to in a given situation .
lumping together all signals at the users on the one hand , and all signals at the base station on the other hand , yields the two fundamental quantities @xmath73 with @xmath74)^n\big)^{{j } } , \qquad z\,=\,\big(l_2([0,t])^n\big)^{{k}}.\ ] ] the two signal spaces @xmath75 and @xmath76 introduced above are equipped with the inner products @xmath77}\big\langle { { \bf s}}^{(1)}_j(t ) , { { \bf s}}^{(2)}_j(t)\big\rangle_n\,dt \\
\left\langle{{\bf r}}^{(1)},{{\bf r}}^{(2)}\right\rangle_{z}&= & \sum_{k=1}^{{k}}\int_{[0,t]}\big\langle { { \bf r}}^{(1)}_k(t ) , { { \bf r}}^{(2)}_k(t)\big\rangle_n\,dt.\end{aligned}\ ] ] the corresponding norms are @xmath78 each user @xmath65 and each antenna @xmath68 at the base station can send a given signal @xmath70 or @xmath71 , respectively .
this gives rise to a source distribution @xmath79 , @xmath66 , or @xmath80 , @xmath69 , respectively . here and
in the following we will use in our notation the following convention .
if one symbol appears in both forms , with and without a hat ( ^ ) on top of this symbol , then all quantities _ with _ the hat symbol are related to the users , and those _ without _ the hat symbol to the antennas at the base station .
each of the sources created by a user or by a base antenna will appear on the right hand side of ( [ sym.1 ] ) as a mathematical source function and gives rise to a corresponding wave field which satisfies ( [ sym.1 ] ) , ( [ sym.2 ] ) .
when solving the system ( [ sym.1 ] ) , ( [ sym.2 ] ) , typically certain sobolev spaces need to be employed for the appropriate description of the underlying function spaces ( see for example @xcite ) . for our purposes
, however , it will be sufficient to assume that both , source functions and wave fields , are members of the following canonical function space @xmath81 which is defined as @xmath82)^{n}\,,\,{{\bf u}}=0\ ; \mbox{on}\;\partial \omega\times [ 0,t]\,,\,\;\|{{\bf u}}\|_{u}<\infty\},\ ] ] and which we have equipped with the usual energy inner product @xmath83}\int_{\omega}\left\langle\gamma({{\bf x}}){{\bf u}}({{\bf x}},t ) , { { \bf v}}({{\bf x}},t)\right\rangle_n\,d{{\bf x}}dt,\ ] ] and the corresponding energy norm @xmath84 also here , in order to simplify the notation , we will use the same space when considering functions in the shifted time interval @xmath72 $ ] instead of @xmath20 $ ] .
typically , when a user or an antenna at the base station transforms a signal into a source distribution , it is done according to a very specific antenna characteristic which takes into account the spatial extension of the user or the antenna .
we will model this characteristic at the user by the functions @xmath85 , @xmath66 , and for base antennas by the functions @xmath86 , @xmath69 . with these functions
, we can introduce the linear source operators @xmath87 and @xmath88 mapping signals @xmath89 at the set of users and @xmath90 at the set of base antennas into the corresponding source distributions @xmath91 and @xmath6 , respectively .
they are given as @xmath92 we will assume that the functions @xmath85 are supported on a small neighbourhood @xmath93 of the user location @xmath94 , and that the functions @xmath86 are supported on a small neighbourhood @xmath95 of the antenna location @xmath96 . moreover , all these neighbourhoods are strictly disjoint to each other . for example
, the functions @xmath85 could be assumed to be @xmath97-approximations of the dirac delta measure @xmath98 concentrated at the user locations @xmath94 , and the functions @xmath86 could be assumed to be @xmath97-approximations of the dirac delta measure @xmath99 concentrated at the antenna locations @xmath96 .
both , users and base antennas can also record incoming fields @xmath100 and transform the recorded information into signals .
also here , this is usually done according to very specific antenna characteristics of each user and each base antenna . for simplicity
( and without loss of generality ) , we will assume that the antenna characteristic of a user or base antenna for receiving signals is the same as for transmitting signals , namely @xmath85 for the user and @xmath86 for a base antenna .
( the case of more general source and measurement operators is discussed in section [ pm.sec ] . ) with this , we can define the linear measurement operators @xmath101 and @xmath102 , respectively , which transform incoming fields into measured signals , by @xmath103 finally , we define the linear operator @xmath104 mapping sources @xmath105 to states @xmath106 by @xmath107 where @xmath106 solves the problem ( [ sym.1 ] ) , ( [ sym.2 ] ) . as already mentioned , we assume that the domain @xmath25 is chosen sufficiently large and that the boundary @xmath108 is sufficiently far away from the users and base antennas , such that there is no energy reaching the boundary in the time interval @xmath20 $ ] ( or @xmath72 $ ] ) due to the finite speed of signal propagation . therefore , the operator @xmath104 is well - defined .
formally , we can now introduce the two linear * communication operators * @xmath109 and @xmath110 which are at the main focus of this paper .
they are defined as @xmath111 the operator @xmath109 models the following situation .
the base station emits the signal @xmath112 which propagates through the complex environment . the users measure the arriving wave fields and transform them into measurement signals
the measured signals at the set of all users is @xmath113 .
the operator @xmath110 describes exactly the reversed situation .
all users emit together the set of signals @xmath114 , which propagate through the given complex environment and are received by the base station .
the corresponding set of measured signals at all antennas of the base station is just @xmath115 .
no time - reversal is involved so far .
in the following , we outline a typical problem arising in communication , which gives rise to a mathematically well - defined inverse problem .
a specified user of the system , say @xmath116 , defines a ( typically but not necessarily short ) pilot signal @xmath117 which he wants to use as a template for receiving the information from the base station .
the base station wants to emit a signal @xmath118 which , after having travelled through the complex environment and arriving at the user @xmath116 , matches this pilot signal as closely as possible .
neither the base station nor any other user except of @xmath116 are required ( or expected ) to know the correct form of the pilot signal @xmath117 for this problem . as an additional constraint
, the base station wants that at the other users @xmath65 , @xmath119 , as little energy as possible arrives when communicating with the specified user @xmath116 .
this is also in the interest of the other users , who want to use a different channel for communicating at the same time with the base antenna , and want to minimize interference with the communication initiated by user @xmath116 .
the complex environment itself in which the communication takes place ( i.e. the channel ) is assumed to be unknown to all users and to the base station . in order to arrive at a mathematical description of this problem
, we define the ideal signal @xmath120 received by all users as @xmath121 each user only knows his own component of this signal , and the base antenna does not need to know any component of this ideal signal at all . * the inverse problem of communication*:in the terminology of inverse problems , the above described scenario defines an inverse source problem , which we call for the purpose of this paper the inverse problem of communication. the goal is to find a source distribution @xmath122 at the base station which satisfies the data @xmath123 at the users : @xmath124 the state equation relating sources to data is given by the symmetric hyperbolic system ( [ sym.1 ] ) , ( [ sym.2 ] ) .
notice that the basic operator @xmath109 in ( [ opt.2 ] ) is unknown to us since we do not know the complicated medium in which the waves propagate .
if the operator @xmath109 ( together with @xmath125 ) would be known at the base station by some means , the inverse source problem formulated above could be solved using classical inverse problems techniques , which would be computationally expensive but in principle doable . in our situation
, we are able to do physical experiments , which amounts to applying the communication operator @xmath109 to a given signal .
determining the operator @xmath109 explicitly by applying it to a set of basis functions of @xmath75 would be possible , but again too expensive .
we will show in the following that , nevertheless , many of the classical solution schemes known from inverse problems theory can be applied in our situation even without knowing this operator .
the basic tool which we will use is a series of time - reversal experiments , applied to carefully designed signals at the users and the base station .
a practical template for an * iterative scheme for finding an optimal signal * at the base station can be envisioned as follows .
user @xmath116 starts the communication process by emitting an initial signal @xmath126 into the complex environment .
this signal , after having propagated through the complex environment , finally arrives at the base station and is received there usually as a relatively long and complicated signal due to the multiple scattering events it experienced on its way .
when the base station receives such a signal , it processes it and sends a new signal @xmath127 back through the environment which is received by all users . after receiving this signal , all
users become active .
the user @xmath116 compares the received signal with the pilot signal .
if the match is not good enough , it processes the received signal in order to optimize the match with the pilot signal when receiving the next iterate from the base station .
all other users identify the received signal as unwanted noise , and process it with the goal to receive in the next iterate from the base station a signal with lower amplitude , such that it does not interfere with their own communications .
all users send now their processed signals , which together define @xmath128 , back to the base station .
the base station receives them all at the same time , again usually as a long and complicated signal , processes this signal and sends a new signal
@xmath129 back into the environment which ideally will match the desired signals at all users better than the previously emitted signal @xmath127 .
this iteration stops when all users are satisfied , i.e. , when user @xmath116 receives a signal which is sufficiently close to the pilot signal @xmath117 , and the energy or amplitude of the signals arriving at the other users has decreased enough in order not to disturb their own communications . after this learning process of the channel has been completed , the user @xmath116 can now start communicating safely with the base antenna using the chosen pilot signal @xmath117 for decoding the received signals .
similar schemes have been suggested in @xcite in a single - step fashion , and in @xcite performing multiple steps of the iteration .
the main questions to be answered are certainly which signals each user and each base antenna needs to emit in each step , how these signals need to be processed , at which stage this iteration should be terminated , and which optimal solution this scheme is expected to converge to .
_ one goal of this paper is to provide a theoretical framework for answering these questions by combining basic concepts of inverse problems theory with experimental time - reversal techniques . _
a standard approach for solving problem ( [ opt.2 ] ) in the situation of noisy data is to look for the least - squares solution @xmath130 in order to practically find a solution of ( [ mat.2 ] ) , we introduce the cost functional @xmath131 in the * basic approach * we propose to use the * gradient method * for finding the minimum of ( [ mat.3 ] ) . in this method , in each iteration a correction @xmath132 is sought for a guess @xmath90 which points into the negative gradient direction @xmath133 of the cost functional ( [ mat.3 ] ) . in other words , starting with the initial guess @xmath134 , the iteration of the gradient method goes as follows : @xmath135 where @xmath136 is the step - size at iteration number @xmath137 . notice
that the signals @xmath138 are measured at the base station , whereas the difference @xmath139 is determined at the users .
in particular , @xmath140 is the pilot signal only known by the user who defined it , combined with zero signals at the remaining users .
@xmath141 is the signal received by the users at the @xmath137-th iteration step .
we see from ( [ mat.4 ] ) that we will have to apply the adjoint operator @xmath142 repeatedly when implementing the gradient method . in this section
we provide practically useful expressions for applying this operator to a given element of @xmath76 .
[ lemma.adj.1 ] we have @xmath143 _ proof : _ this follows from @xmath144.@xmath145 [ theorem.adj.1 ] we have @xmath146 _ proof : _ the proof is given in appendix a.@xmath145 next , we want to find an expression for @xmath147 , @xmath148 , where @xmath149 is the adjoint of the operator @xmath104 .
[ theorem.adj.2 ] let @xmath150 be the solution of the adjoint symmetric hyperbolic system @xmath151 @xmath152 @xmath153.\ ] ] then @xmath154 * proof : * the proof is given in appendix b.@xmath145 this procedural characterization of the adjoint operator is often used in solution strategies of large scale inverse problems , where it naturally leads to so - called backpropagation strategies. see for example @xcite and the references given there .
notice that in the adjoint system ( [ adj.3])([adj.5 ] ) final value conditions are given at @xmath155 in contrast to ( [ sym.1])([sym.4 ] ) where initial value conditions are prescribed at @xmath156 .
this corresponds to the fact that time is running backward in ( [ adj.3])([adj.5 ] ) and forward in ( [ sym.1])([sym.4 ] ) .
in the following we want to define an operator @xmath157 such that @xmath158 holds . we will call this operator @xmath157 the *acoustic time - reversal operator*. we will also define the *acoustic time - reversal mirrors* @xmath159 and @xmath160 , which act on the signals instead of the sources or fields . we consider the acoustic system @xmath161 @xmath162 with @xmath9 $ ] and zero boundary conditions at @xmath163 $ ] .
we want to calculate the action of the _ adjoint operator _
@xmath149 on a vector @xmath164 .
[ theorem.atrm.1 ] let @xmath164 and let @xmath165 be the solution of the adjoint system @xmath166 @xmath167 with @xmath9 $ ] and zero boundary conditions at @xmath163 $ ] .
then we have @xmath168 _ proof:_this theorem is just an application of theorem [ theorem.adj.2 ] to the acoustic symmetric hyperbolic system . for the convenience of the reader , we will give a direct proof as well in appendix c. we define the * acoustic time - reversal operator * @xmath157 by putting for all @xmath169 ( and similarly for all @xmath170 ) @xmath171 we define the * acoustic time - reversal mirrors * @xmath159 and @xmath160 by putting for all @xmath172 and all @xmath173 @xmath174 the following lemma is easy to verify .
[ lemma.atrm.1 ] we have the following commutations @xmath175 [ theorem.atrm.2 ] for @xmath164 we have @xmath176 _ proof:_for the proof it is convenient to make the following definition . *
acoustic time - reversal experiment : * for given @xmath164 define @xmath177 and @xmath178 and perform the following physical experiment @xmath179 @xmath180 with @xmath181 $ ] and zero boundary conditions .
doing this experiment means to process the data in the following way : time - reverse all data @xmath182 according to @xmath183 , @xmath9 $ ] , and , in addition , reverse the directions of the velocities @xmath184 .
we call this experiment the acoustic time - reversal experiment. notice that the time is running forward in this experiment
. the solution @xmath185 of this experiment can obviously be represented by @xmath186 in order to show that the so defined time - reversal experiment is correctly modelled by the adjoint system derived above , we make the following change in variables : @xmath187 which just corresponds to the application of the operator @xmath157 to @xmath185 .
we have @xmath188 $ ] . in these variables the time - reversal system ( [ atrm.11])-([atrm.13 ] ) gets the form @xmath189 @xmath190 taking into account the definition of @xmath191 and @xmath192 , we see that @xmath193 where @xmath194 and @xmath195 solve the adjoint system ( [ atrm.4])([atrm.6 ] ) . therefore , according to theorem [ theorem.atrm.1 ] : @xmath196 since we have with ( [ atrm.14 ] ) , ( [ atrm.15 ] ) also @xmath197 the theorem is proven .
in the following we want to define an operator @xmath198 such that @xmath199 holds . we will call this operator @xmath198 the *electromagnetic time - reversal operator*. we will also define the *electromagnetic time - reversal mirrors* @xmath200 and @xmath201 , which act on the signals instead of the sources or fields . we consider maxwell s equations @xmath202 @xmath203 with @xmath9 $ ] and zero boundary conditions .
we want to calculate the action of the _ adjoint operator _
@xmath149 on a vector @xmath164 .
[ theorem.etrm.1 ] let @xmath164 and let @xmath204 be the solution of the adjoint system @xmath205 @xmath206 with @xmath9 $ ] and zero boundary conditions .
then we have @xmath207 _ proof:_this theorem is just an application of theorem [ theorem.adj.2 ] to the electromagnetic symmetric hyperbolic system .
again , we will give a direct proof in appendix d as well .
we define the * electromagnetic time - reversal operator * @xmath198 by putting for all @xmath208 ( and similarly for all @xmath209 ) @xmath210 we define the * electromagnetic time - reversal mirrors * @xmath200 and @xmath201 by putting for all @xmath211 and for all @xmath212 @xmath213 the following lemma is easy to verify .
[ lemma.etrm.1 ] we have the following commutations @xmath214 [ theorem.etrm.2 ] for @xmath164 we have @xmath215 _ proof:_for the proof it is convenient to make the following definition . *
electromagnetic time - reversal experiment : * for a given vector @xmath164 define @xmath216 and @xmath217 and perform the physical experiment @xmath218 @xmath219 with @xmath181 $ ] and zero boundary conditions .
doing this experiment means to process the data in the following way : time - reverse all data according to @xmath183 , @xmath9 $ ] , and , in addition , reverse the directions of the electric field component by @xmath184 .
we call this experiment the electromagnetic time - reversal experiment. notice that the time is running forward in this experiment
. the solution @xmath220 of this experiment can obviously be represented by @xmath221 in order to show that the so defined time - reversal experiment is correctly modelled by the adjoint system derived above , we make the following change in variables : @xmath222 which just corresponds to the application of the operator @xmath198 to @xmath220 .
we have @xmath188 $ ] . in these variables
the time - reversal system ( [ etrm.11])([etrm.13 ] ) gets the form @xmath223 @xmath224 taking into account the definition of @xmath225 and @xmath226 , we see that @xmath227 where @xmath228 and @xmath229 solve the adjoint system ( [ etrm.4])([etrm.6 ] ) .
therefore , according to theorem [ theorem.etrm.1 ] : @xmath230 since we have with ( [ etrm.14 ] ) , ( [ etrm.15 ] ) also @xmath231 the theorem is proven .
@xmath145 + for electromagnetic waves , there is formally an alternative way to define the _ electromagnetic time - reversal operator _ ,
namely putting for all @xmath208 @xmath232 accompanied by the analogous definitions for the _ electromagnetic time - reversal mirrors_. with these alternative definitions , theorem [ theorem.etrm.2 ] holds true as well , with only very few changes in the proof .
which form to use depends mainly on the preferred form for modelling applied antenna signals in the given antenna system .
the first formulation directly works with applied electric currents , whereas the second form is useful for example for magnetic dipole sources .
define @xmath233 , @xmath234 , @xmath235 for the acoustic case , and @xmath236 , @xmath237 , @xmath238 for the electromagnetic case .
we call @xmath239 the * time - reversal operator * and @xmath240 , @xmath241 the * time - reversal mirrors*. we combine the results of lemma [ lemma.atrm.1 ] and lemma [ lemma.etrm.1 ] into the following lemma .
[ lemma.tra.1 ] we have the following commutations @xmath242 moreover , combining theorem [ theorem.atrm.2 ] and theorem [ theorem.etrm.2 ] we get [ theorem.tra.1 ] for @xmath164 we have @xmath243 with this , we can prove the following theorem which provides the fundamental link between time - reversal and inverse problems .
[ theorem.tra.2 ] we have @xmath244 _ proof : _ recall that the adjoint operator @xmath142 can be decomposed as @xmath245 . with theorem [ theorem.adj.1 ] , theorem [ theorem.tra.1 ] , and lemma [ lemma.tra.1 ] , it follows therefore that @xmath246 which proves the theorem .
@xmath145 the above theorem provides a direct link between the adjoint operator @xmath142 , which plays a central role in the theory of inverse problems , and a physical experiment modelled by @xmath110 .
the expression @xmath247 defines a time - reversal experiment. we will demonstrate in the following sections how we can make use of this relationship in order to solve the inverse problem of communication by a series of physical time - reversal experiments .
we only mention here that the above results hold as well for * elastic waves * with a suitable definition of the * elastic time - reversal mirrors*. we leave out the details for brevity .
the results achieved above give rise to the following experimental procedure for applying the gradient method ( [ mat.4 ] ) to the inverse problem of communication as formulated in sections [ opt.sec ] and [ mat.sec ] .
first , the pilot signal @xmath120 is defined by user @xmath116 as described in ( [ mat.1 ] ) .
moreover , we assume that the first guess @xmath248 at the base station is chosen to be zero . then , using theorem [ theorem.tra.2 ] , we can write the gradient method ( [ mat.4 ] ) in the equivalent form @xmath249 or , expanding it , @xmath250 in a more detailed form , we arrive at the following * experimental procedure for implementing the gradient method * , where we fix in this description @xmath251 for all iteration numbers @xmath137 for simplicity . 1 .
the user @xmath116 chooses a pilot signal @xmath117 which he wants to use for communicating with the base station .
the objective signal at all users is then @xmath252 . the initial guess @xmath248 at the base station is defined to be zero , such that @xmath253 .
2 . the user @xmath116 initiates the communication by sending the time - reversed pilot signal into the environment .
this signal is @xmath254 .
all other users are quiet .
3 . the base station receives the pilot signal as @xmath255 .
it time - reverses this signal and sends this time - reversed form , namely @xmath256 , back into the medium .
the new signal arrives at all users as @xmath257 .
all users compare the received signals with their components of the objective signal @xmath120 .
they take the difference @xmath258 , and time - reverse it .
they send this new signal @xmath259 back into the medium . 5 . the base station receives this new signal , time - reverses it , adds it to the previous signal @xmath127 , and sends the sum back into the medium as @xmath129 . 6 .
this iteration is continued until all users are satisfied with the match between the received signal @xmath260 and the objective signal @xmath120 at some iteration number @xmath137 .
alternatively , a fixed iteration number @xmath137 can be specified a - priori for stopping the iteration .
needless to say that , in practical implementations , the laboratory time needs to be reset to zero after each time - reversal step .
the experimental procedure which is described above is practically equivalent to the experimental procedure which was suggested and experimentally verified in @xcite .
therefore , our basic scheme provides an alternative derivation and interpretation of this experimental procedure .
we mention that several refinements of this scheme are possible and straightforward .
for example , a weighted inner product can be introduced for the user signal space @xmath76 which puts different preferences on the satisfaction of the user objectives during the iterative optimization process .
for example , if the importance of suppressing interferences with other users is valued higher than to get an optimal signal quality at the specified user @xmath116 , a higher weight can be put into the inner product at those users which did not start the communication process . a user who does not care about these interferences , simply puts a very small weight into his component of the inner product of @xmath76 .
notice that there is no mechanism directly built into this procedure which prevents the energy emitted by the base antenna to increase more than the communication system can support .
for example , if the subspace of signals @xmath261 is not empty , then it might happen that during the iteration described above ( e.g. due to noise ) an increasing amount of energy is put into signals emitted by the base station which are in this subspace and which all produce zero contributions to the measurements at all users .
more generally , elements of the subspace of signals @xmath262 for a very small threshold @xmath263 , might cause problems during the iteration if the pilot signal @xmath120 chosen by the user has contributions in the subspace @xmath264 ( i.e. in the space of all @xmath265 with @xmath266 ) .
this is so because in the effort of decreasing the mismatch between @xmath267 and @xmath120 , the base antenna might need to put signals with high energy into the system in order to get only small improvements in the signal match at the user side . since the environment ( and therefore the operator @xmath109 ) is unknown a - priori , it is difficult to avoid the existence of such contributions in the pilot signal .
one possible way to prevent the energy emitted by the base station to increase artificially would be to project the signals @xmath268 onto the orthogonal complements of the subspaces @xmath269 or @xmath270 ( if they are known or can be constructed by some means ) prior to their emission .
alternatively , the iteration can be stopped at an early stage before these unwanted contributions start to build up .
( this in fact has been suggested in @xcite ) . in the following subsection
we introduce an alternative way of ensuring that the energy emitted by the base station stays reasonably bounded in the effort of fitting the pilot signal at the users .
consider the regularized problem @xmath271 with some suitably chosen regularization parameter @xmath272 . in this problem formulation a trade - off is sought between a signal fit at the user side and a minimized energy emission at the base station .
the trade - off parameter is the regularization parameter @xmath273 .
instead of ( [ mat.3 ] ) we need to consider now @xmath274 the negative gradient direction is now given by @xmath275 , such that the regularized iteration reads : @xmath276 where we have replaced @xmath142 by @xmath247 .
the time - reversal iteration can be expanded into the following practical scheme @xmath277 comparing with ( [ gmc.2 ] ) , we see that the adaptations which need to be applied in the practical implementation for stabilizing the basic algorithm can easily be done .
in this section we want to propose an alternative scheme for solving the inverse problem of communication . as mentioned above
, a major drawback of the basic approach ( [ mat.2 ] ) is that the energy emitted by the base station is not limited explicitly when solving the optimization problem .
the regularized version presented above alleviates this problem .
however , we want to mention here that , under certain assumptions , there is an alternative scheme which can be employed instead and which has an energy constraint directly built in . under the formal assumption that there exists at least one ( and presumably more than one ) solution of the inverse problem at hand ( i.e. the formally underdetermined case )
, we can look for the * minimum norm solution * @xmath278 in hilbert spaces this solution has an explicit form .
it is @xmath279 here , the operator @xmath280 acts as a filter on the pilot signal @xmath125 . instead of sending the pilot signal to the base station , the users send the filtered version of it .
certainly , a method must be found in order to apply the filter @xmath280 to the pilot signal .
one possibility of doing so would be to try to determine the operator @xmath281 explicitly by a series of time - reversal experiments on some set of basis functions of @xmath76 , and then invert this operator numerically
. however , this might not be practical in many situations .
( it certainly would be slow and it would involve a significant amount of signal - processing , which we want to avoid here . ) therefore , we propose an alternative procedure .
first , we notice that there is no need to determine the whole operator @xmath280 , but that we only have to apply it to one specific signal , namely @xmath125 .
let us introduce the short notation @xmath282 in this notation , we are looking for a signal @xmath283 such that @xmath284 .
we propose to solve this equation in the least squares sense : @xmath285 moreover , as a suitable method for practically finding this solution , we want to use the * gradient method*. starting with the initial guess @xmath286 , the gradient method reads @xmath287 where @xmath288 is the adjoint operator to @xmath289 and @xmath136 is again some step - size . expanding this expression , and taking into account @xmath290 and @xmath291 ,
we arrive at @xmath292 in the practical implementation , we arrive at the following iterative scheme : @xmath293 this iteration can be implemented by a series of time - reversal experiments , without the need of heavy signal - processing
. the final step of the above algorithm amounts to applying @xmath142 to the result of the gradient iteration for calculating @xmath294 , which yields then @xmath295 .
this will then be the signal to be applied by the base station during the communication process with the user @xmath116 .
in some situations it might be expected that the operator @xmath289 is ill - conditioned , such that its inversion might cause instabilities , in particular when noisy signals are involved . for those situations , a regularized form of
the minimum norm solution is available , namely @xmath296 where @xmath297 denotes the identity operator in @xmath76 and @xmath272 is some suitably chosen regularization parameter .
the necessary adjustments in the gradient iteration for applying @xmath298 to @xmath299 are easily done .
we only mention here the resulting procedure for the implementation of this gradient method by a series of time - reversal experiments : @xmath300 again , the last step shown above is a final application of @xmath142 to the result of the gradient iteration for calculating @xmath301 , which yields then @xmath302 .
this will then be the signal to be applied by the base station during the communication process with the user @xmath116 .
we have introduced above the regularized least squares solution of the inverse problem of communication , namely @xmath303 with @xmath272 being the regularization parameter . in hilbert spaces ,
the solution of ( [ rls.1 ] ) has an explicit form .
it is @xmath304 where @xmath305 is the identity operator in @xmath75 .
it is therefore tempting to try to implement also this direct form as a series of time - reversal experiments and compare its performance with the gradient method as it was described above . as our last strategy
which we present in this paper , we want to show here that such an alternative direct implementation of ( [ rls.1 ] ) is in fact possible .
notice that in ( [ rls.1 ] ) the filtering operator @xmath306 is applied at the base station , in contrast to the previous case where the user signal was filtered by the operator @xmath307 .
analogously to the previous case , we need to find a practical way to apply this filter to a signal at the base station .
we propose again to solve the equation @xmath308 in the least squares sense , where @xmath309 .
defining @xmath310 and using @xmath290 and @xmath291 , we arrive at the following * gradient iteration * for solving problem ( [ rls.3 ] ) : @xmath311 this gives rise to the following practical implementation by a series of time - reversal experiments : @xmath312 @xmath313 will then be the signal to be applied by the base station during the communication process with the user @xmath116 .
in many practical applications , only partial measurements of the whole wave - field are available .
for example , in ocean acoustics often only pressure is measured , whereas the velocity field is not part of the measurement process .
similarly , in wireless communication only one or two components of the electric field might be measured simultaneously , but the remaining electric components and all magnetic components are missing .
we want to demonstrate in this section that all results presented above are valid also in this situation of partial measurements , with the suitable adaptations .
mathematically , the measurement operator needs to be adapted for the situation of partial measurements .
let us concentrate here on the special situation that only one component @xmath314 @xmath315 of the incoming wave field @xmath106 is measured by the users and the base station .
all other possible situations will then just be combinations of this particular case .
it might also occur the situation that users can measure a different partial set of components than the base station .
that case also follows directly from this canonical situation .
we introduce the new signal space at the base station @xmath316)^{{k}}$ ] and the corresponding signal projection operator @xmath317 by putting @xmath318 we see immediately that its adjoint @xmath319 is given by @xmath320 where @xmath321 appears on the right hand side at the @xmath322-th position . our new measurement operator @xmath323 , and the new source operator @xmath324 ,
are then defined by @xmath325 analogous definitions are done for @xmath326 , @xmath327 , @xmath328 and @xmath329 at the users .
obviously , we will have to replace now in the above derivation of the iterative time - reversal procedure all measurement operators @xmath330 by @xmath323 ( and @xmath331 by @xmath328 ) and all source operators @xmath88 by @xmath324 ( and @xmath87 by @xmath329 ) .
in particular , the new communication operators are now given by @xmath332 in the following two theorems we show that the main results of this paper carry over to these newly defined operators .
[ theorem.pm.1 ] we have @xmath333 _ proof : _ the proof is an easy exercise using ( [ pm.3 ] ) and theorem [ theorem.adj.1 ] .
@xmath145 [ theorem.pm.2 ] it is @xmath334 _ proof : _ the proof is now identical to the proof of theorem [ theorem.tra.2 ] , using theorem [ theorem.pm.1 ] instead of theorem [ theorem.adj.1 ] .
@xmath145 in fact , it is easy to verify that all results of this paper remain valid for _ arbitrarily defined linear measurement operators _ @xmath335 where @xmath336 and @xmath337 are any meaningful signal spaces at the users and the base antennas , respectively .
the only requirement is that it is experimentally possible to apply signals according to the source operators defined by @xmath338 where @xmath339 and @xmath340 are the formal adjoint operators to @xmath341 and @xmath342 with respect to the chosen signal spaces @xmath336 and @xmath337 .
in addition , the measurement and source operators as defined above are required to satisfy the commutation relations as stated in lemma [ lemma.tra.1 ] . under these assumptions ,
we define the _ generalized communication operators _ @xmath343 and @xmath344 by @xmath345 now the proof to theorem [ theorem.tra.2 ] directly carries over to this generalized situation , such that we have also here @xmath346 this yields iterative time - reversal schemes completely analogous to those presented above .
we have derived in this paper a direct link between the time - reversal technique and the adjoint method for imaging . using this relationship ,
we have constructed several iterative time - reversal schemes for solving an inverse problem which arises in ocean acoustic and wireless communication .
each of these schemes can be realized physically as a series of time - reversal experiments , without the use of heavy signal processing or computations .
one of the schemes which we have derived ( and which we call the basic scheme ) , is practically equivalent to a technique introduced earlier in @xcite using different tools .
therefore , we have given an alternative theoretical derivation of that technique , with a different mathematical interpretation .
the other schemes which we have introduced are new in this application .
they represent either generalizations of the basic scheme , or alternatives which follow different objectives .
many questions related to these and similar iterative time - reversal approaches for telecommunication are still open .
the experimental implementation has been investigated so far only for one of these techniques in @xcite , for the situation of underwater sound propagation .
a thorough experimental ( or , alternatively , numerical ) verification of the other schemes is necessary for their practical evaluation .
an interesting and practically important problem is the derivation of quantitative estimates for the expected focusing quality of each of these schemes , for example following the ideas of the work performed for a single step in @xcite .
certainly , it is expected that these estimates will again strongly depend on the randomness of the medium , on the geometry and distribution of users and base antennas , and on technical constraints as for example partial measurements .
also , different types of noise in the communication system need to be taken into account .
the performance in a time - varying environment is another interesting issue of practical importance .
all schemes presented here can be adapted in principle to a dynamic environment by re - adjusting the constructed optimal signals periodically .
practical ways of doing so need to be explored .
finally , we want to mention that speed is an important factor in the application of communication . in the time - reversal experiment , in a certain sense
a physical process is used for applying a mathematical operator ( namely @xmath142 ) in order to solve a given inverse problem . in a complex environment
, this can be much faster than using a powerful computer , even if the complex environment would be known to the users and the base station .
the author thanks george papanicolaou , hongkai zhao and knut solna for many exciting and useful discussions on time - reversal .
he thanks dominique lesselier for many fruitful discussions , and for making the research stay at suplec possible .
he thanks mathias fink , claire prada , gabriel montaldo , francois vignon , and the group at laboratoire ondes et acoustique in paris for pointing out to him their recent work on iterative communication schemes , which stimulated part of this paper . the author also thanks bill kuperman and the group at scripps institution of oceanography in san diego for useful discussions . financial support from the office of naval research ( under grant n00014 - 02 - 1 - 0090 ) and the cnrs ( under grant no .
8011618 ) is gratefully acknowledged .
for arbitrary @xmath347 and @xmath100 we have @xmath348}\left\langle \int_{\omega}\gamma_k({{\bf x}}){{\bf u}}({{\bf x}},t)d{{\bf x}}\,,\ , { { \bf r}}(t)\right\rangle_{n}dt \\ & = & \int_{[0,t]}\int_{\omega}\left\langle\gamma({{\bf x}}){{\bf u}}({{\bf x}},t ) \,,\,\gamma^{-1}({{\bf x}})\sum_{k=1}^{{k}}\gamma_k({{\bf x } } ) { { \bf r}}(t)\right\rangle_{n}d{{\bf x}}dt.\end{aligned}\ ] ] therefore , @xmath349 with @xmath350 doing the analogous calculation for @xmath351 we get @xmath352 taking the adjoint of ( [ pt1.1 ] ) we see that @xmath353 holds as well .
this proves the theorem .
we have the following version of _ green s formula _ @xmath354}\int_{\omega } \left\langle \gamma({{\bf x}})\frac{\partial { { \bf u}}}{\partial t}\,+\,\sum_{i=1}^{3 } d^i\frac{\partial { { \bf u}}}{\partial x_i}\,+\,\phi({{\bf x}}){{\bf u}}\,,\,{{\bf z}}\right\rangle_n\,d{{\bf x}}dt\ ] ] @xmath355}\int_{\omega } \left\langle \gamma({{\bf x } } ) { { \bf u}}({{\bf x}},t ) , { { \bf v}}({{\bf x}},t ) \right\rangle_n \,d{{\bf x}}dt \ , + \ , \int_{[0,t]}\int_{\omega } \left\langle { { \bf q}}({{\bf x}},t)\,,\,{{\bf z}}({{\bf x}},t ) \right\rangle_n \,d{{\bf x}}dt\ ] ] @xmath356}\int_{\omega } \left\langle { { \bf u}}\,,\ , - \gamma({{\bf x}})\frac{\partial { { \bf z}}}{\partial t } \,-\,\sum_{i=1}^{3 } d^i\frac{\partial { { \bf z}}}{\partial x_i}\,+\,\phi({{\bf x}}){{\bf z}}\right\rangle_n \,d{{\bf x}}dt\ ] ] @xmath357}\int_{\omega}\left\langle \gamma({{\bf x } } ) { { \bf u}}({{\bf x}},t),{{\bf v}}({{\bf x}},t)\right\rangle_n\,d{{\bf x}}dt \ , + \ , \int_{[0,t]}\int_{\omega } \left\langle { { \bf q}}({{\bf x}},t)\,,\,{{\bf z}}({{\bf x}},t ) \right\rangle_n \,d{{\bf x}}dt\ ] ] @xmath358 @xmath357}\int_{\partial\omega}\sum_{i=1}^{3 } \left\langle d^i{{\bf u}},{{\bf z}}\right\rangle_n\nu_i({{\bf x } } ) \,d\sigma dt,\ ] ] where @xmath359 is the outward normal at @xmath108 in the point @xmath360 . notice that we have augmented green s formula in ( [ pt2.1 ] ) by some terms which appear in identical form on the left hand side and on the right hand side .
we will assume here that the boundary is far away from the sources and receivers and that no energy enters @xmath25 from the outside , such that during the time interval of interest @xmath20 $ ] all fields along this boundary are identically zero .
this is expressed by the boundary conditions given in ( [ sym.4 ] ) and ( [ adj.5 ] ) .
let @xmath5 be a solution of ( [ sym.1]),([sym.2]),([sym.4 ] ) , and @xmath361 a solution of ( [ adj.3])([adj.5 ] ) .
then the first term on the left hand side of ( [ pt2.1 ] ) and the third term on the right hand side cancel each other because of ( [ sym.1 ] ) .
the second term on the left hand side and the first term on the right hand side cancel each other because of ( [ adj.3 ] ) .
the @xmath362-term and the @xmath363 term vanish due to ( [ adj.4 ] ) and ( [ sym.2 ] ) , respectively , and the boundary integral vanishes because of ( [ sym.4 ] ) and ( [ adj.5 ] ) . the remaining terms ( i.e. the third term on the left hand side and the second term on the right hand side ) can be written as @xmath364 with @xmath365 as defined in ( [ adj.6 ] ) . @xmath145
we prove the lemma by using _
greens formula _ : @xmath366 \,d{{\bf x}}dt\ ] ] @xmath367\,d{{\bf x}}dt + \,\int_{0}^{t}\int_{\omega}\left [ { { \bf q}}_{{{\bf v}}}{{\bf v}}_a \,+\ , { { \bf q}}_p
p_a \right]\,d{{\bf x}}dt\ ] ] @xmath368\,d{{\bf x}}dt\ ] ] @xmath367\,d{{\bf x}}dt + \,\int_{0}^{t}\int_{\omega}\left [ { { \bf q}}_{{{\bf v}}}{{\bf v}}_a \,+\ , { { \bf q}}_p
p_a \right]\,d{{\bf x}}dt\ ] ] @xmath369 @xmath370\,d{{\bf x}}\,+\,\int_{\omega } \kappa \big [ ( p_{f}p_{a})({{\bf x}},t )
-(p_{f}p_{a})({{\bf x}},0)\big]\,d{{\bf x}}.\ ] ] this equation has the form ( [ pt2.1 ] ) .
notice that we have augmented green s formula in ( [ pt3.1 ] ) , as already shown in ( [ pt2.1 ] ) , by some terms which appear in identical form on the left hand side and on the right hand side . the first term on the left hand side of equation ( [ pt3.1 ] ) and the third term on the right hand side
cancel each other due to ( [ atrm.1]),([atrm.2 ] ) . the second term on the left hand side and the first term on the right hand side
cancel each other because of ( [ atrm.4]),([atrm.5 ] ) .
the ( @xmath155)-terms and the ( @xmath156)-terms vanish due to ( [ atrm.6 ] ) , ( [ atrm.3 ] ) , respectively , and the boundary terms vanish because of zero boundary conditions .
we are left over with the equation
@xmath371\,d{{\bf x}}dt \,=\,\int_{0}^{t}\int_{\omega}\left [ { { \bf q}}_{{{\bf v}}}{{\bf v}}_a \,+\ , { { \bf q}}_p p_a \right]\,d{{\bf x}}dt\ ] ] defining @xmath149 by ( [ atrm.7 ] ) , this can be written as @xmath372 therefore , @xmath149 is in fact the adjoint of @xmath104 , and the lemma is proven .
we prove the lemma by using _ green s formula _ : @xmath373 \,d{{\bf x}}dt\ ] ] @xmath374\,d{{\bf x}}dt \,+\,\int_{0}^{t}\int_{\omega}\left[{{\bf q}}_e { { \bf e}}_a\,+\,{{\bf q}}_h{{\bf h}}_a \right]\,d{{\bf x}}dt\ ] ] @xmath375\,d{{\bf x}}dt\ ] ] @xmath376\,d{{\bf x}}dt \,+\,\int_{0}^{t}\int_{\omega}\left[{{\bf q}}_e { { \bf e}}_a\,+\,{{\bf q}}_h{{\bf h}}_a \right]\,d{{\bf x}}dt\ ] ] @xmath377 @xmath378\,d{{\bf x}}\,+\,\int_{\omega}\mu\big [ ( { { \bf h}}_{f}{{\bf h}}_{a})(t)- ( { { \bf h}}_{f}{{\bf h}}_{a})(0)\big]\,d{{\bf x}}\ ] ] this equation has the form ( [ pt2.1 ] ) .
notice that we have augmented green s formula in ( [ pt4.1 ] ) , as already shown in ( [ pt2.1 ] ) , by some terms which appear in identical form on the left hand side and on the right hand side . the first term on the left hand side of equation ( [ pt4.1 ] ) and the third term on the right hand side
cancel each other because of ( [ etrm.1 ] ) and ( [ etrm.2 ] ) . the second term on the left hand side and the first term on the right hand side
cancel each other because of ( [ etrm.4 ] ) , ( [ etrm.5 ] ) .
the ( @xmath156)-terms and the ( @xmath155)-terms vanish due to ( [ etrm.3 ] ) and ( [ etrm.6 ] ) .
the boundary terms vanish because of zero boundary conditions .
we are left over with the equation @xmath379\,d{{\bf
x}}dt \,=\ , \int_{0}^{t}\int_{\omega}\left[{{\bf q}}_e { { \bf e}}_a\,+\,{{\bf q}}_h{{\bf h}}_a \right]\,d{{\bf x}}dt .\ ] ] defining @xmath149 by ( [ etrm.7 ] ) , this can be written as @xmath380 therefore , @xmath149 is in fact the adjoint of @xmath104 , and the lemma is proven .
99 kuperman w a , hodgkiss w s , song h c , akal t , ferla c , and jackson d r 1998 phase conjugation in the ocean : experimental demonstration of an acoustic time reversal mirror _ j. acoust .
* 103 * 25 - 40 | we establish a direct link between the time - reversal technique and the so - called adjoint method for imaging . using this relationship
, we derive new solution strategies for an inverse problem which arises in telecommunication .
these strategies are all based on iterative time - reversal experiments , which try to solve the inverse problem _ experimentally _ instead of computationally .
we will focus in particular on examples from underwater acoustic communication and wireless communication in a multiple - input multiple - output ( mimo ) setup . |
the goal of this paper is to provide a numerical investigation of the so - called _ soliton dynamics _
behaviour for the nonlinear schrdinger equation with an external time independent smooth potential @xmath2 @xmath3 that is the qualitative behaviour of the solution @xmath4 in the semi - classical regime , namely for @xmath5 ( which of course plays the rle of planck s constant ) going to zero , by taking as initial datum a ( bump like ) function of the form @xmath6 we shall assume that @xmath7 , @xmath8 , @xmath9 is the imaginary unit and @xmath10 is the unique @xcite ( up to translations ) positive and radially symmetric solution of the elliptic problem @xmath11 for some value @xmath12.finally , @xmath13 and @xmath14 are given vectors in @xmath15 that should be conveniently thought ( in the transition from quantum to classical mechanics ) as corresponding to the _ initial position _ and _ initial velocity _ respectively of a point particle . in this framework , since has a conservative nature , the typical expected behaviour is that the solution travels with the shape of @xmath16 ( hence its support shrinks , as @xmath5 gets small ) along a suitable concentration line @xmath17 merely depending on the potential @xmath2 and starting at @xmath13 with initial slope @xmath14 .
on the basis of the analytical results currently available in literature ( see the discussion in section [ theoryfacts ] ) , we believe that providing some numerical study is useful to complete the overall picture of this phenomenon and furnish some practical machinery for the computation of the solutions of and , in turn , of - .
the authors are not aware of any other contribution in the literature on this issue . for the linear schrdinger
, some results can be found in @xcite .
it is well - known that , given a positive real number @xmath18 , the afore mentioned ( ground state ) solution @xmath0 of ( where the value of @xmath19 depends on @xmath18 ) can obtained through the following variational characterization on the sphere of @xmath20 @xmath21 where @xmath22 is the @xmath23 energy functional @xmath24 furthermore , there exists a suitable choice of @xmath18 yielding @xmath25 as eigenvalue in equation .
the restriction to the values of @xmath26 below @xmath27 is strictly related to the global well - posedness of for any choice of initial data @xmath28 in @xmath29 .
if @xmath26 is larger than or equal to @xmath27 , then the solution can blow - up in finite time ( see e.g. the monograph by t. cazenave @xcite ) . in particular , in the two dimensional case , @xmath26 will be picked in @xmath30 . from the analytical side ,
it has been rigorously known since 2000 that the solution @xmath4 of remains close to the ground state @xmath0 , in the sense stated here below , locally uniformly in time , as @xmath5 is pushed to zero .
as we said , this dynamical behaviour is typically known as soliton dynamics ( for a recent general survey on solitons and their stability , see the work of t. tao @xcite ) .
for the nonlinear equation , rigorous results about the soliton dynamics were obtained in various papers by j.c .
bronski , r.l .
jerrard @xcite and s. keraani @xcite .
we also refer to @xcite for a complete study of the problem with the additional presence of an external time independent magnetic vector potential @xmath31 , and to @xcite for a study of a system of two coupled nonlinear schrdinger equations , a topic which is rapidly spreading in the last few years . the arguments are mainly based upon the following ingredients : the energy convexity estimates proved by m. weinstein @xcite to get the so called modulational stability , the use of conservation laws ( mass and energy ) satisfied by the equation and by the associated hamiltonian system in @xmath15 built upon the guiding external potential @xmath2 , that is the classical _ newton law _ @xmath32 under reasonable assumptions on @xmath2 ( e.g. uniform boundedness of the second order partial derivatives ) , equation admits a unique global solution @xmath33 which satisfies the following conservation law @xmath34 let us now define a suitable scaling of the standard norm of @xmath35 @xmath36 the precise statement of the soliton dynamics reads as follows [ anal11 ] let @xmath4 be the solution to problem corresponding to the initial datum .
then there exists a family of shifts @xmath37 such that , as @xmath5 goes to zero , @xmath38 is equal to the function @xmath39 } , \qquad x\in{{\mathbb r}}^n,\,\,t>0,\ ] ] up to an error function @xmath40 such that @xmath41 , locally uniformly in time .
it is important to stress that , in the particular case of _ standing wave solutions _ of , namely special solutions of of the form @xmath42 where @xmath43 is a real - valued function , there is an enormous literature regarding the semi - classical limit for the corresponding elliptic equation @xmath44 see the recent book @xcite by a. ambrosetti and a. malchiodi and the references therein . to this regard notice
that , if @xmath45 ( null initial velocity ) and @xmath13 is a critical point of the potential @xmath2 , as equation admits the trivial solution @xmath46 and @xmath47 for all @xmath48 , formula reduces to @xmath49 so that the concentration of @xmath4 is _ static _ and takes place at @xmath13 , instead occurring along a smooth concentration curve in @xmath15 .
this is consistent with the literature for the standing wave solutions mentioned above .
for other achievements about the full dynamics of , see also @xcite ( in the framework of orbital stability of standing waves ) as well as @xcite ( in the framework of non - integrable perturbation of integrable systems ) .
similar results were investigated in geometric optics by a different technique ( wkb method ) , namely writing formally the solution as @xmath50 , with @xmath51 where @xmath52 and @xmath53 are solutions , respectively , of a hamilton - jacobi type equation ( the eikonal equation ) and of a system of transport equations . in presence of a constant external potential , the orbital stability issue for problem
was investigated by t. cazenave and p.l .
lions @xcite , and by m. weinstein in @xcite .
then , a. soffer and m. weinstein proved in @xcite the asymptotic stability of nonlinear ground states of .
see also the following important contributions : v. buslaev and g. perelman @xcite , v. buslaev and c. sulem @xcite , j. frhlich , s. gustafson , l. jonsson , i.m .
sigal , t .-
tsai and h .-
yau @xcite , j. holmer and zworski @xcite , a. soffer and m. weinstein @xcite , t .- p .
tsai and h .-
yau @xcite .
another interesting problem concerns the case where the initial datum is multibump ( for simplicity two bumps ) , say , @xmath54 where @xmath55 are solutions to the problem @xmath56 for some fixed @xmath57 , @xmath58 and @xmath59 are taken as initial data for @xmath60 then we state the following let @xmath4 be the solution to corresponding to the initial datum .
then there exist two families of shifts @xmath61 such that , as @xmath5 goes to zero , @xmath38 is equal to the function @xmath62 } + r_2\big(\frac{x - y(t)}{{\varepsilon}}\big)e^{\frac{{\rm i}}{{\varepsilon}}\left[x\cdot \dot y(t)+\teta_{{\varepsilon}}^2(t)\right]},\ ] ] up to an error function @xmath40 depending both on @xmath5 and on the initial relative velocity @xmath63 ( the larger is @xmath64 the smaller is the error ) , locally uniformly in time .
see figure [ fig:2 ] in the final section for a movie showing this behaviour .
in the numerical simulations included in the last section of the paper , we shall consider the two dimensional case . on the other hand , here we consider the general case .
our purpose is to solve the schrdinger equation @xmath65 where @xmath66 , so that @xmath67 is the solution of @xmath68 being @xmath69 real , positive and minimizing the energy under the constraint @xmath70 . instead of a direct minimization of the energy ( see , e.g. , @xcite ) ,
here we consider the following parabolic differential equation @xmath71 with vanishing boundary conditions , where the map @xmath72 is defined by @xmath73 this approach is similar to the imaginary time method ( see , e.g. , @xcite ) , based on the propagation of the schrdinger equation along imaginary time @xmath74 and projection to the @xmath75 sphere of radius @xmath76 . in equation
, projection is not necessary and the energy decreases : in fact , if we multiply equation by @xmath77 and integrate over @xmath78 , we easily get @xmath79 and if we multiply equation by @xmath80 and integrate over @xmath78 , we get @xmath81 hence , the steady - state solution @xmath82 of satisfies @xmath83 and has a minimal energy .
in fact , notice that by the results of @xcite , for any @xmath12 there exists a unique ( up to translations ) positive and radially symmetric solution @xmath84 of . in turn , given @xmath85 , if @xmath86 denote , respectively , the positive radial solutions of the equations @xmath87 then it is readily verified that @xmath88 which tells us that that , up to a scaling , the solution corresponding to different values of @xmath19 is unique .
notice now that , due to the choice of the bump like initial datum ( gaussian like , see ) in the iterations to compute @xmath89 ( see the discussion below ) , it turns out that @xmath90 , defined as @xmath91 , is negative and @xmath89 is positive , radially symmetric ( see figure [ fig:0 ] ) and solves @xmath92 where @xmath93 . if @xmath94 denotes the ground state solution ( with the corresponding positive eigenvalue denoted by @xmath95 ) , then we have @xmath96 of with @xmath97 and @xmath98 . of course , in the computation of @xmath89 , there is a spurious imaginary part of maximum value around @xmath99 , since the complex fft algorithm is involved .
the corresponding value of @xmath90 is @xmath100 . ] on the other hand , by construction , we have @xmath101 namely @xmath102 . finally , by the definition of @xmath103 and @xmath104 in , we get @xmath105 and @xmath106 , yielding from the desired conclusion , that is @xmath107 moreover , @xmath108 is a solution of
. we will take @xmath109 as our candidate initial condition for the time - dependent nonlinear schrdinger equation . from a numerical point of view , it is convenient to compute directly @xmath110 instead of @xmath111 and to apply the change of variable @xmath112{\varepsilon^n}\phi_\varepsilon(x , t)$ ] , @xmath113 , to the nonlinear schrdinger equation , and hence to equation .
altogether , we need to solve @xmath114 with @xmath115 where @xmath116{\varepsilon^n}$ ] . since it is not possible to numerically integrate the equation up to an infinite time , we will consider @xmath117 the steady - state as soon as @xmath118 is stabilized within a prescribed tolerance .
the initial condition @xmath119 can be arbitrarily chosen ( in the class of bump like functions ) , but an initial solution with small energy will shorten the `` steady - state '' time @xmath120 . among the family of the gaussian functions parameterized by @xmath121 @xmath122{\left(\frac{\varepsilon}{\pi}\right)^n}\ ] ] with @xmath123 it is possible to choose the one with minimal energy .
in fact @xmath124 if we define @xmath125 the minimum for @xmath126 is attained for @xmath127 the quantities @xmath128 and @xmath129 can be analytically computed and give @xmath130 with the normalization introduced above , the nonlinear schrdinger equation to solve is @xmath131 a well - established numerical method for the cubic schrdinger equation ( focusing or defocusing case ) is the strang splitting @xcite .
it is based on a split of the full equation into two parts , in which the first is spectrally discretized in space and then exactly solved in time and the second has an analytical solution .
we used the strang splitting method as well .
the first part is @xmath132 thus , the fourier coefficients of @xmath133 restricted to a sufficiently large space domain satisfy a linear and diagonal system of odes , which can be exactly solved .
the second part is @xmath134 it is easy to show that the quantity @xmath135 is constant in time for this equation .
then it has an analytical solution .
given the approximated solution @xmath136 of equation , a single time step of the strang splitting fourier spectral method can be summarized in 1 .
take @xmath137 as initial solution at time @xmath138 for and solve for a time step @xmath139 , obtaining @xmath140 ; 2 .
take @xmath140 as initial solution at time @xmath138 for and solve for a time step @xmath141 , obtaining @xmath142;[realspace ] 3 .
take @xmath142 as initial solution for and solve for a time step @xmath139 , obtaining @xmath143 .
the result @xmath143 is an approximation of @xmath144 .
since the solutions of the first part and the second part are trivial to compute in the spectral space and in the real space , respectively , it is necessary to transform the solution from spectral space to real and from real space to spectral before and after step above , respectively .
all the transformations can be carried out by the fft algorithm .
the method turns out to be spectrally accurate in space and of the second order in time .
therefore , we used the fourier spectral decomposition for the solution of equation , too .
together with the galerkin method , it yields a nonlinear system of odes @xmath145 where @xmath146 is the vector of fourier coefficients , @xmath147 the diagonal matrix of the eigenvalues of the laplace operator and @xmath148 the truncated fourier expansion of the whole nonlinear part of equation .
for the solution of equation we used an exponential runge
kutta method of order two ( see , e.g. ,
@xcite ) , with the embedded exponential euler method .
given the approximation @xmath149 , a single time step of the method is 1 .
set @xmath150 and @xmath151 ; 2 .
compute @xmath152 ( exponential euler method ) ; 3 . compute @xmath153[error ] where @xmath154 and @xmath155 are the analytic functions @xmath156 the result is an approximation of @xmath157 .
exponential integrators are explicit and do not suffer of time step restrictions .
however , they require the computation of matrix functions . in our case ,
the matrices involved @xmath158 are diagonal and the computation of the matrix functions @xmath159 , @xmath160 and @xmath161 is trivial . in order to compute the terms @xmath162 and @xmath163 , it is necessary to recover the functions in the real space corresponding to the fourier spectral coefficients @xmath164 and @xmath165 , respectively , then compute the nonlinear part of equation and finally to compute its fourier transform .
all the transformations can be carried out by the fft algorithm .
the term @xmath166 in step ( [ error ] ) above can be used as an error estimate for @xmath167 and then it is possible to derive a variable time step integrator .
this is particularly useful for our aim of computing the steady - state of the equation : in fact , we expect that the as soon as the solution approaches the steady - state it is possible to enlarge the time step , thus reducing the computational cost .
the method turns out to be spectrally accurate in space and of the second order in time .
in this section , in order to provide some examples , we reduce to the two dimensional setting and focus on the physically relevant case of harmonic potential @xmath168 well - established in the theory of bose - einstein condensates . in the two movies starting from figure [ fig:1 ]
we show the dynamics of the solitary wave along two lissajous curves , periodic in the left side and ergodic for the right side . in the movie starting from figure [ fig:2 ]
we report the soliton dynamics in the case of an initial datum exhibiting a double bump behaviour ( with a sufficiently large distance between the centers ) up to the collision time .
it is important to stress that in these figures the paths have an analytic expression and are plotted before the dynamics starts .
the movies will then show that the centers of mass of the solitons follow adherently these curves up to the final computation time .
an analysis of the error ( in the single bump case ) arising when the modulus of the solution @xmath169 is replaced by the modulus of the expression in the representation formula , namely @xmath16 , is indicated in figure [ fig:3 ] . as predicted by the analytical property [ anal11 ]
, the error in the @xmath170 is below the order @xmath171 .
run : movie1bh.avi [ , @xmath172 , @xmath173 , @xmath174 , @xmath175 . in the left movie
, we chose @xmath176 and @xmath177 ( rational ratio ) . in the right movie , we chose @xmath178 and @xmath177 ( irrational ratio ) . notice that , although the ratios @xmath179 are very close in the two examples , the soliton dynamics is ergodic in the right movie .
of course the figures refer to the ( squared modulus of the ) solution at the time @xmath180 and contain the concentration paths ( admitting an analytic expression ) that the soliton is going to travel on.,title="fig : " ] ] run : movie1bhirr.avi [ , @xmath172 , @xmath173 , @xmath174 , @xmath175 . in the left movie
, we chose @xmath176 and @xmath177 ( rational ratio ) . in the right movie , we chose @xmath178 and @xmath177 ( irrational ratio ) . notice that , although the ratios @xmath179 are very close in the two examples , the soliton dynamics is ergodic in the right movie .
of course the figures refer to the ( squared modulus of the ) solution at the time @xmath180 and contain the concentration paths ( admitting an analytic expression ) that the soliton is going to travel on.,title="fig : " ] ] , @xmath186 , @xmath97 , @xmath187 and a final time @xmath188 . with the change of variable we used , @xmath113 and @xmath189 , we have @xmath190 and @xmath191 .
hence , the numerical error is computed through formula ( written for the 2d case ) @xmath192 . as predicted by the analytical property [ anal11 ]
, the error in the @xmath170 is below the order @xmath171 . ] | we provide some numerical computations for the soliton dynamics of the nonlinear schrdinger equation with an external potential . after computing the ground state solution @xmath0 of a related elliptic equation
we show that , in the semi - classical regime , the center of mass of the solution with initial datum modelled on @xmath0 is driven by the solution of @xmath1 .
finally , we provide some examples and analyze the numerical errors in the two dimensional case when @xmath2 is an harmonic potential . |
it is widely believed that spacetime must be both orientable and time - orientable @xcite .
arguments are that there is no evidence of a lack of orientability and that a non - orientable spacetime would be incompatible with the observed violations of p ( parity ) and t ( time reversal ) invariance .
these arguments are shown to be false . this paper aims to answer the questions : _ how can time - orientability be tested ? _ and _ is a non - time orientable spacetime consistent with observations?_. the focus of this paper is on experiments a physicist would have to do to test the orientability of a spacetime region .
the work is motivated by results suggesting that non - time orientable spacetime regions might be used to explain quantum theory @xcite , the existence of electric charge @xcite and spin half @xcite .
topological models of elementary particles within the equations and framework of general relativity require a breakdown of the causal structure to allow topology changing interactions .
a non - timeorientable spacetime structure can also exhibit net electric charge from source free maxwell equations and would naturally have the transformation properties of a spinor under rotations .
mathematically , the orientability of a manifold is a simple and unambiguous property .
a few mathematical definitions and theorems are given in the next section .
this paper looks at how the mathematical definitions can be used in practice to answer questions about spacetime .
some simple experiments are postulated ; they follow the mathematical definitions very closely , but the interpretation remains ambiguous .
this paper is exclusively concerned with connected spacetimes . locally spacetime is always orientable .
spacetime is modelled by a 3 + 1 dimensional semi - riemannian manifold .
therefore mappings to flat minkowski spacetime exist in a region of every point .
minkowski space - time is orientable hence spacetime is locally orientable .
it follows that orientability is a global rather than a local property . on a non - orientable spacetime
there exists a closed path along which a consistent orientation can not be defined .
[ rem : path ] this follows from the definition of orientability and topological arguments .
if spacetime is not orientable then through every point there exists a path on which an orientation can not be defined .
this follows simply from remark [ rem : path ] above and assumed connectedness of spacetime .
if a has a non - orientable path , c , through it and x is any other point .
then by connectedness there exists a path , p , from x to a and a path , p , from a to x. a non - timeorientable closed path through x is constructed by joining the paths p - c - p. putting the three remarks together , spacetime could be orientable for all paths except those going through a small remote region . for all normal purposes such a spacetime would appear orientable . only an experiment that probed the remote region
could detect the non - orientability .
however , non - orientability in a connected spacetime can not be strictly localised as there would be a curve through every spacetime point that went through the non - timeorientable region .
mathematically , orientability is a global property of spacetime . by analogy with a mbius strip it is possible that paths comparable in size to the universe could be non - orientable .
however a non - orientable region could be microscopic in size : for example a non - orientable wormhole with dimensions comparable to the plank length .
this work is motivated by the possibility that elementary particles may be regions of non - trivial topology , but the arguments are generally applicable .
it could of course be possible for all _ timelike _ paths in a non - orientable spacetime to be orientable .
a non - orientable region behind an event horizon would be such an example .
the observed violation of t - invariance implies that an observer can unambiguously define the direction of time .
this is compatible with a localised time - reversing region , because the boundary conditions ( the expanding universe ) define a time direction and because a consistent time direction could still be defined if the experimenter never probed the time - reversing region .
microscopic localised nonorientable regions as described in @xcite are therefore consistent with observations .
this section examines classical tests of the orientability of spacetime - they are based upon classical general relativity and classical physics .
it can not be overemphasised that this argument is not based on quantum field theory or even quantum mechanics , it is classical - classical mechanics and classical general relativity .
consider a space that is flat and space - orientable everywhere except for a region .
in spacetime would be a tube .
the construction of spacetimes such as this , with all permutations of orientability , are described in @xcite : sorkin @xcite constructs a non - orientable wormhole , which is also asymptotically flat .
gibbons and herdeiro @xcite give an asymptotically flat example for a supersymmetric rotating black hole .
the arguments in this paper are independent of the exact form of the metric in .
the orientation of is tested by sending a probe into .
the probe contains a triad of unit vectors , and .
the triad defines a space - orientation continuously along the path taken by the probe .
the path ( timeline ) of the probe and the path ( timeline ) of the observer together define a closed loop .
an orientation can be defined in the laboratory , excluding region .
if the probe that emerges from has a different orientation from that defined in the rest of the laboratory then space is not orientable ( see figure [ fig : space ] ) .
an example of a region that is not time - orientable is easily constructed ( see for example ref .
@xcite ) : remove a ball from space - that is a world - tube from spacetime .
then identify opposite points of the sphere using the antipodal map . at the same time identify @xmath0 with @xmath1 in a similar manor to the creation of a mbius strip . like a mbius strip this example is continuous and flat .
a similar test of the time orientability of would need a probe that defined a time direction continuously along a trajectory through and back to the observer .
a clock would be a suitable probe , with the positive time direction defined by increasing times displayed on the clock .
however the experiment is not so simple .
consider a clock that enters at an observer time @xmath2 and emerges at a later observer time but counting backwards ( see figure [ fig : time1 ] ) .
this is not a demonstration of non - time orientability , because in this experiment , the clock increases in value and then decreases . at some point in the path
it attains a maximum reading and at that point it does not define a time direction . in a true test of time orientability on a region through
which time can not be oriented , the clock readings would increase steadily along the path taken by the clock . before entering the observer sees the time values increasing on the clock . when the clock exits at a time @xmath3 the increasing clock times would be at ever decreasing values for the observer time .
the observer would still see a backwards counting clock , but only at times before @xmath3 ( see figure [ fig : time2 ] ) .
this might seem like a clear example of a successful demonstration of a failure of time orientability .
the author would be sympathetic to such a view ! however the observer sees the backwards ( clock - time counting backwards ) moving clock entering as well as the forward counting clock entering . at observer times greater than @xmath3
the observer sees no clock at all - neither inward nor outward moving .
so the observer can interpret the experiment as a clock and an anti - clock entering and annihilating each other .
it must be emphasised that this is still a description in terms of classical physics .
the expression _ anti - clock _ is a logical one - it does not refer to anti - particles in the normal sense - indeed it is totally independent of the construction of the clock .
if you expect the number of clocks to be preserved , then describing the backward counting clock as an anti - clock achieves the objective . using the concept of an anti - clock , the experiment has zero clocks most of the time . before @xmath2
there is a clock and anti - clock after @xmath3 there are no clocks .
the alternative is to postulate two rather different real objects ( the clock and the backward counting clock ) that both enter and vanish .
the analogy shares with anti - particles the fact that both the clock and anti - clock must exist for this to happen .
unlike the conventional particle -antiparticle annihilation there does not appear to be any conservation of energy .
gibbons and herdeiro make similar observations for a particle traversing their supersymmetric rotating black hole
. in general @xmath2 and @xmath3 would be different . in a normal particle - antiparticle annihilation event
they would be equal .
the difference between @xmath2 and @xmath3 depends upon the geometry . for a similar construction to that described above the identification of @xmath4 with @xmath5 ( rather than @xmath0 and @xmath1 ) would result in @xmath2 and @xmath3 being equal
. a simple would be just a special case of this to make matters even worse , the experiment is dependent upon the observer . since any attempt to stop the backward counting clock from entering would be inconsistent with the clock entering and leaving a time reversing region .
so not only is the anti - clock description a possible alternative description , the time reversing event _ requires _ the existence of anti - clocks ( time reversed clocks ) to exist before the experiment takes place .
since a clock is normally a macroscopic object composed of atoms , anticlocks do not exist in the sense of clocks composed of antiparticles .
therefore , an apparently simple test may be physically impossible to carry out .
this highlights a weakness in the way the problem was posed .
space orientability is a property of space .
however timeorientability can only be a property of spacetime .
furthermore it is a global property of spacetime .
it would appear that the existence or otherwise of a time reversing region is dependent upon the observer . a probe and an anti - probe
must both be prepared for a consistent result , in which case it could be argued that the existence of the time reversing region depends upon the experiment being performed to test it .
although the preceding argument referred to anti - clocks , this was a purely logical classical description .
quantum field theory , which correctly describes both particles and antiparticles - and indeed requires anti - particles for completeness , was not being used .
the similarities of these classical arguments to some quantum phenomena is not coincidental .
it has already been shown @xcite that quantum logic can be derived from classical general relativity with acausal spacetimes .
it is not possible to define a spinor field on a non - time orientable spacetime @xcite .
since fermions exist , this well - known result has been cited as evidence that time must be orientable .
however the argument relies on a realist interpretation of the wavefunction and the false assumption that a wavefunction is defined at each spacetime point .
in fact a wavefunction is a function defined on a 3n - dimensional configuration space where , n , is the number of particles .
see ballentine @xcite chapter 4 for the clearest discussion of arguments against a realist interpretation of a wavefunction ( also @xcite ) .
fermion fields are used to calculate the probabilities of results of experiments .
the extension of a fermion field to inaccessible microscopic non - orientable regions is neither relevant nor particularly meaningful , since measurements can not take place within these regions .
while on a spacetime with accessible time - reversing regions , the inability to define a spinor field , simply means that qft can not be used to calculate probabilities of the results of experiments on a space time that is not timeorientable a conclusion that is hardly surprising !
although the analysis is clear , the interpretation is by no means obvious .
the following statements can all be supported by the arguments above : 1 .
spacetime is not time orientable .
particle antiparticle annihilation events are evidence of this
. 2 . a failure of time
orientability and particle antiparticle annihilation are indistinguishable .
they are alternative descriptions of the same phenomena .
time orientability is untestable .
non time orientability can not be an objective property of spacetime because the outcome of our test would depend upon the observer .
the conventional view of the experiment in figure [ fig : time2 ] is that it depicts a particle antiparticle annihilation .
this view has developed and now predominates for historical reasons .
it is an interpretation that conforms to the classical paradigm - modelling nature in terms of initial conditions , unique evolution and observer independent outcomes .
the classical paradigm is clearly inconsistent with a failure of time - orientation .
the explanatory power of a non - timeorientable spacetime in relation to interactions in geon models , electric charge and spin - half , adds weight to the interpretation of figure [ fig : time2 ] as a demonstration of the non - timeorientability of spacetime .
epsf ( 150,200 ) ( 15,-100 ) ( 98,125 ) ( 150,200 ) ( 15,-100 ) ( 40,102 ) ( 98,125 ) ( 150,200 ) ( 15,-100 ) ( 98,125 ) ( 40,121 ) ( 40,102 )
99 visser m 1996 .
woodbury , n.y .
hadley m j 1999 14811492 diemer t and hadley m j 1999 35673577 hadley m j 2000 41874194 sorkin r d 1977 717725 gibbons g w and herdeiro c a r 1999 36193652 hadley m j 1997 4360 geroch r 1968 17391744 ballentine l e 1989 .
prentice hall , new york ballentine l e 1970 358381 | contrary to established beliefs , spacetime may not be time - orientable . by considering an experimental test of time orientability
it is shown that a failure of time - orientability of a spacetime region would be indistinguishable from a particle antiparticle annihilation event . |
it is common for dynamical systems to have two or more coexisting attractors . in predicting the long - term behavior of a such a system , it is important to determine sets of initial conditions of orbits that approach each attractor ( i.e. , the basins of attraction ) .
the boundaries of such sets are often fractal ( @xcite , chapter 5 of @xcite , and references therein ) .
the fine - scale fractal structure of such a boundary implies increased sensitivity to errors in the initial conditions : even a considerable decrease in the uncertainty of initial conditions may yield only a relatively small decrease in the probability of making an error in determining in which basin such an initial condition belongs @xcite . for discussion of fractal basin boundaries in experiments ,
see chapter 14 of @xcite .
thompson and soliman @xcite showed that another source of uncertainty induced by fractal basin boundaries may arise in situations in which there is slow ( adiabatic ) variation of the system .
for example , consider a fixed point attractor of a map ( a node ) . as
a system parameter varies slowly , an orbit initially placed on the node attractor moves with time , closely following the location of the solution for the fixed point in the absence of the temporal parameter variation . as the parameter varies , the node attractor may suffer a saddle - node bifurcation .
for definiteness , say that the node attractor exists for values of the parameter @xmath0 in the range @xmath1 , and that the saddle - node bifurcation of the node occurs at @xmath2 .
now assume that , for a parameter interval @xmath3 $ ] with @xmath4 , in addition to the node , there are also two other attractors a and b , and that the boundary of the basin of attractor a , attractor b and the node is a fractal basin boundary .
we are interested in the typical case where , before the bifurcation , the saddle lies on the fractal basin boundary , and thus , at the bifurcation , the merged saddle - node orbit is on the basin boundary .
in such a case an arbitrarily small ball about the saddle - node at @xmath2 contains pieces of the basins of both a and b. thus , as @xmath0 slowly increases through @xmath5 , it is unclear whether the orbit following the node will go to a or to b after the node attractor is destroyed by the bifurcation . in practice , noise or round - off error may lead the orbit to go to one attractor or the other , and the result can often depend very sensitively on the specific value of the slow rate at which the system parameter varies .
we note that the study of orbits swept through an indeterminate saddle - node bifurcation belongs to the theory of dynamical bifurcations .
many authors have analyzed orbits swept through other bifurcations , like the period doubling bifurcation @xcite , the pitchfork bifurcation @xcite , and the transcritical bifurcation @xcite . in all these studies of the bifurcations listed above , the local structure before _ and _ after
the bifurcation includes stable invariant manifolds varying smoothly with the bifurcation parameter ( i.e. , a stable fixed point that exists before or after the bifurcation , and whose location varies smoothly with the bifurcation parameter ) .
this particular feature of the local bifurcation structure , not shared by the saddle - node bifurcation , allows for well - posed , locally defined , problems of dynamical bifurcations .
the static saddle - node bifurcation has received much attention in theory and experiments @xcite , but so far , no dynamical bifurcation problems have been defined for the saddle - node bifurcation . in this work , we demonstrate that , in certain common situations , global structure ( i.e. , an invariant cantor set or a fractal basin boundary ) adds to the local properties of the saddle - node bifurcation and allows for well - posed problems of dynamical bifurcations
. situations where a saddle - node bifurcation occurs on a fractal basin boundary have been studied in two dimensional poincar maps of damped forced oscillators @xcite .
several examples of such systems are known @xcite , and it seems that this is a common occurence in dynamical systems . in this work ,
we first focus on saddle - node bifurcations that occur for one parameter families of smooth one dimensional maps having multiple critical points ( a critical point is a point at which the derivative of the map vanishes ) .
since one dimensional dynamics is simpler than two dimensional dynamics , indeterminate bifurcations can be more simply studied , without the distraction of extra mathematical structure .
taking advantage of this , we are able to efficiently investigate several scaling properties of these bifurcations .
in particular , we investigate the scaling of ( 1 ) the fractal basin boundary of the static ( i.e. , unswept ) system near the saddle - node bifurcation ( secs .
[ sec : dim ] and [ sec : scdim ] ) , ( 2 ) the dependence of the orbit final destination on the sweeping rate ( sec .
[ sec : sweep ] ) , ( 3 ) the dependence of the time it takes for an attractor to capture the swept orbit following the bifurcation on the sweeping rate ( sec .
[ sec : capture ] ) , and ( 4 ) the dependence of the final attractor capture probability on the noise level ( sec .
[ sec : noise ] ) .
following our one - dimensional investigations , we explain that these results apply to two dimensional systems .
we show , through numerical experiments on the periodically forced duffing oscillator , that the scalings we have found also apply to higher dimensional systems ( sec . [
sec : duff ] ) . for one - dimensional maps , a situation dynamically similar to that in which there is indeterminacy in which attractor captures the orbit can also occur in cases where there are two rather than three ( or more ) attractors ( sec .
[ sec : model2 ] ) .
in particular , we can have the situation where one attractor persists for all values of the parameters we consider , and the other attractor is a node which is destroyed via a saddle - node bifurcation on the basin boundary separating the basins of the two attractors .
in such a situation , an orbit starting on the node , and swept through the saddle - node bifurcation , will go to the remaining attractor .
it is possible to distinguish different ways that the orbit initially on the node approaches the remaining attractor .
we find that the way in which this attractor is approached can be indeterminate .
we consider the general situation of a one dimensional real map @xmath6 depending on a parameter @xmath0 . we assume the following : ( 1 ) the map is twice differentiable with respect to @xmath7 , and once differentiable with respect to @xmath0 ( the derivatives are continuous ) ;
( 2 ) @xmath8 has at least two attractors sharing a fractal basin boundary for parameter values in the vicinity of @xmath5 ; and ( 3 ) an attracting fixed point @xmath9 of the map @xmath6 is destroyed by a saddle - node bifurcation as the parameter @xmath0 increases through a critical value @xmath5 , and this saddle - node bifurcation occurs on the common boundary of the basins of the two attractors .
we first recall the saddle - node bifurcation theorem ( see for example @xcite ) .
if the map @xmath6 satisfies : ( a ) @xmath10 , ( b ) @xmath11 , ( c ) @xmath12 , and ( d ) @xmath13 , then the map @xmath8 undergoes a backward saddle - node bifurcation ( i.e. , the node attractor is destroyed at @xmath9 as @xmath0 increases through @xmath5 ) . if the inequality in either ( c ) or ( d ) is reversed , then the map undergoes a forward saddle - node bifurcation , while , if both these inequalities are reversed , the bifurcation remains backward .
a saddle - node bifurcation in a one dimensional map is also called a tangent or a fold bifurcation . as an illustration of an indeterminate saddle - node bifurcation in a one - dimensional map , we construct an example in the following way .
we consider the logistic map for a parameter value where there is a stable period three orbit .
we denote this map @xmath14 and its third iterate @xmath15}(x)$ ] .
the map @xmath15}(x)$ ] has three stable fixed points .
we perturb the map @xmath15}(x)$ ] by adding a function ( which depends on a parameter @xmath0 ) that will cause a saddle - node bifurcation of one of the attracting fixed points but not of the other two [ see figs .
1(a ) and 1(b ) ] .
we investigate @xmath16}(x)+\mu\sin(3\pi x),\quad \mbox{where } \quad g(x)=3.832\,x(1-x).\end{aligned}\ ] ] numerical calculations show that the function @xmath6 satisfies all the conditions of the saddle - node bifurcation theorem for having a backward saddle - node bifurcation at @xmath17 and @xmath18 .
figure 2(a ) displays how the basins of the three attracting fixed points of the map @xmath8 change with variation of @xmath0 . for @xmath19
the third iterate of the logistic map is unperturbed , and it has three attracting fixed points whose basins we color - coded with blue , green and red . for every value of @xmath0
, the red region @xmath20 $ ] is the set of initial conditions attracted to the rightmost stable fixed point which we denote @xmath21 .
the green region @xmath22 $ ] is the set of initial conditions attracted to the middle stable fixed point which we denote @xmath23 .
the blue region @xmath24 $ ] is the set of initial conditions attracted to the leftmost stable fixed point which we denote @xmath25 . for @xmath1 ,
each of these colored sets has infinitely many disjoint intervals and a fractal boundary . as @xmath0 increases
, the leftmost stable fixed point @xmath25 is destroyed via a saddle - node bifurcation on the fractal basin boundary .
in fact , in this case , for @xmath26 , every boundary point of one basin is a boundary point for all three basins .
( that is , an arbitrarily small @xmath7-interval centered about any point on the boundary of any one of the basins contains pieces of the other two basins . )
the basins are so - called wada basins @xcite .
this phenomenon of a saddle - node bifurcation on the fractal boundary of wada basins also occurs for the damped forced oscillators studied in refs .
alternatively , if we look at the saddle - node bifurcation as @xmath0 decreases through the value @xmath5 , then the basin @xmath24 $ ] of the newly created stable fixed point immediately has infinitely many disjoint intervals and its boundary displays fractal structure . according to the terminology of robert et al .
@xcite , we may consider this bifurcation an example of an ` explosion ' .
figure 3 graphs the computed dimension @xmath27 of the fractal basin boundary versus the parameter @xmath0 . for @xmath1 , we observe that @xmath27 appears to be a continuous function of @xmath0 .
park et al .
@xcite argue that the fractal dimension of the basin boundary near @xmath5 , for @xmath1 , scales as @xmath28 with @xmath29 the dimension at @xmath2 ( @xmath29 is less than the dimension of the phase space ) , and @xmath30 a positive constant .
figure 3 shows that the boundary dimension @xmath27 experiences a discontinuous jump at the saddle - node bifurcation when @xmath2 .
we believe that this is due to the fact that the basin @xmath24 $ ] suddenly disappears for @xmath31 .
the existence of a fractal basin boundary has important practical consequences .
in particular , for the purpose of determining which attractor eventually captures a given orbit , the arbitrarily fine - scaled structure of fractal basin boundaries implies considerable sensitivity to small errors in initial conditions .
if we assume that initial points can not be located more precisely than some @xmath32 , then we can not determine which basin a point is in , if it is within @xmath33 of the basin boundary .
such points are called @xmath33-uncertain .
the lebesgue measure of the set of @xmath33-uncertain points ( in a bounded region of interest ) scales like @xmath34 , where @xmath35 is the dimension of the phase space ( @xmath36 for one dimensional maps ) and @xmath27 is the box - counting dimension of the basin boundary @xcite . for the case of a fractal basin boundary @xmath37 .
when @xmath38 is small , a large decrease in @xmath33 results in a relatively small decrease in @xmath34 .
this is discussed in ref .
@xcite which defines the uncertainty dimension , @xmath39 , as follows .
say we randomly pick an initial condition @xmath7 with uniform probability density in a state - space region @xmath40 .
then we randomly pick another initial condition @xmath41 in @xmath40 , such that @xmath42 .
let @xmath43 be the probability that @xmath7 and @xmath41 are in different basins .
[ we can think of @xmath43 as the probability that an error will be made in determing the basin of an initial condition if the initial condition has uncertainty of size @xmath33 . ]
the uncertainty dimension of the basin boundary @xmath39 is defined as the limit of @xmath44 as @xmath33 goes to zero @xcite .
thus , the probability of error scales as @xmath45 , where for fractal basin boundaries @xmath46 .
this indicates enhanced sensitivity to small uncertainty in initial conditions .
for example , if @xmath47 , then a decrease of the initial condition uncertainty @xmath33 by a factor of 10 leads to only a relative small decrease in the final state uncertainty @xmath43 , since @xmath48 decreases by a factor of about @xmath49 .
thus , in practical terms , it may be essentially impossible to significantly reduce the final state uncertainty . in ref .
@xcite it was conjectured that the box - counting dimension equals the uncertainty dimension for basin boundaries in typical dynamical systems . in ref .
@xcite it is proven that the box - counting dimension , the uncertainty dimension and the hausdorff dimension are all equal for the basin boundaries of one and two dimensional systems that are uniformly hyperbolic on their basin boundary .
we now explain some aspects of the character of the dependence of @xmath27 on @xmath0 ( see fig . 3 ) .
from refs .
@xcite it follows that the box - counting dimension and the hausdorff dimension coincide for all intervals of @xmath0 for which the map @xmath8 is hyperbolic on the basin boundary , and that the dimension depends continuously on the parameter @xmath0 in these intervals . for @xmath50 , there are many parameter values for which the map has a saddle - node bifurcation of a periodic orbit on the fractal basin boundary . at such parameter values , which we refer to as saddle - node bifurcation parameter values ,
the dimension is expected to be discontinuous ( as it is at the saddle - node bifurcation of the fixed point , @xmath2 , see fig . 3 ) .
in fact , there exist sequences of saddle - node bifurcation parameter values converging to @xmath5 @xcite .
furthermore , for each parameter value @xmath50 for which the map undergoes a saddle - node bifurcation , there exists a sequence of saddle - node bifurcation parameter values converging to that parameter value .
the basins of attraction of the periodic orbits created by saddle - node bifurcations of high period exist only for very small intervals of the parameter @xmath0 .
we did not encounter them numerically by iterating initial conditions for a discrete set of values of the parameter @xmath0 , as we did for the basin of our fixed point attractor .
just past @xmath5 , the remaining green and red basins display an alternating stripe structure [ see fig . 2(b ) ] .
the red and green stripes are interlaced in a fractal structure .
as we approach the bifurcation point , the interlacing becomes finer and finer scaled , with the scale approaching zero as @xmath0 approaches @xmath5 .
similar fine scaled structure is present in the neighborhood of all preiterates of @xmath9 .
if one changes the horizontal axis of figs .
2(a , b ) from @xmath0 to @xmath51 , then , the complex alternating stripe structure appears asymptotically periodic [ see fig .
[ thus , with identical horizontal scale , the dimension plot in fig .
4(b ) appears asymptotically periodic , as well .
] we now explain why this is so .
we restrict our discussion to a small neighborhood of @xmath9 .
consider the second order expansion of @xmath8 in the vicinity of @xmath9 and @xmath5 @xmath52 and @xmath53 .
the trajectories of @xmath54 in the neighborhood of @xmath55 , for @xmath56 close to zero , are good approximations to trajectories of @xmath8 in the neighborhood of @xmath57 , for @xmath0 close to @xmath5 .
assume that we start with a certain initial condition for @xmath54 , @xmath58 , and we ask the following question : what are all the positive values of the parameter @xmath56 such that a trajectory passes through a fixed position @xmath59 at some iterate @xmath60 ?
for any given @xmath61 which is not on the fractal basin boundary , there exists a range of @xmath0 such that iterates of @xmath61 under @xmath8 evolve to the same final attractor , for all values of @xmath0 in that range .
in particular , once @xmath62 appreciably exceeds @xmath56 , the subsequent evolution is approximately independent of @xmath56 .
thus , we can choose @xmath63 , but still small enough so that it lies in the region of validity of the canonical form .
there exists a range of such @xmath64 values satisfying these requirements provided that @xmath65 is small enough . since consecutive iterates of @xmath54 in the neighborhood of @xmath55 for @xmath56 close to zero differ only slightly , we approximate the one dimensional map , @xmath66 by the differential equation @xcite , @xmath67 where in @xmath60 is considered as a continuous , rather than a discrete , variable . integrating from @xmath68 to @xmath64 yields @xmath69 close to the saddle - node bifurcation ( i.e. , @xmath70 , and @xmath71 close to zero ) ,
@xmath54 is a good approximation to @xmath8 . for @xmath72 eq .
becomes @xmath73 the values of @xmath74 satisfying eq .
increase with @xmath60 in step of @xmath75 . for our example
we have @xmath53 , thus @xmath76 .
counting many periods like those in fig . 4 in the region of @xmath77 , the closest critical point to @xmath9
[ see fig .
4(a ) ] , we find that the period of the stripe structure is 3.015 , which is in good agreement with our theoretical value . in order to investigate the structure of the fractal basin boundary in the vicinity of the saddle - node bifurcation ( i.e. , @xmath68 close to @xmath78 ) , we consider in the case where we demand only @xmath79 .
thus , eq . becomes @xmath80 let @xmath81 denote the solution of eq .
for @xmath82 .
equation implies the behavior of @xmath81 as function of @xmath83 and @xmath60 as sketched in fig .
5 . for a fixed @xmath60
, @xmath74 has a horizontal asymptote at the value @xmath84 as @xmath85 , and a vertical asymptote to infinity at @xmath86 . for @xmath87
, we have an infinite number of values of the parameter @xmath88 , for which an orbit of @xmath54 starting at @xmath83 passes through the same position @xmath89 , after some number of iterations . for @xmath90 ( i.e. , @xmath91 )
, we also have an infinite number of @xmath92 , but with constant step @xmath93 rather than @xmath75 ( see the intersections marked with black dots in the fig . 5 ) .
this is hard to verify from numerics , since @xmath94 increases with @xmath95 , and the stripes become very tilted in the neighborhood of @xmath96 .
[ see fig .
4(a ) , where the approximate positions of @xmath77 and @xmath9 on the vertical axis are indicated . ] for @xmath97 , @xmath74 has only a limited number of values with @xmath98 . in order to understand the consequences of a saddle - node bifurcation on a fractal basin boundary for systems experiencing slow drift , we imagine the following experiment . we start with the dynamical system @xmath8 at parameter @xmath99 , with @xmath100 on the attractor to be destroyed at @xmath2 by a saddle - node bifurcation ( i.e. , @xmath25 ) .
then , as we iterate , we slowly change @xmath0 by a small constant amount @xmath101 per iterate , thus increasing @xmath0 from @xmath102 to @xmath103 , @xmath104 when @xmath105 we stop sweeping the parameter @xmath0 , and , by iterating further , we determine to which of the remaining attractors of @xmath106 the orbit goes . numerically , we observe that , if @xmath107 is not too small , then , by the time @xmath108 is reached , the orbit is close to the attractor of @xmath106 to which it goes .
[ from our subsequent analysis , ` not too small @xmath109 ' translates to choices of @xmath101 that satisfy @xmath110 .
] we repeat this for different values of @xmath101 and we graph the final attractor position for the orbit versus @xmath101 [ see fig .
for convenience in the graphical representation of figs .
6(a , b ) , we have represented the attractor of the green region @xmath22 $ ] , denoted @xmath111 , as a 0 , and the attractor of the red region @xmath20 $ ] , denoted @xmath112 , as a 1 . in fig .
6(a ) we use of 25,000 points having the vertical coordinate either 0 or 1 , which we connect with straight lines . in an interval of @xmath101 for which the system reaches the same final attractor ( either 0 or 1 ) , the lines connecting the points are horizontal .
such intervals appear as white bands in fig . 6 , if they are wider than the width of the plotted lines connecting 0 s and 1 s .
for example , in fig 6(a ) , the white band centered at @xmath113 has at the bottom a thick horizontal line , which indicates that for the whole of that interval , the orbit reaches the attractor @xmath111 which we represented by 0 .
adjacent intervals of width less than the plotted lines appear as black bands . within such black bands ,
an uncertainty in @xmath101 of size equal to the width of the plotted line makes the attractor that the orbit goes to indeterminate .
figure 6(a ) shows that the widths of the white bands decrease as @xmath101 decreases , such that , for small @xmath101 , we see only black . if @xmath107 is large enough ( i.e. , @xmath114 ) , numerics and our subsequent analysis show that fig .
6 is independent of @xmath108 .
this fact can be understood as follows .
once @xmath115 , the orbit typically lands in the green or the red basin of attraction and goes to the corresponding attractor .
due to sweeping , it is possible for the orbit to switch from being in one basin of attraction of the _ time - independent _ map @xmath8 to the other , since the basin boundary between @xmath22 $ ] and @xmath20 $ ] changes with @xmath0 .
however , the sweeping of @xmath0 is slow ( i.e. , @xmath101 is small ) , and , once @xmath116 is large enough , the orbit is far enough from the fractal basin boundary , and the fractal basin boundary changes too little to switch the orbit between @xmath22 $ ] and @xmath20 $ ] .
we also find numerically that figs .
6(a , b ) are independent of the initial condition @xmath100 , provided that it is in the blue basin @xmath117 $ ] , sufficiently far from the fractal basin boundary , and that @xmath118 is not too small ( i.e. , @xmath119 ) . if one changes the horizontal scale of fig .
6(a ) from @xmath101 to @xmath120 [ see fig .
6(b ) ] , the complex band structure appears asymptotically periodic .
furthermore , we find that the period in @xmath121 of the structure in fig .
6(b ) asymptotically approaches @xmath122 as @xmath101 becomes small . in order to explain this result
, we again consider the map @xmath54 , the local approximation of @xmath8 in the region of the saddle - node bifurcation .
equations can be approximated by @xmath123 we perform the following numerical experiment .
we consider orbits of our approximate two dimensional map given by eq .
starting at @xmath124 .
we define a final state function of an orbit swept with parameter @xmath101 in the following way .
it is 0 if the orbit has at least one iterate in a specified fixed interval far from the saddle - node bifurcation , and is 1 , otherwise .
in particular , we take the final state of a swept orbit to be 0 if there exists @xmath60 such that @xmath125 , and to be 1 otherwise .
figure 6(c ) graphs the corresponding numerical results .
similar to fig .
6(b ) , we observe periodic behavior in @xmath120 with period @xmath126 .
in contrast to fig .
6(b ) where the white band structure seems fractal , the structure within each period in fig .
6(c ) consists of only one interval where the final state is 0 and one interval where the final state is 1 .
this is because @xmath127 is a single interval , while the green basin [ denoted 0 in fig .
6(b ) ] has an infinite number of disjoint intervals and a fractal boundary ( see fig . 2 ) .
+ with the similarity between figs .
6(b ) and 6(c ) as a guide , we are now in a position to give a theoretical analysis explaining the observed periodicity in @xmath120 . in particular , we now know that this can be explained using the canonical map , and that the periodicity result is thus universal [ i.e. , independent of the details of our particular example , eq . ]
. for slow sweeping ( i.e. , @xmath101 small ) , consecutive iterates of in the vicinity of @xmath55 and @xmath128 differ only slightly , and we further approximate the system by the following ricatti differential equation , @xmath129 the solution of eq . can be expressed in terms of the airy functions @xmath130 and @xmath131 and their derivatives , denoted by @xmath132 and @xmath133 , @xmath134 where @xmath135 and @xmath136 is a constant to be determined from the initial condition .
we are only interested in the case of slow sweeping , @xmath137 , and @xmath138 ( which is the stable fixed point of @xmath139 destroyed by the saddle - node bifurcation at @xmath140 ) .
in particular , we will consider the case where @xmath141 and @xmath142 ( i.e. , @xmath143 ) .
using @xmath144 to solve for @xmath136 yields @xmath145\gg 1 $ ] . for positive large values of @xmath146 ( i.e. , for @xmath60 small enough ) , using the corresponding asymptotic expansions of the airy functions @xcite , the lowest order in @xmath101 approximation to is @xmath147 with the correction term of higher order in @xmath101 being negative .
thus , for @xmath60 sufficiently smaller than @xmath148 , the swept orbit lags closely behind the fixed point for @xmath139 with @xmath56 constant . for @xmath149 , we use the fact that @xmath136 is large to approximate as @xmath150 note that @xmath151 gives the lag of the swept orbit relative to the fixed point attractor evaluated at the saddle - node bifurcation .
equation does not apply for @xmath152 , where @xmath153 is the value of @xmath60 for which @xmath154 , the largest root of @xmath155 ( i.e. , @xmath156 ) . at @xmath157 ,
the normal form approximation predicts that the orbit diverges to @xmath158 .
thus , for @xmath60 near @xmath153 , the normal form approximation of the dynamical system ceases to be valid .
note , however , that can be valid even for @xmath146 close to @xmath159 .
this is possible because @xmath101 is small .
in particular , we can consider times up to the time @xmath160 where @xmath160 is determined by @xmath161 , ( @xmath162 is small , ) provided @xmath163 so that the normal form applies .
that is , we require @xmath164(\delta\mu / a^2)^{1/3}\ll 1 $ ] , which can be satisfied even if @xmath164 $ ] is large .
furthermore , we will take the small quantity @xmath165 to be not too small ( i.e. , @xmath166 ) , so that @xmath167 .
we then consider in the range , @xmath168 , where the normal form is still valid .
we use eq . for answering the following question : what are all the values of the parameter @xmath101 ( @xmath101 small ) for which an orbit passes exactly through the same position @xmath59 , at some iterate @xmath169 ?
all such orbits would further evolve to the same final attractor , independent of @xmath101 , provided @xmath170 ; i.e. , @xmath64 is large enough that @xmath171 does not much influence the orbit after @xmath172 reaches @xmath64 .
[ denote @xmath173 as @xmath174 . ] using we can estimate when this occurs , @xmath175 ^ 2 ( \delta\mu^2/a)^{1/3}\gg ( \hat\mu_s+ n_f\,\delta\mu)$ ] or @xmath176 ^ 2\gg \xi_f$ ] .
this inequality is satisfied when @xmath177 gets near @xmath178 , which is the largest zero of @xmath130 ( i.e. , @xmath179 , where @xmath165 is a small positive quantity ) .
we now rewrite eq . in the following way @xmath180^{1/3}k\left [ \left(\frac{a^2}{\delta\mu}\right)^{1/3}\,\hat x_f\right]},\end{aligned}\ ] ] representing a transcedental equation in @xmath101 where @xmath181 and @xmath64 are fixed
, @xmath169 is a large positive integer ( i.e. , @xmath182 is the integer part of @xmath183 ) , and @xmath184 is the inverse function of @xmath185 in the neighborhood of @xmath186 .
thus @xmath187|\lesssim |k(\infty)|=|\tilde\xi|$ ] .
the difference @xmath188 $ ] , where @xmath189 is the solution of eq . , yields the limit period of the attracting state versus @xmath120 graph ( see fig .
we denote this limit period by @xmath190 . for small @xmath101 , the term involving @xmath191 $ ] in eq .
can be neglected , and we get @xmath192 . figure 7 graphs numerical results of @xmath193^{-1}$ ] versus @xmath102 for our map example given by eq . .
the fit line is @xmath193^{-1}=-0.9986\mu_s+0.0028 $ ] , which agrees well with the prediction of the above analysis and our numerical value for @xmath0 at the bifurcation , @xmath18 .
an alternate point of view on this scaling property is as follows . for @xmath194
( i.e. , @xmath1 ) and slow sweeping ( i.e. , @xmath101 small ) , the orbit closely follows the stable fixed point attractor of @xmath54 , until @xmath195 , and the saddle - node bifurcation takes place . however , due to the discreteness of @xmath60 , the first nonnegative value of @xmath56 depends on @xmath196 and @xmath101 ( see fig .
now consider two values of @xmath101 , one @xmath197 satisfying @xmath198 , and another @xmath199 satisfying @xmath200 . because @xmath197 and @xmath199 are very close ( for large @xmath201 )
_ and _ both lead @xmath202 to pass through @xmath203 ( one at time @xmath204 , and the other at time @xmath205 ) , it is reasonable to assume that their orbits for @xmath206 are similar ( except for a time shift @xmath207 ) ; i.e. , they go to the same attractor .
thus , the period of @xmath120 is approximately @xmath208 .
+ we now consider the intervals of @xmath120 between the centers of consecutive wide white bands in fig .
figure 9 graphs the calculated fractal dimension @xmath209 of the boundary between white bands in these consecutive intervals versus their center value of @xmath120 . from fig . 9
, we see that as @xmath120 increases , the graph of the fractal dimension @xmath209 does not converge to a definite value , but displays further structure .
nevertheless , numerics show that as @xmath120 becomes large ( i.e. , in the range of @xmath210 ) , @xmath209 varies around the value 0.952 .
this is consistent with the numerics presented in fig .
4(b ) which graphs the dimension of the fractal basin boundary for the time - independent map @xmath8 , at fixed values of the parameter @xmath0 where @xmath50 . thus , for large @xmath120
, @xmath209 provides an estimate of the dimension of the fractal basin boundary in the absence of sweeping at @xmath50 .
+ we now discuss a possible experimental application of our analysis .
the conceptually most straightforward method of measuring a fractal basin boundary would be to repeat many experiments each with precisely chosen initial conditions . by determining the final attractor corresponding to each initial condition , basins of attraction
could conceivably be mapped out @xcite .
however , it is commonly the case that accurate control of initial conditions is not feasible for experiments . thus , the application of this direct method is limited , and , as a consequence , fractal basin boundaries have received little experimental study , in spite of their fundamental importance . if a saddle - node bifurcation occurs on the fractal basin boundary , an experiment can be arranged to take advantage of this . in this case , the purpose of the experiment would be to measure the dimension @xmath209 as an estimate of the fractal dimension of the basin boundary @xmath27 .
the measurements would determine the final attractor of orbits starting at the attractor to be destroyed by the saddle - node bifurcation , and swept through the saddle - node bifurcation at different velocities ( i.e. , the experimental data corresponding to the numerics in fig . 6 ) .
this does not require precise control of the initial conditions of the orbits .
it is sufficient for the initial condition to be in the basin of the attractor to be destroyed by the saddle - node bifurcation ; after enough time , the orbit will be as close to the attractor as the noise level allows .
then , the orbit may be swept through the saddle - node bifurcation .
the final states of the orbits are attractors ; in their final states , orbits are robust to noise and to measurement perturbations .
the only parameters which require rigorous control are the sweeping velocity ( i.e. , @xmath101 ) and the initial value of the parameter to be swept ( i.e. , @xmath102 ) ; precise knowledge of the parameter value where the saddle - node bifurcation takes place ( i.e. , @xmath5 ) is not needed .
[ it is also required that the noise level be sufficiently low ( see sec .
[ sec : noise ] ) . ]
a question of interest is how much time it takes for a swept orbit to reach the final attracting state .
namely , we ask how many iterations with @xmath50 are needed for the orbit to reach a neighborhood of the attractor having the green basin .
due to slow sweeping , the location of the attractor changes slightly on every iterate .
if @xmath211 is a fixed point attractor of @xmath8 ( with @xmath0 constant ) , then a small change @xmath101 in the parameter @xmath0 , yields a change in the position of the fixed point attractor , @xmath212 we consider the swept orbit to have reached its final attractor if consecutive iterates differ by about @xmath213 ( which is proportional to @xmath101 ) . for numerical purposes , we consider that the orbit has reached its final state if @xmath214 . in our numerical experiments ,
this condition is satisfied by every orbit before @xmath0 reaches its final value @xmath108 .
we refer to the number of iterations with @xmath50 needed to reach the final state as the _ capture time _ of the corresponding orbit .
figure 10 plots the capture time by the attractor @xmath111 [ having the green basin in fig . 2 ] versus @xmath120 for a range corresponding to one period of the structure in fig .
no points are plotted for values of @xmath101 for which the orbit reaches the attractor @xmath112 .
the capture time graph has fractal features , since for many values of @xmath101 the orbit gets close to the fractal boundary between @xmath20 $ ] and @xmath22 $ ] . using the fact that the final destination of the orbit versus @xmath120 is asymptotically periodic
[ see fig .
6(b ) ] , we can provide a further description of the capture time graph .
we consider the series of the largest intervals of @xmath120 for which the orbit reaches the attractor @xmath111 [ see fig .
6(b ) ; we refer to the wide white band around @xmath215 and the similar ones which are ( asymptotically ) separated by an integer number of periods ] .
orbits swept with @xmath101 at the centers of these intervals spend only a small number of iterations close to the common fractal boundary of @xmath20 $ ] and @xmath22 $ ] .
thus , the capture time of such similar orbits does not depend on the structure of the fractal basin boundary .
we use eq . as an approximate description of these orbits .
a swept orbit reaches its final attracting state as @xmath216 becomes large .
then , the orbit is rapidly trapped in the neighborhood of one of the swept attractors of @xmath8 .
thus , we equate the argument of the airy function in the denominator to its first root [ see ] , solve for @xmath60 , and substract @xmath148 ( the time for @xmath56 to reach the bifurcation value ) .
this yields the following approximate formula for the capture time @xmath217 where @xmath156 is the largest root of the airy function @xmath130 .
thus , we predict that for small @xmath101 , a log - log plot of the capture time of the selected orbits versus @xmath101 is a straight line with slope -1/3 .
figure 11 shows the corresponding numerical results .
the best fitting line ( not shown ) has slope -0.31 , in agreement with our prediction @xcite .
we now consider the addition of noise .
thus , we change our swept dynamical system to @xmath218 where @xmath219 is random with uniform probability density in the interval @xmath220 $ ] , and @xmath221 is a parameter which we call the noise amplitude . see fig .
6(a ) which shows the numerical results of the final destination of the orbits versus @xmath101 in the case @xmath222 .
the graph exhibits fractal features of structure at arbitrarily small scales .
the addition of small noise is expected to alter this structure , switching the final destination of orbits . in this case , it is appropriate to study the probability of orbits reaching one of the final destinations . for every @xmath221 , we compute the final attractor of a large number of orbits having identical initial condition and parameters , but with different realizations of the noise .
we estimate the probability that an orbit reaches a certain attractor by the fraction of such orbits that have reached the specified attractor in our numerical simulation .
figure 12 graphs the probability that an orbit reaches the attractor @xmath111 versus the noise amplitude @xmath221 .
we present five graphs corresponding to five different values of @xmath101 equally spaced in a range of @xmath223 centered at @xmath224 ( i.e. , @xmath225 , @xmath226 and @xmath227 ) .
we notice that the probability graphs have different shapes , but a common horizontal asymptote in the limit of large noise .
the value of the horizontal asymptote , approximately equal to 0.5 , is related to the relative measure of the corresponding basin . as in the previous subsection
, we take advantage of the asymptotically periodic structure of the noiseless final destination graph versus @xmath120 [ see fig .
we consider centers of the largest intervals of @xmath120 for which an orbit reaches the middle attractor in the absence of noise .
we chose five such values of @xmath101 , spread over two decades , where the ratio of consecutive values is approximately 3 .
figure 13(a ) graphs the probability that an orbit reaches the middle fixed point attractor versus the noise amplitude @xmath221 , for the five selected values of @xmath101 .
we notice that all the curves have qualitatively similar shape . for a range from zero to small @xmath221 ,
the probability is 1 , and as @xmath221 increases , the probability decreases to a horizontal asymptote .
the rightmost curve in the family corresponds to the largest value of @xmath101 ( @xmath228 ) , and the leftmost curve corresponds to the smallest value of @xmath101 ( @xmath229 ) .
figure 13(b ) shows the same family of curves as in fig .
13(a ) , but with the horizontal scale changed from @xmath221 to @xmath230 .
all data collapse to a single curve , indicating that the probability that a swept orbit reaches the attractor @xmath111 depends only on the reduced variable @xmath230 .
later , we provide a theoretical argument for this scaling . in order to gain some understanding of this result ,
we follow the idea of sec . [
sec : sweep ] , and use the canonical form @xmath231 to propose a simplified setup of our problem .
we modify by the addition of a noise term @xmath232 in the right hand side of the first equation of .
we are interested in the probability that a swept orbit has at least one iterate , @xmath233 , in a specified fixed interval far from the vicinity of the saddle - node bifurcation .
more precisely , we analyze how this probability changes versus @xmath221 and @xmath101 .
depending on the choice of interval and the choice of @xmath101 , the probability versus @xmath221 graph ( not shown ) has various shapes . for numerical purposes ,
we choose our fixed interval to be the same as that of sec .
[ sec : sweep ] , @xmath234 .
we then select values of @xmath101 for which a noiseless swept orbit , starting at @xmath124 , reaches exactly the center of our fixed interval .
the inverse of these values of @xmath101 are centers of intervals where the final state of the swept orbits is 0 [ see fig .
we consider five such values of @xmath101 , where the ratio of consecutive values is approximately 3 .
figure 14(a ) shows the probability that a swept orbit has an iterate in our fixed interval versus the noise amplitude for the selected values of @xmath101 .
figure 14(a ) shares the qualitative characteristics of fig .
13(a ) , with the only noticeable difference that the value of the horizontal asymptote is now approximately 0.1 . figure 14(b ) shows the same family of curves as in fig .
14(a ) , where the horizontal scale has been changed from @xmath221 to @xmath230 . as for fig .
12(b ) , this achieves good collapse of the family of curves .
+ we now present a theoretical argument for why the probability of reaching an attractor depends on @xmath101 and @xmath221 only through the scaled variable @xmath230 when @xmath101 and @xmath221 are small . from our results in figs .
14 , we know that the scaling we wish to demonstrate should be obtainable by use of the canonical form @xmath231 .
accordingly , we again use the differential equation approximation , but with a noise term added , @xmath235 where @xmath236 is white noise , @xmath237 and we have redefined the origin of the time variable @xmath60 so that the parameter @xmath56 sweeps through zero at @xmath238 ( i.e. , we replaced @xmath60 by @xmath239 ) . because we are only concerned with scaling , and not with the exact solution of , a fairly crude analysis will be sufficient .
first we consider the solution of with the noise term omitted , and the initial condition [ see ] @xmath240 we define a characteristic point of the orbit , @xmath241 , where @xmath242 . for @xmath243 , @xmath244 , and we can approximate the noiseless orbit as @xmath245 where @xmath246 is a slowly varying function of @xmath60 of order 1 ( @xmath247 for @xmath243 ) . setting @xmath248 , we find that @xmath249 is given by @xmath250 corresponding to [ c.f . , eq . ] @xmath251 for @xmath252 ( i.e. , @xmath253 ) , eq . can be approximated as @xmath254 .
starting at @xmath255 , integration of this equation leads to explosive growth of @xmath172 to infinity in a time of order @xmath256 , which is of the same order as @xmath249 .
thus , the relevant time scale is @xmath256 [ this agrees with eq . in sec .
[ sec : capture ] ] . now consider the action of noise .
for @xmath243 , we neglect the nonlinear term @xmath62 , so that becomes @xmath257 .
the solution of this equation is the linear superposition of the solutions of @xmath258 and @xmath259 , or @xmath260 ; @xmath261 is given by @xmath262 , and @xmath263 is a random walk .
thus , for @xmath243 , there is diffusive spreading of the probability density of @xmath172 , @xmath264 this diffusive spreading can blur out the structure in fig .
how large does the noise amplitude @xmath221 have to be to do this ?
we can estimate @xmath221 by noting that the periodic structure in figs .
6(b , c ) results from orbits that take different integer times to reach @xmath265 .
thus , for @xmath266 we define a scale @xmath267 in @xmath172 corresponding to the periodicity in @xmath120 by [ c.f . , eq . ] @xmath268 which yields @xmath269 if by the time @xmath266 , the diffusive spread of the probability density of @xmath172 becomes as large as @xmath267 , then the noise starts to wash out the periodic variations with @xmath120 . setting @xmath270 from to be of the order of @xmath267 from , we obtain @xmath271 , which with yields @xmath272 thus , we expect a collapse of the two parameter @xmath273 data in fig .
14(a ) by means of a rescaling of @xmath221 by @xmath101 raised to an exponent 5/6 [ i.e. , @xmath230 ] .
in this section we demonstrate the scaling properties of sweeping through an indeterminate saddle - node bifurcation in the case of the periodically forced duffing oscillator @xcite , @xmath275 the unforced duffing system ( i.e. , @xmath19 ) is an example of an oscillator in a double well potential .
it has two coexisting fixed point attractors corresponding to the two minima of the potential energy . for small @xmath0 , the forced duffing oscillator has two attracting periodic orbits with the period of the forcing ( i.e. , @xmath276 ) , one in each well of the potential . at @xmath277 , a new attracting periodic orbit of period @xmath278
arises through a saddle - node bifurcation . in ref .
@xcite , it is argued numerically that for a certain range of @xmath50 the basin of attraction of the @xmath278 periodic orbit and the basins of attraction of the @xmath276 periodic orbits have the wada property .
thus , as @xmath0 decreases through the critical value @xmath5 , the period @xmath278 attractor is destroyed via a saddle - node bifurcation on the fractal boundary of the basins of the other two attractors .
this is an example of an indeterminate saddle - node bifurcation of the duffing system which we study by considering the two - dimensional map in the @xmath279 plane resulting from a poincar section at constant phase of the forcing signal . we consider orbits starting in the vicinity of the period three fixed point attractor , and , as we integrate the duffing system , we decrease @xmath0 from @xmath280 to @xmath281 at a small rate of @xmath101 per one period of the forcing signal .
as @xmath0 approaches @xmath5 , ( with @xmath50 , ) the period three fixed point attractor of the unswept duffing system approaches its basin boundary , and the slowly swept orbit closely follows its location . for @xmath282 small , the orbit will approximately follow the one - dimensional unstable manifold of the @xmath2 period three saddle - node pair .
thus , we can describe the sweeping through the indeterminate bifurcation of the duffing oscillator by the theory we developed for one dimensional discrete maps .
figure 15 shows the final destination graph of a swept orbit initially situated in the vicinity of the period three fixed point of the poincar map .
the final attracting state is represented as a 1 if situated in the potential well where @xmath283 , and is represented as a 0 if situated in the potential well where @xmath284 .
as expected , the structure in fig .
15 appears asymptotically periodic if graphed versus @xmath285 in addition to slowly sweeping the duffing system , consider an additive noise term @xmath286 in the right hand side of , where on every time step @xmath287 is chosen randomly in @xmath220 $ ] , and the time step used is @xmath288 .
figure 16(a ) shows the dependence of the probability of approaching the attractor represented as a 1 versus the noise amplitude @xmath221 for three specially selected values of @xmath101 ( centers of white bands in the structure of fig .
15 where the swept orbit reaches the attracting state represented by 1 ) spread over one decade . figure 16(b ) shows collapse of the data in fig .
16(a ) to a single curve when the noise amplitude @xmath221 is rescaled by @xmath289 , as predicted by our previous one - dimensional analysis ( sec .
[ sec : noise ] ) .
thus , we believe that the scaling properties of the indeterminate saddle - node bifurcation we found in one - dimensional discrete maps are also shared by higher dimensional flows .
in this section we consider the case of a one dimensional map @xmath8 having two attractors a and b , one of which ( i.e. , a ) exists for all @xmath290 $ ] . the other ( i.e. , b ) is a node which is destroyed by a saddle - node bifurcation on the boundary between the basins of a and b , as @xmath0 increases through @xmath5 ( @xmath291 $ ] ) .
when an orbit is initially on b , and @xmath0 is slowly increased through @xmath5 , the orbit will always go to a ( which is the only attractor for @xmath50 ) .
however , it is possible to distinguish between two ( or more ) different ways of approaching a. [ in particular , we are interested in ways of approach that can be distinguished in a coordinate - free ( i.e. , invariant ) manner . ] as we show in this section , the way in which a is approached can be indeterminate . in this case
, the indeterminacy is connected with the existence of an invariant nonattracting cantor set embedded in the basin of a for @xmath50 . as an illustration ,
we construct the following model @xmath292 calculations show that @xmath8 satisfies all the requirements of the saddle - node bifurcation theorem for undergoing a backward saddle - node bifurcation at @xmath293 and @xmath294 .
figure 17(a ) shows the graph of @xmath8 versus @xmath7 at @xmath295 .
figure 17(b ) shows how the basin structure of the map @xmath8 varies with the parameter @xmath0 . for positive values of @xmath0
, @xmath8 has only one attractor which is at minus infinity .
the basin of this attractor is the whole real axis .
as @xmath0 decreases through @xmath294 , a new fixed point attractor is created at @xmath293 .
the basin of attraction of this fixed point has infinitely many disjoint intervals displaying fractal features [ indicated in black in fig .
this is similar to the blue basin @xmath24 $ ] of the attractor @xmath25 of the previous one - dimensional model ( see sec .
[ model1 ] ) . the blue region in fig .
18(a ) is the basin of attraction of the stable fixed point destroyed as @xmath0 increases through @xmath5 . for every value of @xmath0 we consider , the map @xmath8 has invariant cantor sets .
the trajectories of points which are located on an invariant cantor set , do not diverge to infinity .
one way to display such cantor sets , is to select uniquely defined intervals whose end points are on the cantor set .
for example , fig .
18(a ) shows green and red regions .
for every fixed parameter value @xmath0 , the collection of points that are boundary points of the red and green regions , constitutes an invariant cantor set . in order to describe these green and red regions ,
we introduce the following notations . for each parameter value @xmath0 ,
let @xmath296 be the leftmost fixed point of @xmath8 [ see fig .
for every @xmath297 , the sequence of iterates @xmath298}(x_0)\}$ ] is decreasing and diverges to minus infinity .
for each value of @xmath0 , let @xmath299 be the fixed point of @xmath8 to the right of @xmath300 at which @xmath301 .
a point @xmath302 is colored green if its trajectory diverges to minus infinity and it passes through the interval @xmath303 , and it is colored red if its trajectory diverges to minus infinity and it does not pass through the interval @xmath304 .
denote the collection of points @xmath302 that are colored green by @xmath22 $ ] , and the collection of points @xmath302 that are colored red by @xmath20 $ ] .
using the methods and techniques of @xcite , it can be shown that the collection of points @xmath302 which are common boundary points of @xmath22 $ ] and @xmath20 $ ] is a cantor set @xmath305 $ ] @xcite . in particular , the results of @xcite imply that for @xmath306 the point @xmath293 belongs to the invariant cantor set @xmath307 $ ] .
figure 18(b ) is a zoom of figure 18(a ) in the region of the saddle - node bifurcation . for values of @xmath50 , in the vicinity of @xmath308 ,
one notices a fractal alternation of red and green stripes .
the green and red stripe structure in fig .
18(b ) shares qualitative properties with the structure in fig .
all the analysis in sec .
[ sec2 ] can be adapted straightfowardly to fit this situation .
figure 19 shows how the chaotic saddle of the map @xmath8 varies with @xmath0 .
the chaotic saddle is generated numerically using the pim - triple method .
for an explanation of this method see nusse and yorke @xcite . using arguments similar to those in sec .
[ sec : scdim ] , we predict that changing the horizontal axis of fig . 19 from @xmath0 to @xmath51 makes the chaotic saddle asymptotically periodic .
numerical results confirming this are presented in fig .
for @xmath8 given by , we were able to find a parameter value @xmath309 where changing the horizontal axis of fig .
19 from @xmath0 to @xmath310 [ see fig . 20(b ) ] apparently makes the chaotic saddle asymptotically periodic [ with a different period than that of fig .
20(a ) ] . as in the case discussed in sec .
[ sec2 ] , past the saddle - node bifurcation of @xmath8 at @xmath5 , infinitely many other saddle - node bifurcations of periodic orbits take place on the invariant cantor set @xmath305 $ ] .
we believe that @xmath311 is an approximate value of @xmath0 where such a saddle - node of a periodic orbit takes place .
in this paper , we have investigated scaling properties of saddle - node bifurcations that occur on fractal basin boundaries .
such situations are known to be indeterminate in the sense that it is difficult to predict the eventual fate of an orbit that tracks the pre - bifurcation node attractor as the system parameter is swept through the bifurcation .
we have first analyzed the case of one - dimensional discrete maps . using the normal form of the saddle - node bifurcation and general properties of fractal basin boundaries , we established the following universal ( i.e. , model independent ) scaling results * scaling of the fractal basin boundary of the static ( i.e. , unswept ) system near the saddle - node bifurcation , * the scaling dependence of the orbit s final destination with the inverse of the sweeping rate , * the dependence of the time it takes for an attractor to capture a swept orbit with the -1/3 power of the sweeping rate , * scaling of the effect of noise on the final attractor capture probability with the 5/6 power of the sweeping rate .
all these results were demonstrated numerically for a one - dimensional map example .
following our one - dimensional investigations , we have explained and demonstrated numerically that these new results also apply to two - dimensional maps .
our numerical example was a two - dimensional map that results from a poincar section of the forced duffing oscillator . in the last section of the paper ,
we have discussed how the new results listed above apply to the case where a saddle - node bifurcation occurs on an invariant cantor set which is embedded in a basin of attraction , and we have supported our discussion by numerics .
thompson and m.s .
soliman , proc .
lond . a * 432 * , 101 ( 1991 ) ; m.s . soliman and j.m.t .
thompson , proc .
a * 438 * , 511 ( 1992 ) ; j.m.t .
thompson , physica d * 58 * , 260 ( 1992 ) ; m.s .
soliman , proc .
a * 451 * , 471 ( 1995 ) .
c. baesens , physica d * 53 * , 319 ( 1991 ) ; lecture notes in mathematics * 1493 * ( springer - verlag , berlin , 1991 ) , ed .
e. benot , 105 ; h.g .
davies and k. rangavalhula , proc .
a * 453 * , 2043 ( 1997 ) .
j. guckenheimer , inventiones math . *
39 * , 165 ( 1977 ) ; d. whitley , bull .
london math .
soc . * 15 * , 177 ( 1983 ) ; y.a .
kuznetsov `` elements of applied bifurcation theory '' , applied mathematical sciences * 112 * , ( springer - verlag 1995 ) .
j. kennedy and j.a .
yorke , physica d * 51 * , 213 ( 1991 ) ; h.e .
nusse and j.a .
yorke , physica d * 90 * , 242 ( 1996 ) ; h.e .
nusse and j.a .
yorke , science * 271 * , 1376 ( 1996 ) ; ergodic th . & dynam . syst .
* 17 * , 463 ( 1997 ) . for numerical purposes ,
we consider that a swept orbit has reached its final state if @xmath214 . changing
the factor of 10 to other values ranging from 1 to 30 , can slightly change the numerical result for the slope ( by as much as 3% ) . for every @xmath0 ( @xmath312 ) , write @xmath313 and @xmath314 .
the interval @xmath315 $ ] also contains a cantor set . by coloring this whole segment green ,
this information is lost .
therefore , the coloring scheme should be adapted if one wants to have the whole invariant cantor set represented for every @xmath0 .
for example , if a trajectory that diverges to minus infinity contains a point that is greater than @xmath316 then the initial point is colored green , if a trajectory that diverges to minus infinity contains a point that is greater than @xmath317 but not greater than @xmath316 then the initial point is colored yellow .
a point is colored red , if its trajectory diverges to minus infinity and does not have a point that is greater than @xmath317
. then the collection of boundary points ( a point @xmath7 is a boundary point if every open neighborhood of @xmath7 contains points of at least two different colors ) is a cantor set that contains the cantor set described above . | we analyze situations where a saddle - node bifurcation occurs on a fractal basin boundary . specifically , we are interested in what happens when a system parameter is slowly swept in time through the bifurcation .
such situations are known to be indeterminate in the sense that it is difficult to predict the eventual fate of an orbit that tracks the pre - bifurcation node attractor as the system parameter is swept through the bifurcation . in this paper
we investigate the scaling of ( 1 ) the fractal basin boundary of the static ( i.e. , unswept ) system near the saddle - node bifurcation , ( 2 ) the dependence of the orbit s final destination on the sweeping rate , ( 3 ) the dependence of the time it takes for an attractor to capture a swept orbit on the sweeping rate , and ( 4 ) the dependence of the final attractor capture probability on the noise level . with respect to noise
, our main result is that the effect of noise scales with the 5/6 power of the parameter drift rate . our approach is to first investigate all these issues using one - dimensional map models .
the simplification of treatment inherent in one dimension greatly facilitates analysis and numerical experiment , aiding us in obtaining the new results listed above .
following our one - dimensional investigations , we explain that these results can be applied to two - dimensional systems .
we show , through numerical experiments on a periodically forced second order differential equation example , that the scalings we have found also apply to systems that result in two dimensional maps . |
the high - precision combined hera data @xcite for proton s deeply inelastic scattering ( dis ) structure functions constrain parton density functions ( pdfs ) over a wide range of the kinematic variables .
these data have been used for determinations of the collinear pdfs and related studies at the lhc @xcite . on the other hand , qcd applications to multiple - scale scattering problems and complex final - state observables
require in general formulations of factorization @xcite which involve transverse - momentum dependent ( tmd ) , or unintegrated , parton density and parton decay functions @xcite .
tmd pdfs are necessary to describe appropriately nonperturbative physics and to control perturbative large logarithms to higher orders of perturbation theory .
the purpose of this work is to use the combined dis data on structure functions @xcite and charm production @xcite for determination of tmd pdfs .
a general program for tmd pdfs phenomenology has been proposed in @xcite .
our work has a more limited scope than this program as we limit ourselves to considering dis data in the small-@xmath0 kinematic region . on the other hand , from the point of view of tmd pdfs ,
this region is interesting because a well - defined form of tmd factorization holds at high energy @xcite , which has been applied to sum small-@xmath0 logarithmic corrections to dis to all orders in @xmath1 at leading and next - to - leading @xmath2 level @xcite .
furthermore , given the high precision of the combined data @xcite , this analysis provides a compelling test of the tmd approach and of the limitations of the logarithmic approximations used at small @xmath0 .
this is to be contrasted with earlier analyses @xcite based on older and much less precise structure function measurements .
the high - energy factorization @xcite expresses the heavy - quark leptoproduction cross section in terms of the tmd gluon density via well - prescribed , calculable perturbative coefficients .
this framework is extended to dis structure functions in @xcite .
phenomenological applications of this approach require matching of small-@xmath0 contributions with contributions from medium and large @xmath0 @xcite . to do this ,
in this work we further develop the parton branching monte carlo @xcite implementation of the ccfm evolution equation @xcite , which we include in the ` herafitter ` program @xcite .
the tmd gluon distribution at the initial scale @xmath3 of the evolution is determined from fits to dis data , including charm production .
we perform fits to measurements of the @xmath4 structure function @xcite in the range @xmath5 , @xmath6 gev@xmath7 and to measurements of the charm structure function @xmath8 @xcite in the range @xmath9 gev@xmath7 .
we obtain good fits to @xmath4 and @xmath8 , and we make a determination of the tmd gluon density ( as well as of the charm mass @xmath10 and of @xmath11 , or the strong coupling @xmath1 ) based on these . we find that the best fit to @xmath8 gives @xmath12 per degree of freedom @xmath13 , and the best fit to @xmath4 gives @xmath14 .
the method allows one to assign experimental and theoretical uncertainties to pdfs .
we give results for these different kinds of pdf uncertainties .
we also carry out an application of the tmd gluon density determined from hera data fits to lhc physics , by computing predictions for @xmath15-boson + jet production in proton - proton collisions .
the paper is organized as follows . in sec .
[ sec:2 ] we summarize the main elements of the approach based on high - energy factorization and evolution , and discuss a few details on its implementation . in sec .
[ sec:3 ] we describe the fits to charm and dis precision data .
we discuss @xmath4 , @xmath8 , the tmd gluon density determination and associated uncertainties . in sec .
[ sec:4 ] we illustrate the use of tmd gluon density at the lhc .
we give conclusions in sec .
[ sec:5 ] .
in the framework of high - energy factorization @xcite the deeply inelastic scattering cross section is written as a convolution in both longitudinal and transverse momenta of the tmd parton density function @xmath16 with off - shell partonic matrix elements , as follows @xmath17 here @xmath0 and @xmath18 denote the bjorken variable and photon virtuality , and the dis cross sections @xmath19 , with @xmath20 , are related to the customary dis structure functions @xmath4 and @xmath21 by @xcite @xmath22 where @xmath23 is the electromagnetic coupling . the factorization formula ( [ kt - factorisation ] )
allows one to resum logarithmically enhanced @xmath24 contributions to all orders in perturbation theory , both in the hard scattering coefficients and in the parton evolution , taking fully into account the dependence on the factorization scale @xmath25 and on the factorization scheme @xcite .
explicit evaluations have been carried out through next - to - leading logarithmic accuracy @xcite in dis at @xmath26 .
the physical origin of the logarithmically enhanced @xmath24 corrections at higher loops lies in the contribution from regions not ordered in initial - state transverse momenta to the qcd multi - parton matrix elements .
the perturbative higher - loop corrections are generally found to be large at small @xmath0 .
consider for instance the gluonic hard - scattering coefficient function @xmath27 for the dis structure function @xmath4 @xcite .
taking mellin moments with respect to @xmath0 , @xmath28 the perturbative expansion of @xmath29 for @xmath30 , resummed to all orders in @xmath31 via eq .
( [ kt - factorisation ] ) , is given at scale @xmath32 in the @xmath33 minimal subtraction scheme by @xcite @xmath34 where @xmath35 , @xmath36 , @xmath37 .
the @xmath38 poles @xmath39 , @xmath40 , correspond in @xmath0-space via eq .
( [ c2-mom - def ] ) to next - to - leading - logarithmic higher - loop corrections @xmath41 .
the first two terms in eq .
( [ c2-mom - res ] ) are the leading - order ( lo ) @xcite and next - to - leading - order ( nlo ) @xcite small-@xmath0 contributions to @xmath42 .
the next two terms are the three - loop and four - loop small-@xmath0 contributions .
the three - loop coefficient agrees with the complete next - to - next - to - leading - order ( nnlo ) calculation @xcite .
the three - loop and four - loop terms are logarithmically enhanced compared to lower orders . moreover , their numerical coefficients are significantly larger than the one - loop and two - loop ones .
analogous results were obtained in @xcite for the coefficient function @xmath43 of the longitudinal structure function , and confirmed through three loops by the nnlo calculation @xcite . given these results , there is little theoretical justification for treating small-@xmath0 dis by truncating the perturbative expansion to fixed nlo ( or nnlo ) level .
thus , although phenomenologically successful in giving very good fits to structure function data , fixed - order perturbative approaches are theoretically disfavored , and can not be expected to describe the physics of the scaling violation in the region of low values of @xmath0 , where transverse - momentum ordering does not apply .
the motivation of this work is to take a quantitative step toward going beyond fixed - order phenomenology , and confront our results with the high - precision combined data @xcite .
the approach of this work is based on * including the hard - scattering kernels @xmath44 of eq .
( [ kt - factorisation ] ) , whose @xmath45-dependence , once convoluted with the gluon unintegrated green s function @xcite , controls the all - order resummation of the higher - loop terms @xmath41 in the structure functions @xmath4 and @xmath21 ; * including evolution of the transverse momentum dependent gluon density @xmath46 by combining the resummation of small-@xmath0 logarithmic contributions @xcite with medium-@xmath0 and large-@xmath0 contributions to parton splitting @xcite .
this is done via a parton branching monte carlo implementation of the ccfm evolution equation @xcite , which we develop based on @xcite , and make available within the ` herafitter ` program @xcite . in the remainder of this section we briefly recall the basic elements of this approach and give a few technical details . we start in subsec .
[ subsec:2a ] by recalling the main features of tmd matrix elements and evolution . in subsec .
[ subsec:2b ] we include the unintegrated valence quark density according to the method @xcite . in subsec .
[ subsec:2c ] we discuss aspects of the numerical implementation .
we give comments on the general approach in subsec .
[ subsec:2-comment ] .
the dis transverse - momentum dependent partonic cross sections @xmath44 are evaluated in the second paper of @xcite for @xmath47 and in @xcite for @xmath48 , including the effects of finite quark masses .
the small-@xmath0 resummation for dis structure functions based on these results is carried out in @xcite .
for the off - shell process @xmath49 at different values of @xmath18 and quark masses .
we set @xmath50.__,title="fig : " ] for the off - shell process @xmath49 at different values of @xmath18 and quark masses .
we set @xmath50.__,title="fig : " ] for the off - shell process @xmath49 at different values of @xmath18 and quark masses .
we set @xmath50.__,title="fig : " ] let us describe the off - shell process @xmath49 in terms of lightcone momenta @xmath51 where , for any four - momentum , @xmath52 , with @xmath53 .
in the high - energy limit we have @xmath54 in terms of lightcone and transverse momentum components , we have @xmath55 , @xmath56 . in fig .
[ fig : ktmatrel ] we plot the partonic kernel @xmath57 , normalized to the value for @xmath58 , as a function of @xmath59 at fixed @xmath0 and @xmath18 , for various values of quark masses . as shown in @xcite , the large-@xmath45 tail of the kernels in fig .
[ fig : ktmatrel ] , once it is folded with unintegrated gluon distributions @xcite , is responsible for the logarithmically enhanced higher - loop contributions at high energy . for evolution of the tmd gluon density , we require that in the limit @xmath60 this evolves with the full bfkl anomalous dimension @xcite .
the ccfm evolution equation @xcite is an exclusive equation which satisfies this property ( see e.g. appendix c of @xcite , section 7 of the first paper in @xcite , section iii of @xcite ) and , in addition , includes finite-@xmath61 contributions to parton splitting . it incorporates soft gluon coherence for any value of @xmath0 .
the evolution equation for the tmd gluon density @xmath62 , depending on @xmath0 , @xmath63 and the evolution variable @xmath64 , reads @xmath65 the first term in the right hand side of eq .
( [ uglurepr1 ] ) is the contribution of the non - resolvable branchings between the starting scale @xmath3 and the evolution scale @xmath66 , and is given by @xmath67 where @xmath68 is the sudakov form factor , and @xmath69 is the starting distribution at scale @xmath3 . the integral term in the right hand side of eq .
( [ uglurepr1 ] ) gives the -dependent branchings in terms of the sudakov form factor @xmath68 and unintegrated splitting function @xmath70 .
the gluons average momentum does not change with @xmath66 .
the ccfm evolution equation can be written in a differential form , best suited for the backward evolution approach adopted in the monte carlo generator @xcite , as @xmath71 where the splitting variable @xmath72 is given by @xmath73 , @xmath74 , and @xmath75 is the azimuthal angle of @xmath76 .
the sudakov form factor @xmath77 is given by @xmath78 with @xmath79 .
the splitting function @xmath80 for branching @xmath81 is given by @xcite @xmath82 where @xmath83 is the non - sudakov form factor defined by @xmath84 quark masses are treated in the fixed flavor number scheme .
we include the two - loop running coupling @xmath1 , and we apply the kinematic consistency constraint @xcite in the @xmath85 splitting function @xcite , given by @xcite @xmath86 the evolution ( [ uglurepr1 ] ) of the tmd density implies that regions of transverse momenta below @xmath3 can be reached . in this region
the branching is performed purely by the sudakov form factor .
previous determinations of parton distributions based on the ccfm evolution have included only the gluon density @xcite . in this work
we include valence quarks using the method of @xcite .
we consider the branching evolution equation at the transverse - momentum dependent level according to @xmath87 where @xmath66 is the evolution scale .
the quark splitting function @xmath88 is given by @xmath89 with @xmath90 .
( [ integral]),([splitt ] ) the non - sudakov form factor is not included , unlike the ccfm kernel given in the appendix b of the first paper in @xcite , because we only associate this factor to @xmath91 terms .
the term @xmath92 in eq .
( [ integral ] ) is the contribution of the non - resolvable branchings between starting scale @xmath3 and evolution scale @xmath66 , given by @xmath93 where @xmath94 is the sudakov form factor , and the starting distributions at scale @xmath3 are parameterized , using the cteq 6.6 @xcite @xmath95 and @xmath96 valence quark distributions , as @xmath97 \;\ ; .
\label{gauss}\ ] ] in the numerical calculations that follow we will take @xmath98 . for every @xmath66
we ensure that the flavor sum rule is fulfilled .
ccfm evolution can not easily be written in an analytic closed form .
for this reason a monte carlo method is employed , based on @xcite .
the monte carlo solution is however time - consuming , and can not be used in a straightforward manner in a fit program . here
we proceed as follows .
first a kernel @xmath99 is determined from the monte carlo solution of the ccfm evolution equation , and then this is folded with the non - perturbative starting distribution @xmath100 , following the convolution method introduced in @xcite : @xmath101 the kernel @xmath102 incorporates all of the dynamics of the evolution , including sudakov form factors and splitting functions .
it is determined on a grid of @xmath103 bins in @xmath104 .
the binning in the grid is logarithmic , except for the longitudinal variable @xmath0 where we use 40 bins in logarithmic spacing below 0.1 , and 10 bins in linear spacing above 0.1 .
the calculation of the cross section according to eq .
( [ kt - factorisation ] ) involves a multidimensional monte carlo integration which is time consuming and suffers from numerical fluctuations .
this can not be employed directly in a fit procedure involving the calculation of numerical derivatives in the search for the minimum .
instead the following procedure is applied : @xmath105 here , first @xmath106 is calculated numerically with a monte carlo integration on a grid in @xmath0 for the values of @xmath18 used in the fit .
then the last step in eq.([final - convolution ] ) is performed with a fast numerical gauss integration , which can be used in standard fit procedures .
the approach described above , which we are going to confront in the next section with high - precision dis measurements , can be compared with other approaches to dis data in the literature .
on one hand , it can be contrasted with descriptions of data based on the dglap equation @xcite at fixed perturbative order , e.g. nlo or nnlo ( see for instance @xcite and references therein ) .
as recalled at the beginning of this section , these descriptions , however successful phenomenologically , have little theoretical justification at small @xmath0 , due to the structure of the perturbative expansion for @xmath107 .
the need to go beyond fixed - order truncations of perturbation theory leads us to employ a tmd formalism .
in particular , in the approach of this paper the transverse momentum dependence of the gluon density arises both from perturbative and from nonperturbative processes .
both the kernel and the initial condition of the evolution equation are @xmath45-dependent .
these different sources of @xmath45-dependence have distinct physical effects , for instance on the angular distributions of associated jet final states in dis , as analyzed in detail in @xcite .
this feature can be contrasted with tmd approaches focusing on nonperturbative @xmath45-dependence , see e.g. @xcite . on the other hand ,
the approach of this paper can be compared with approaches based on the bfkl equation @xcite .
recent fits to dis data have been performed in this context @xcite .
compared to these works , the main theoretical inputs in the present paper are * the use of tmd matrix elements which can be directly related with the resummation of dis coefficient functions as in eq .
( [ c2-mom - res ] ) , and * the use of the ccfm evolution for the tmd parton density rather than the bfkl evolution .
this results in distinctive properties due to soft gluon coherence of the final states contributing to dis @xcite .
a further , distinctive feature of the framework employed in this paper is that the gluon distribution obtained from dis fits can be directly used to make predictions for final states at the lhc , as we do for example in sec .
[ sec:4 ] for @xmath15-boson production associated with jets .
our approach relies on perturbative factorization theorems , which classify higher - order corrections according to the logarithmic hierarchy based on high @xmath18 and low @xmath0 . for this reason in the next section we will apply this approach to @xmath4 structure function measurements in a range @xmath108 , @xmath109 , where we choose @xmath110 gev@xmath7 , @xmath111 . for asymptotically small @xmath0
one expects the operator product expansion to break down and dis to become dominated by nonperturbative physics .
thus the low-@xmath18 region could require methods beyond the ones applied in this work , see e.g. @xcite .
on the other hand , the evolution approach used in this work may be supplemented with nonlinear corrections @xcite to describe aspects of parton saturation @xcite .
it will therefore be of interest to investigate the extension of the work presented in this paper to low @xmath18 .
the inclusion of data at higher @xmath18 , relaxing the low-@xmath0 kinematic cut , will constitute a further development of the tmd formulation .
high-@xmath0 theoretical issues , including tmd quark distributions , are discussed e.g. in @xcite .
analyses of dis data over the whole available range in @xmath0 and @xmath18 in terms of tmd pdfs will be relevant for the calculational program @xcite of off - shell hard cross sections .
the fit to the hera structure function measurements is performed by applying the ` herafitter ` program @xcite to determine the parameters of the starting distribution @xmath112 at the starting scale @xmath3 .
we perform fits by using two possible parameterizations of @xmath112 : the five - parameter form @xmath113 \;\ ; , \label{a0 - 5par}\end{aligned}\ ] ] and the three - parameter form @xmath114 \;\ ; .
\label{a0 - 3par}\end{aligned}\ ] ] as in eq .
( [ gauss ] ) , we take @xmath98 .
the parameters @xmath115 ( resp .
@xmath116 ) in eq . ( [ a0 - 5par ] ) ( resp . eq . ( [ a0 - 3par ] ) )
are determined by fitting the high - precision structure function measurements @xcite in the range @xmath5 and @xmath6 gev@xmath7 , and the charm production measurements @xcite , which are in the range @xmath9 gev@xmath7 .
the results presented here are obtained with the ` herafitter ` package by treating the correlated systematic uncertainties separately from the uncorrelated statistical and systematic uncertainties .
.__the values of @xmath117 corresponding to the best fit for charm structure function @xmath8 , for inclusive structure function @xmath4 , and for the combination of @xmath118 and @xmath4 . _
_ [ cols="^,^,^,^",options="header " , ] we fit the charm leptoproduction data @xcite and the inclusive structure function data @xcite based on high - energy factorization and ccfm evolution as described in sec . [ sec:2 ] .
in particular , we also include two - loop running coupling , gluon splitting and consistency constraint as in subsec .
[ subsec:2a ] , and , in addition to the gluon - induced process @xmath119 , the contribution from valence quarks is included via @xmath120 as in subsec .
[ subsec:2b ] by using a ccfm evolution of valence quarks @xcite . to obtain a reasonable fit to structure function data , we vary the starting scale @xmath3 , the qcd scale @xmath11 and the charm quark mass @xmath121 . the results in table [ tablechi ] and in fig . [ fig : f2datafit ] are obtained for @xmath122 gev , @xmath123 gev at @xmath124 , @xmath125 gev .
table [ tablechi ] reports the values of @xmath12 per degree of freedom for the best fit to the charm structure function @xmath8 @xcite in the full data range @xmath9 gev@xmath7 , and to the inclusive structure function @xmath4 @xcite in the data range @xmath5 , @xmath6 gev@xmath7 and for a combination of both , in the cases of the three - parameter fit ( [ a0 - 3par ] ) and five - parameter fit ( [ a0 - 5par ] ) .
the best - fit @xmath117 is below 1 for the charm structure function , while it is around 1.18 for the inclusive structure function .
this is in accord with the expectation that charm production is dominated by the gluon distribution , while the inclusive structure function receives significant contributions from quark channels , for which an improved treatment at unintegrated level is needed .
also , the gluon density determined from charm measurements turns out to be steeper at small @xmath0 than that determined from the inclusive structure function .
the resulting tension between the two fits may be related to the the fact that the starting scale @xmath126 is not far from the charm threshold mass @xmath127 .
the @xmath117 for the fit to both @xmath8 and @xmath4 is 1.43 for the three - parameter fit , and is not significantly changed by using the five - parameter form .
[ fig : f2datafit ] shows the description of the charm leptoproduction measurements @xcite and inclusive structure function measurements @xcite , by the individual fits to @xmath8 and @xmath4 and by the combined fit . plotted
are the reduced cross sections defined in @xcite . in fig .
[ fig : charmscan ] we show the results of a scan in the charm mass , by plotting the values of @xmath12 for the fit to charm and inclusive measurements as a function of the charm quark mass .
the minimum is reached for @xmath128 gev . for the fit to dis high - precision data including charm leptoproduction @xcite versus the charm quark mass . _ ] here we present two sets of unintegrated pdfs determined from the fits to high - precision dis measurements described in the previous subsection : jh-2013-set1 is determined from the fit to inclusive @xmath4 data only ; jh-2013-set2 is determined from the fit to both @xmath8 and @xmath4 data .
the unintegrated tmd gluon density is shown in figs .
[ fig : updf ] and [ fig : updfbis ] at different evolution scales , versus the longitudinal momentum fraction @xmath0 and versus the transverse momentum @xmath63 .
the results are compared with the older parton distribution set a0 @xcite , which did not use the precision data and did not include two - loop running coupling , kinematic consistency constraint , nonsingular terms in the gluon splitting function , and unintegrated valence quark density .
-boson mass , @xmath129 : ( top ) as a function of @xmath0 for different values of @xmath130 ; ( bottom ) as a function of @xmath130 for different values of @xmath0 .
the results are compared with set a0 @xcite . _ ] we see from the evolution to scale @xmath131 equal to the @xmath132-boson mass in fig . [
fig : updf ] that the fit to the dis high - precision measurements gives significant differences in the tmd gluon density compared to earlier determinations , especially in the region of medium to low @xmath63 . for the lower @xmath131 scale in fig .
[ fig : updfbis ] the differences are less pronounced but still important especially for small values of @xmath0 .
gev@xmath7 : ( top ) as a function of @xmath0 for different values of @xmath130 ; ( bottom ) as a function of @xmath130 for different values of @xmath0 .
the results are compared with set a0 @xcite . _ ] in this subsection we consider separately experimental and theoretical uncertainties of the tmd parton distributions . for different values of transverse momentum at @xmath129 .
the yellow band gives the uncertainty from the factorization scale variation ; the curves indicate the uncertainties from the other sources .
_ , title="fig : " ] for different values of transverse momentum at @xmath129 .
the yellow band gives the uncertainty from the factorization scale variation ; the curves indicate the uncertainties from the other sources .
_ , title="fig : " ] experimental uncertainties are obtained within the ` herafitter ` package from a variation of the individual parameter uncertainties , following the procedure described in @xcite applying @xmath133 .
these result in 10 to 20 percent gluon uncertainty for medium and large @xmath0 .
the experimental uncertainties on the gluon at small @xmath0 are small ( much smaller than those obtained in standard fits based on integrated pdfs ) , since only the gluon density is fitted .
the experimental uncertainties are shown by the dot - dashed red curves in fig .
[ fig : updf - uncertainty ] as a function of @xmath0 for different values of transverse momentum at the evolution scale @xmath131 equal to the @xmath132-boson mass .
similarly in fig .
[ fig : updf - uncertainty - bis ] for lower evolution scale @xmath134 gev@xmath7 . for different values of transverse momentum at @xmath134 gev@xmath7 .
the yellow band gives the uncertainty from the factorization scale variation ; the curves indicate the uncertainties from the other sources .
_ , title="fig : " ] for different values of transverse momentum at @xmath134 gev@xmath7 .
the yellow band gives the uncertainty from the factorization scale variation ; the curves indicate the uncertainties from the other sources .
_ , title="fig : " ] next we consider theoretical uncertainties .
the first such kind of uncertainty is the dependence on the starting scale @xmath3 for gluon density evolution . in figs . [
fig : updf - uncertainty ] and [ fig : updf - uncertainty - bis ] the dotted blue curves show the effect on the gluon distribution from variation in the starting scale @xmath135 .
these uncertainties are small at small @xmath0 , while they become very large at large @xmath0 because in this region , since we fit @xmath4 in the range @xmath5 and @xmath6 gev@xmath7 , there is little constraint from data .
we also consider theoretical uncertainties on the tmd gluon density from variation of the factorization scale and renormalization scale .
this approach is different from that usually followed in determinations of ordinary , collinear pdfs from fixed - order perturbative treatments @xcite . in this case , no uncertainty on the pdfs is considered from scale variation .
only when computing predictions for any specific observable the theoretical uncertainty on the predictions is estimated by scale variation . in our approach we are interested to estimate the uncertainty from varying scales in the theoretical calculation used to determine the pdf . in figs .
[ fig : updf - uncertainty ] and [ fig : updf - uncertainty - bis ] the renormalization scale ( blue dashed curves ) and in the factorization scale ( yellow band ) are varied by a factor of 2 . for a cross check with the integrated pdfs we now compute the integral over transverse momenta of the tmd parton distributions . in fig .
[ fig : integrated ] we plot the results for gluon and valence quark distributions , obtained from the set jh-2013-set1 of this paper ( and also , for comparison , from the gluon in the older set a0 @xcite ) , at two different evolution scales .
for comparison we plot the ordinary , integrated distributions obtained from the nlo - dglap cteq 6.6 @xcite fit .
we observe good agreement for the integral of valence quark distributions at low scales while differences arise from the different evolution at larger scales . for the gluon case , the difference between the integral of tmd and cteq reflects the shuffling of flavor singlet contributions between sea quark and gluon in the two formalisms .
gev@xmath7 ; ( bottom ) @xmath136 .
_ , title="fig : " ] gev@xmath7 ; ( bottom ) @xmath136 .
_ , title="fig : " ] gev@xmath7 ; ( bottom ) @xmath136 .
_ , title="fig : " ] gev@xmath7 ; ( bottom ) @xmath136 .
_ , title="fig : " ] we conclude this section by noting that the ccfm evolution kernel can be approximated to one loop by using collinear ordering @xcite .
this constitutes the dglap limit of the evolution equation .
if we perform fits to the high - precision @xmath4 data by using eqs .
( [ kt - factorisation]),([uglurepr1 ] ) in the one - loop approximation mode we find that this approximation is unable to give a good fit based on the tmd gluon only , @xmath137 . we interpret this as a check on the consistency of the physical picture , signaling the need for introducing quark - initiated processes in the collinear framework .
the tmd parton distributions determined in sec . [ sec:3 ] from fits to the high - precision dis data can be used to make predictions for hadron - hadron collider processes .
an example is the drell - yan vector boson production .
we here consider @xmath138-boson production in association with jets .
this process is important both for standard model physics and for new physics searches at the lhc . in particular
it is relevant to studies of parton distribution functions and of monte carlo event generators @xcite , including signals of multi - parton interactions , for which @xmath139 jets is a classic channel @xcite .
jets production using the unintegrated tmd gluon density jh-2013-set2 : ( left ) inclusive jet multiplicity ; ( right ) leading jet transverse momentum spectrum .
the experimental data are from @xcite .
the yellow band is the experimental uncertainty .
the blue band is the theory uncertainty.__,title="fig : " ] jets production using the unintegrated tmd gluon density jh-2013-set2 : ( left ) inclusive jet multiplicity ; ( right ) leading jet transverse momentum spectrum .
the experimental data are from @xcite .
the yellow band is the experimental uncertainty .
the blue band is the theory uncertainty.__,title="fig : " ] to compute predictions for @xmath138-boson + jets final states @xcite , we use the ccfm gluon and valence quark distributions determined in the previous section , convoluted with high - energy matrix elements @xcite with off - shell partons @xcite for weak boson production .
we present results for the inclusive jet multiplicity distribution and leading jet transverse momentum spectrum .
the results ( obtained with the rivet - package @xcite ) are shown in fig .
[ fig : wjets ] along with the atlas measurements @xcite .
the solid red curves in fig .
[ fig : wjets ] are the predictions from jh-2013-set2 , with the blue band corresponding to the pdf uncertainty . both the jet multiplicity and the transverse momentum are reasonably well described by the predictions .
further discussion on @xmath15 + jets will be given in @xcite .
the production of final states with @xmath15 boson and multiple jets at the lhc receives sizeable contributions from large separations in rapidity between final - state particles .
however , the cross sections computed in fig . [ fig : wjets ] are not dominated by very small values of @xmath0 . as a result ,
the uncertainty band due to the uncertainties in the tmd pdfs is significant .
a comparison with nlo - matched results and corresponding uncertainties is presented in @xcite .
it is conceivable that combining @xmath140 measurements on vector boson production with the dis measurements analyzed in this paper may help to constrain tmd pdfs especially at medium to large values of @xmath0 .
in this work we have performed the first determination of the tmd gluon density function from high - precision dis measurements , including experimental and theoretical uncertainties .
we have presented fits , based on qcd high - energy factorization and ccfm evolution , to hera charm - quark leptoproduction data for the structure function @xmath8 @xcite in the range @xmath9 gev@xmath7 , and to hera @xmath4 structure function data @xcite in the range @xmath5 and @xmath6 gev@xmath7 . in this approach
the charm structure function can be regarded as a physical probe of the unintegrated tmd gluon density .
we fit the combined hera charm - quark data @xcite over the whole kinematic range of the measurement , and obtain that the best fit gives @xmath12 per degree of freedom @xmath13 .
the inclusive @xmath4 structure function involves both gluon - density and quark - density channels .
we fit the combined hera @xmath4 data @xcite in the kinematic range @xmath5 , @xmath6 gev@xmath7 , and obtain that the best fit gives @xmath14 . despite the restricted kinematic range ,
the great precision of the data provides a highly nontrivial test of the approach .
we find a good fit to both charm - quark and inclusive data .
based on this , we make a determination of the tmd gluon density ( as well as of the qcd scale @xmath11 and the charm mass @xmath121 ) and present new unintegrated pdf sets , jh-2013 . as a result of the high - precision data ,
the jh-2013 distributions differ significantly from earlier sets
. we also present experimental and theoretical uncertainties associated with the tmd pdfs .
we compute predictions based on the tmd pdfs for @xmath15-boson plus jets production at the lhc , and find that the results compare well with the measurements @xcite of jet multiplicities and transverse momentum spectra within the pdf uncertainties .
the approach of this work is based on the use of transverse momentum dependent matrix elements and evolution .
both the kernel and the initial condition of the evolution equation are @xmath45-dependent .
the transverse momentum dependence of the gluon density arises from both perturbative and nonperturbative processes .
the physical picture of dis scaling violation underlying this approach differs from that of finite - order perturbative qcd fits , e.g. at the nlo level , because it takes into account corrections to the collinear ordering in the initial state evolution to all orders in the qcd coupling @xmath1 . on the other hand
, it also differs from bfkl evolution because it takes into account , for any @xmath0 , color coherence associated with soft multi - gluon emission . in this paper
we have developed a parton branching monte carlo implementation of the ccfm evolution equation and we have included it in the ` herafitter ` program @xcite .
the choice of the kinematic range for the @xmath4 data considered in this paper is motivated by the fact that our approach relies on perturbative factorization theorems , which classify higher - order corrections according to logarithmic hierarchy based on high @xmath18 and low @xmath0 .
however , we note that the choice made in this paper is conservative , and the physical picture lends itself to extensions to lower @xmath18 and higher @xmath0 . on one hand
, this picture goes beyond dglap by including non - collinear emissions to all orders in @xmath1 . on the other hand
, it goes beyond bfkl by including large-@xmath0 terms according to the ccfm prescription .
furthermore , the study of @xmath15-boson + jets made in this paper suggests that @xmath140 measurements of vector boson production at the lhc may be used to extend experimental investigations of tmd parton density functions .
many thanks to the herafitter team for invaluable advice .
we acknowledge useful discussions and comments on the paper draft from m. cooper - sarkar , s. glazov , r. placakyte , s. prestel and v. radescu .
fh thanks the terascale physics helmholtz alliance and desy for support and hospitality .
g. altarelli , r. ball and s. forte , nucl .
b*799 * ( 2008 ) 199 ; m. ciafaloni , d. colferai , g.p . salam and a. stasto , jhep * 0708 * ( 2007 ) 046 ; r.s .
thorne , phys .
d*60 * ( 1999 ) 054031 ; r.k .
ellis , f. hautmann and b.r .
webber , phys .
* b * 348 ( 1995 ) 582 .
`` '' ( 2012 ) , https://www.herafitter.org/ ; f. aaron _ et al . _ ,
eur . phys .
j. * c64 * ( 2009 ) 561 , http://arxiv.org/abs/0904.3513[arxiv:0904.3513 [ hep - ex ] ] ; f. james and m. roos , comput .
commun .
* 10 * ( 1975 ) 343 .
j. pumplin , d. stump and w .- k .
tung , phys .
rev . * d65 * ( 2001 ) 014011 ; j. pumplin , d. stump , r. brock , d. casey , j. huston , j. kalk , h .-
lai and w .- k .
tung , http://dx.doi.org/10.1103/physrevd.65.014013[phys .
* d65 * ( 2001 ) 014013 ] ; http://dx.doi.org/10.1103/physrevd.65.014012[phys .
* d65 * ( 2001 ) 014012 ] . | the combined measurements of proton s structure functions in deeply inelastic scattering at the hera collider provide high - precision data capable of constraining parton density functions over a wide range of the kinematic variables .
we perform fits to these data using transverse momentum dependent qcd factorization and ccfm evolution .
the results of the fits to precision measurements are used to make a determination of the nonperturbative transverse momentum dependent gluon density function , including experimental and theoretical uncertainties .
we present an application of this density function to vector boson + jet production processes at the lhc . |
while the sun appears to be a single star , binary and multiple stars are commonplace .
the comprehensive survey of all solar - type ( f6 to k3 ) main - sequence stars within 25 pc by raghavan et al .
( 2010 ) found @xmath3 56% of the primary stars to have no confirmed companions , @xmath3 33% to have a binary companion , @xmath3 8% to be in triple systems , and @xmath3 1% to be in even higher - order systems .
when counting the total number of stars involved , then , this means that at least @xmath3 2/3 of all stars are members of binary or multiple star systems .
in fact , increasingly higher binary star frequencies are found in increasingly younger stellar and protostellar populations ( e.g. , class o protostars ; chen et al .
2013 ) , implying that essentially all stars could be born in multiple systems , a fraction of which eventually decay through orbital evolution and close encounters , leading to the ejection of single stars , and leaving behind stable binary and hierarchical multiple systems ( reipurth et al .
the leading mechanism for the formation of binary and multiple stars is fragmentation ( reipurth et al .
2014 ) , where a dense molecular cloud core undergoes dynamic collapse and rapidly breaks up into two or more protostellar objects ( boss & bodenheimer 1979 ) .
brown dwarf binaries with total system masses as small as @xmath4 have also been detected ( choi et al .
2013 ) , implying that binary fragmentation continues to operate in this sub - stellar mass regime as well . observations of zeeman splitting in molecular clouds have shown that magnetic fields can often be detected , with field strengths significant enough to affect the outcome of the fragmentation process ( e.g. , girart et al . 2013 ) .
crutcher ( 2012 ) concluded that while magnetic field strengths generally are more important than turbulence for supporting molecular clouds against collapse , these fields are not strong enough to overcome gravity , at least not for clouds with column densities higher than @xmath5 @xmath6 : such relatively dense clouds appear to be magnetically supercritical , and hence should be either contracting or susceptible to contraction . as a result ,
3d mhd calculations of cloud collapse have become commonplace in recent years ( reipurth et al .
2014 ) , supplanting the 3d hydrodynamics ( hd ) that had previously been considered state - of - the - art ( e.g. , bonnell et al .
hennebelle et al .
( 2011 ) used an amr mhd code to follow the collapse of a relatively massive , 100 @xmath0 cloud , finding that the initial magnetic field strength reduced the amount of fragmentation by as much as a factor of two , compared to a nonmagnetic cloud collapse .
clearly magnetic fields need to be considered when assessing the efficiency of the fragmentation mechanism for forming binary and multiple star systems .
observations of starless , pre - collapse molecular cloud cores have shown that dense cloud cores are centrally - condensed ( e.g. , ward - thompson , motte , & andr 1999 ; schnee et al . 2010 ; ruoskanen et al . 2011 ; butler & tan 2012 ) , highly non - spherical , and often turbulent .
velocity gradients indicative of overall rotation in dense cloud cores lead to estimates of the ratio of rotational to gravitational energy ranging from @xmath7 to @xmath8 ( caselli et al .
2002 ; harsono et al .
prolate and oblate shapes have been inferred from observations of suspected pre - collapse molecular cloud cores ( e.g. , jones , basu , & dubinski 2001 ; curry & stahler 2001 ; cai & taam 2010 ; gholipour & nejad - asghar 2013 ; lomax , whitworth , & cartwright 2013 ) . while boss & keiser ( 2013 ; hereafter bk13 ) used enzo to study the collapse of initially uniform density , rotating , spherical clouds , here we consider the collapse of more realistic , centrally - condensed , spheroidal molecular cloud cores , namely the initially prolate and oblate cloud cores considered by boss ( 2009 ) , who used a pseudo - mhd code to model magnetic field effects and ambipolar diffusion ( e.g. , boss 2005 , 2007 ) .
we intend in part to see if these pseudo - mhd calculations produced results similar to those obtained with a true mhd code like enzo .
machida , inutsuka , & matsumoro ( 2014 ) used an mhd code to study the effects of starting from either initially uniform density or centrally - condensed cloud cores , and found that the initial density profile had a significant effect on the formation of protostellar disks . here
we focus on the effects of the initial magnetic field strength on fragmentation in initially centrally - condensed cloud cores .
we seek to return to the questions of the extent to which the fragmentation mechanism occurs in dense molecular cloud cores when magnetic fields are included , and when it does occur , what this implies for the formation of single , binary , and multiple protostellar systems .
while bk13 used the enzo2.0 code , here we have employed the updated enzo2.2 version , which includes several code improvements as well as bug fixes ( e.g. , bk13 found a bug in an early version of enzo2.0 dealing with the gas mean molecular weight , which was corrected in later versions ) .
enzo offers several choices for the hydrodynamics scheme ( collins et al .
2010 ; bryan et al .
2014 ) , which are designed to conserve mass and energy .
following bk13 , we used the runge - kutta third - order - based muscl ( monotone upstream - centered schemes for conservation laws ) algorithm ( hydromethod = 4 ) , which is based on the godunov ( 1959 ) shock - handling hd method , along with the harten - lax - van leer ( hll ) riemann solver .
ideal mhd was performed using the dedner et al .
( 2002 ) divergence cleaning method ( wang et al . 2008 ) .
the divergence of the magnetic field was monitored throughout the calculations , in order to ensure that this quantity , which should remain zero for ideal mhd , remained small throughout the evolutions .
more details about these and other numerical matters may be found in bk13 .
the standard mhd and hd models were initialized in enzo2.2 on a 3d cartesian top grid with 64 grid points in each direction ; several models used 128 grid points in each direction for comparison testing .
the standard models had a maximum of 6 levels of refinement that could be implemented as needed during the collapses , each level with a factor of 2 refinement over the previous level .
hence the maximum possible effective grid resolution for the standard models was @xmath9 times higher than the initial top grid resolution of @xmath10 , i.e. , @xmath11 .
several comparison models explored the effects of changing the maximum number of levels of refinement from 6 to as high as 10 and to as low as 2 levels .
grid refinement was performed automatically whenever necessary to ensure that the jeans length constraint ( e.g. , truelove et al .
1997 ) was satisfied by a factor of 8 or more .
reflecting boundary conditions were applied on each face of the top grid s cubic box , after tests with the other boundary conditions possible with enzo showed that reflecting boundary conditions did the best job of conserving mass and angular momentum ( to within 1% and 6% , respectively ) during non - magnetic collapse to central densities of @xmath12 g @xmath13 . changing the top grid resolution from @xmath14 to @xmath10 , on the other hand , had no effect on the conservation of total mass and angular momentum .
the maximum number of green s functions used to calculate the gravitational potential was 10 , with 10 iterations performed on the gravitational potential at each step .
the time step used was 0.3 of the limiting courant time step .
the second - order - accurate interpolationmethod = 1 option was adopted to maximize mass and angular momentum conservation .
the yt astrophysical analysis and visualization toolkit ( turk et al . 2011 ) was used to analyze the results .
collapsing molecular cloud cores remain isothermal at @xmath3 10 k until the central regions reach densities of @xmath15 g @xmath13 ( e.g. , boss 1986 ) , at which point the centers become optically thick to infrared radiation , and compressional heating begins to raise the clouds temperatures .
while radiative transfer is necessary thereafter to follow the subsequent evolution properly ( e.g. , boss 1986 ; boss et al . 2000 ) , here we explore the nonisothermal regime by using the `` barotropic '' equation employed by , e.g. , price & bate ( 2007 ) , brzle et al .
( 2011 ) , bk13 , and many others .
the barotropic pressure @xmath16 depends on the gas density @xmath17 as @xmath18 , where the polytropic exponent @xmath19 equals 1 for @xmath20 g @xmath13 and 7/5 for @xmath21 g @xmath13 .
the factor @xmath22 , where @xmath23 is the isothermal sound speed ( 0.2 km s@xmath24 ) , below the critical density of @xmath25 g @xmath13 , and @xmath26 for @xmath27 .
this results in continuity of the pressure across the critical density .
the barotropic approximation is a variation of the `` adiabatic '' approximation for the nonisothermal regime introduced by boss ( 1981 ) , who noted that such calculations could be scaled to clouds of arbitrary mass .
the adiabatic approximation assumes @xmath28 with @xmath29 , and since the boss ( 1981 ) models all started with very high densities , @xmath30 g @xmath13 , these adiabatic approximation models were effectively the same as barotropic approximation models .
future calculations will include radiative transfer effects in the flux - limited diffusion approximation , which has been implemented in enzo ( reynolds et al .
2009 ) , but which will increase the computational burden substantially , compared to the barotropic approximation .
radiative transfer was included in 3d hd orion amr calculations by krumholz et al .
( 2007 ) and offner et .
al ( 2009 ) , in 3d mhd ramses amr calculations by commercon et al .
( 2010 ) , in 3d hd sph calculations by whitehouse & bate ( 2006 ) and bate ( 2009 , 2011 , 2012 ) , and in 3d mhd sph calculations by price & bate ( 2009 ) , all of whom found that radiative transfer had important effects not captured by the barotropic approximation . rather than attempt to follow the nonisothermal collapse phase through to the formation of the first and second protostellar cores ( e.g. , bate , tricco , & price 2014 ) , in some models we employ sink particles to represent the highest density clumps that form during protostellar collapse and fragmentation .
sink particles are the moving versions of the stationary sink cells first introduced by boss & black ( 1982 ) and used to study the accretion of infalling gas by protostars .
sink particles have been developed for use in enzo by wang et al .
( 2010 ) and used by zhao & li ( 2013 ) .
sink particles have also been developed for use in smoothed particle hydrodynamics ( sph ) calculations ( bate , bonnell , & price 1995 ; hubber , walch , & whitworth 2013 ) , in the athena code ( e.g. , gong & ostriker 2013 ) , in gadget-2 ( e.g. , brzle et al . 2011 ; klapp et al .
2014 ) , and elsewhere .
recently machida , inutsuka , & matsumoto ( 2014 ) found that sink cells need to be less than 1 au in size , and to have threshold densities greater than @xmath31 @xmath13 , in order to properly study the formation of disks during the main accretion phase of protostellar collapse .
we used the enzo sink particle coding described in some detail by wang et al .
sink particles are created in grid cells that have already been refined to the maximum extent permitted by the specified number of levels of grid refinement , but where the gas density still exceeds that consistent with the jeans length criterion for avoiding spurious fragmentation ( truelove et al .
1997 ; boss et al .
the sink particles thereafter accrete gas from their host cells at the modified bondi - hoyle accretion rate proposed by ruffert ( 1994 ) .
two parameters control the conditions under which sink particles can be merged together : the merging mass ( sinkmergemass ) and the merging distance ( sinkmergedistance ) .
the former of these two parameters is used to divide the sink particles into either large or small particles .
particles with less mass than sinkmergemass are first subjected to being combined with any large particles that are located within the sinkmergedistance .
any surviving small particles after this first step are then merged with any other small particles within the sinkmergedistance .
the merging process is performed in such a way as to ensure conservation of mass and linear momentum .
note that the magnetic field is not accreted by the sink particles , which can be considered as a crude means of depicting decoupling of the magnetic field from high density gas ( zhao & li 2013 ) .
wang et al . (
2010 ) and zhao & li ( 2013 ) discuss the extent to which amr code results depend on the choices made for the sinkmergemass and sinkmergedistance parameters ; we have performed a similar investigation , with relatively large changes in these two sink particle parameters ( see section 4.4 ) .
ambipolar diffusion has been incorporated into the flash amr code by duffin & pudritz ( 2008 ) , the zeus 3d code ( mac low et al .
1995 ) , and into other codes as well ( e.g. , choi et al .
ambipolar diffusion was included in early versions of enzo , but it is not supported in the most recent versions ( bryan et al .
collins et al .
( 2011 ) argued that ambipolar diffusion was unimportant for their enzo calculations of the collapse of cloud cores because the initial magnetic field strengths were assumed to be relatively weak ( i.e. , @xmath32 g for @xmath33 g @xmath13 ) .
boss ( 1997 ) derived an approximation for ambipolar diffusion in a 3d pseudo - mhd code , based on true mhd collapse models , and used this formalism in subsequent papers in the series ( e.g. , boss 2005 , 2007 , 2009 ) .
while one of our goals is to compare our enzo results with those of boss ( 2009 ) , we do not attempt to include ambipolar diffusion in the present models , for several reasons .
first , current versions of enzo do not support ambipolar diffusion .
second , the approximation derived by boss ( 1997 ) was essentially one of lowering the magnetic field strength in a linear manner until dynamic collapse began , after which collapse occurs so rapidly that the field is essentially frozen - in to the gas , i.e. , ideal mhd dominates .
third , crutcher ( 2012 ) has concluded on observational grounds that ambipolar diffusion is not important , but that magnetic reconnection is more likely to be important for solving the magnetic flux problem of star formation ( lazarian , esquivel , & crutcher 2012 ) .
magnetic reconnection is effective on large scales in turbulent clouds ( e.g. , santos - lima , de gouveia dal pino , & lazarian 2013 ) .
we leave for future work the consideration of initially turbulent clouds , which are likely to experience significant magnetic reconnection .
we study the collapse of initially prolate and oblate molecular cloud cores , the same initial conditions studied by boss ( 2009 ) . boss ( 2009 )
used a 3d hd code with radiative transfer in the eddington approximation , but only a pseudo - mhd handling of magnetic fields . here
we employ a true mhd code , enzo2.2 , but resort to a simple barotropic equation of state to represent the effects of radiative transfer in optically thick regions on the gas pressure . as in boss ( 2009 )
, the initial cloud cores are centrally condensed , with gaussian radial density profiles ( boss 1997 ) of the form @xmath34 where @xmath35 g @xmath13 is the initial central density .
the initial central density is roughly a factor of 20 times higher than that at the edge of the cloud for all of the models .
the clouds are initially either prolate or oblate , with axis ratios of 2 to 1 .
the prolate clouds have @xmath36 and @xmath37 , where @xmath38 is the cloud radius , while the oblate clouds have @xmath39 and @xmath40 .
the cloud radius is @xmath41 cm = 0.032 pc ; the top level amr grid is a cube @xmath42 cm on a side .
random numbers in the range [ 0,1 ] are used to add noise to these initial density distributions by multiplying @xmath43 from the above equation by the factor @xmath44 $ ] .
this form of density perturbation was used , instead of a velocity perturbation , in order to make the initial conditions as close as possible to those of boss ( 2009 ) , who used this same density perturbation with no velocity perturbation .
the cloud masses are 1.73 @xmath0 and 2.73 @xmath0 , respectively , for the prolate and oblate clouds , slightly higher than for the corresponding clouds in boss ( 2009 ) , 1.5 @xmath0 and 2.1 @xmath0 , respectively , because of the cubical volume of the enzo computational grid compared to the spherical volume of the boss ( 2009 ) spherical coordinate code grid . with an initial temperature of 10 k ,
the initial ratio of thermal to gravitational energy for the boss ( 2009 ) models is @xmath45 for the prolate clouds and 0.30 for the oblate clouds ; such clouds are unstable to collapse in the absence of rotation or magnetic fields .
the initial magnetic field strengths assumed for the cloud cores range from 6.3 @xmath1 g to 100 @xmath1 g .
estimates of the line - of - sight component of the magnetic field in dense cloud cores with column densities appropriate for these models ( @xmath46 @xmath6 ) are on the order of @xmath47 g to @xmath48 g ( figure 7 in crutcher 2012 ) .
the critical mass @xmath49 is given by @xmath50 , where @xmath51 is the magnetic flux ( nakano & nakamura 1978 ; crutcher 2012 ) and @xmath52 is the gravitational constant .
magnetically subcritical clouds have masses less than @xmath49 , while magnetically supercritical clouds have masses greater than @xmath49 and should be able to contract and collapse .
the critical mass to flux ratio is then defined by @xmath53 . for the gaussian profile prolate clouds studied here ,
the initial mass to flux ratios range from @xmath3 28 to 1.8 @xmath54 for the initial field strengths of 6.3 to 100 @xmath1 g , respectively , i.e. , from clouds that are strongly to marginally supercritical , respectively .
the initial mass to flux ratios are a factor of @xmath55 times higher for the oblate clouds , i.e. , ranging from 44.9 to 11.4 @xmath54 for 6.3 to 25 @xmath1 g , respectively . boss ( 2009 )
started all models with a magnetic field strength of 200 @xmath1 g , resulting in initially magnetically stable clouds with subcritical mass to flux ratios .
the approximate inclusion of ambipolar diffusion allowed the boss ( 2009 ) clouds to collapse eventually .
solid body rotation is assumed , with the angular velocity about the @xmath56 axis taken to be @xmath57 = @xmath58 , @xmath59 , @xmath60 , or @xmath61 rad s@xmath24 .
these choices of @xmath57 result in initial ratios of rotational to gravitational energy ( @xmath62 ) ranging from @xmath7 to @xmath8 , for both the prolate and oblate clouds .
solid body rotation is a reasonable approximation to continue to employ for the initial conditions for dense cloud cores , as indications are that turbulence is dominant only on larger scales than on the scale of individual dense cloud cores , where the velocity field is more likely to be ordered ( klapp et al .
the orthogonal alignment of bipolar outflows with the position angles of protostellar binary systems similarly implies that initial turbulence does not dominate the fragmentation mechanism ( tobin et al .
sph calculations of the collapse of turbulent molecular cloud cores lead to misalignments of the disks formed with the orbital planes of the binary or multiple star systems ( tsukamoto & machida 2013 ) . the models with non - zero magnetic fields all began with the same initial orientation : initially straight magnetic fields that were aligned with the @xmath56 axis ( i.e. , parallel to the cloud s rotation axis ) .
krumholz , crutcher , & hull ( 2013 ) found that magnetic field lines and rotational axes are randomly aligned in cloud cores
. however , chapman et al . (
2013 ) found a positive correlation between bipolar outflow directions and core magnetic field directions , with the former presumably being indicative of the protostar s rotation axis .
li , krasnopolsky , & shang ( 2013 ) studied the effects of magnetic field misalignments with rotation axes , as did price & bate ( 2007 ) and bk13 .
large - scale magnetic fields do not appear to be correlated with the orientations of spheroidal cloud cores , implying they do not shape the cores ( frau et al .
the boss ( 2009 ) pseudo - mhd models assumed the magnetic field to be aligned with the rotation axis , and used an approximate treatment ( boss 2007 ) of magnetic tension forces to model the effects of magnetic braking .
bk13 s true mhd calculations found that the magnetic field alignment with respect to the cloud rotation axis could change the critical initial field strength leading to collapse and fragmentation by a factor of about two .
finally , we note that bk13 found that the angular momentum loss due to magnetic fields initially aligned with the rotation axis was small ( only a few percent , comparable to that lost due to non - conservation of angular momentum by enzo during non - magnetic collapse ) unless the initial mass to flux ratio was less than the critical value , i.e. , unless the clouds were initially magnetically sub - critical , which is not the case for any of the present set of models .
we begin with a discussion of several test models with variations in the parameters that control the spatial resolution of the enzo code , a critical factor for fragmentation models ( e.g. , truelove et al .
1997 ; boss et al .
the intention is to settle upon a reasonable set of standard enzo parameters that can be used for the bulk of the mhd models .
bk13 found that a standard spatial resolution for the top grid of @xmath14 and a maximum of 6 levels of refinement was sufficient to demonstrate the same outcome for the standard isothermal test case ( boss & bodenheimer 1979 ) that had been found by previous high spatial resolution hd calculations , namely a runaway thin bar , and obtained a similar result with a top grid of @xmath10 .
here we use @xmath10 as our standard top grid resolution , though one model did explore a @xmath63 top grid ( model bnpp2r2 in table 1 ) .
we also varied the maximum number of levels of refinement ( mlr ) from mlr = 2 to 4 to 6 to 10 for variations on the collapse of several of the non - magnetic , prolate cloud models ( pp2r through pp2e in table 1 ) , all of which resulted in collapse and fragmentation into multiple systems when mlr = 6 was used . with mlr
= 4 , model pp2r-4 collapsed and formed a multiple system , but with mlr = 2 , model pp2r-2 collapsed and formed a single system , as the fragments formed merged together early in the collapse . on the other hand , with mlr = 10 , models pp2a-10 , pp2c-10 , and pp2e-10 all collapsed to form thin bars with such high densities that the amr routines generated a number of sub - grid points large enough to exceed the memory available on the carnegie xenia cluster nodes and halt the computations .
this occurred so early in the collapses ( @xmath64 , as defined below ) that the bar wind - up and fragmentation phase ( see below ) could not be reached ( at @xmath65 ) .
given these results , a choice of mlr = 6 seemed to be best for achieving reasonably high resolution as well as computational efficacy .
finally , with mlr = 4 , five models were run with variations in the choice of other enzo mesh refinement parameters , namely the minimummassforrefinement and the cellflaggingmethod ( e.g. , using the baryon mass density for refinement ) , to ensure that these parameters produced the correct outcome whether the jeans length criterion was employed or not .
all the models presented below used a standard safety factor of 8 in the jeans length constraint .
table 1 lists the initial conditions and basic results for the mhd models starting with prolate cloud cores .
the models span the applicable range of initial cloud rotation rates and increasingly magnetically supercritical clouds .
the results column describes the density configurations ( generally in the midplane , @xmath66 ) at the final time computed @xmath67 , expressed in terms of the free fall time at the initial central density , @xmath68 yr , where @xmath69 g @xmath13 . the final time was chosen to be when the maximum density was achieved , typically @xmath70 g @xmath13 to @xmath12 g @xmath13 .
many models were evolved considerably farther in time for curiosity s sake , into a realm where the evolution of the highest density regions was probably not being accurately depicted , and hence those later phases are not reported here .
figure 1 shows the typical evolution of a fragmenting cloud , in this case the non - magnetic model pp2a .
starting from the initial prolate cloud , seeded with random density perturbations ( figure 1(a ) ) , the cloud collapses on the free - fall time scale to form a thin bar ( figure 1(b ) ) , as occurs in the isothermal collapse of clouds with initial @xmath71 density perturbations ( e.g. , boss & bodenheimer 1979 , bk13 ) .
however , once densities rise above @xmath58 g @xmath13 , the nonisothermal evolution phase of collapse begins , and the rapidly rising thermal pressure stops collapse at the center of the cloud . as a result of this bounce , combined with the ongoing accretion of higher angular momentum gas , the bar begins to wind up and form a rotationally flattened protostellar disk ( figure 1(c ) ) , which continues to gain mass until it becomes gravitationally unstable and fragments into a multiple protostellar system ( figure 1(d ) ) .
the clumps seen in figure 1(d ) have masses ranging from 0.0051 to 0.13 @xmath0 , a typical range of clump masses for the models listed in table 1 at the final times .
these clumps are the first protostellar cores , composed primarily of molecular hydrogen , with maximum temperatures of about 300 k. figure 2 depicts the evolution of model bmpp2a , which is identical to model pp2a , except for having a magnetic field with an initial mass to flux ratio of 7.2 ( table 1 ) .
bmpp2a also collapses to form a high density bar ( figure 2(a ) ) , which proceeds to wind up and form a protostellar disk ( figure 2(b ) ) .
however , in this case , the added effects of the magnetic field lead to the formation of the transient protostellar ring ( off - axis density maximum ) seen in figure 2(c ) .
two density maxima form in the ring , but the magnetic field effects are sufficiently strong to prohibit the robust fragmentation seen in figure 1(d ) for model pp2a , and instead a single central protostar forms ( figure 2(d ) ) with a central temperature of 300 k. the spiral arms evident in figure 2(d ) , in combination with magnetic torques , are sufficient to transport angular momentum outward , allowing the central protostar to gain mass while avoiding fragmentation into a multiple system . the disk begins to expand slightly as a result of the outward angular momentum transport and the continued infall of higher angular momentum portions of the initial cloud core .
figure 3 illustrates the expanded size of the disk in model bmpp2a ( green regions ) compared to model pp2a , as well as the much thicker disk aspect ratio , and the filamentary structures associated with ideal mhd effects .
figure 4 shows the results for model bnpp2a , identical to models pp2a and bmpp2a except for having a mass to flux ratio of 14.4 , twice that of bmpp2a .
model bnpp2a collapses to form a bar that begins to wind up and form a disk ( figure 4(a ) ) , but in this case , the reduced magnetic effects are unable to prevent the formation of a disk whose outer regions rapidly fragment into a multiple system surrounding a central protostar ( figure 4(b ) ) .
this multiple system undergoes a chaotic evolution , with clumps forming and merging ( figure 4(c ) ) as the maximum density continues to increase ( figure 4(d ) ) . at the final time , three distinct clumps remain in the form of an inner binary and a hierarchical triple system , with clump masses of 0.034 , 0.086 , and 0.20 @xmath0 , though other lower mass clumps persist as well ( figure 4(d ) ) .
the densest clump in the central binary has a maximum temperature close to 500 k. a set of four models with a top grid of @xmath63 was run ( bnpp2r2 , bnpp2a2 , bnpp2c2 , and bnpp2e2 in table 1 ) , which were otherwise identical to the corresponding models with a @xmath10 top grid ( bnpp2r , bnpp2a , bnpp2c , and bnpp2e in table 1 ) . the results were qualitatively similar for both sets of models , with the exception of models bnpp2a2 and bnpp2a , where the higher resolution model ( bnpp2a2 ) did not fragment by the final time calculated .
this was a result of the initially higher resolution allowing a higher central density at an earlier time than model bnpp2r , so that the central clump in the protostellar disk was able to dominate the disk , gaining mass as its spiral arms transported angular momentum outward , and preventing the formation of competing clumps .
however , the higher initial rotation rate of model bnpp2r2 was able to overcome this effect , and the central protostar and disk eventually fragmented into a binary protostar system with masses of 0.040 and 0.019 @xmath0 .
table 1 also notes that two models formed single objects with disks whose midplanes were oriented out of the expected @xmath72 plane , namely model bpp2e ( @xmath73 ) and model blpp2e ( @xmath74 ) .
both of these models started with the smallest initial rotation rates and the strongest initial magnetic fields , and hence were the models where magnetic field effects were expected to dominate the collapse .
evidently in these models the magnetic fields were able to flip the orientation of the disks orbiting the central protostars .
this effect may be related to an effect found in the 3d mhd enzo calculations by zhao et al .
( 2011 ) , who found that magnetic field - matter decoupling occurred in their sink particle calculations , leading to a magnetic pressure build - up in a low density region close to the central sink particle , through which the field escaped . while sink particles were not employed here in models bpp2e and blpp2e , in both models a similar density and magnetic field configuration developed as seen by zhao et al .
( 2011 ) : low density bubbles filled with strong magnetic fields , adjoining the highest density , central region .
this interesting effect is worthy of further investigation ( see also krasnopolsky et al .
2012 ) , though it does not appear to have a major influence on whether a given cloud can fragment or not , which is the focus of the present work . in a similar context , the models with initially high values of @xmath75 and 100 @xmath1 g in table 1 all formed single protostars with small - scale , wide outflows along the @xmath56 axis , even model bpp2e with the flipped disk .
such outflows have also been found in many previous mhd collapse models ( e.g. , tomisaka 2002 ; machida et al .
2006 , 2008 ; brzle et al .
2011 ) .
table 2 lists the initial conditions and basic results for the mhd models starting with oblate cloud cores , most of which collapsed and fragmented .
figure 5 displays the evolution of a representative oblate cloud core , model bnpo2a , which collapsed to form a dense oblate disk ( figure 5(b ) ) before undergoing rotational and magnetic rebound ( figure 5(c ) ) and fragmenting into a small cluster of clumps , with masses ranging from 0.038 to 0.53 @xmath0 , the latter with a central temperature of nearly 500 k. whereas table 1 showed that a mass to flux ratio of about 14 times the critical ratio separates prolate cloud cores that collapse and fragment from those that collapse to form single protostars , table 2 shows that for initially oblate cloud cores , the initial mass to flux ratio only has to be less than about 11 in order for fragmentation to be avoided .
figure 6 compares the results for four oblate clouds that are identical except for their initial magnetic to flux ratios , namely models bmpo2r ,
bnpo2r , bopo2r , and po2r , with @xmath76 ratios of 11.4 , 22.8 , 44.9 , and @xmath77 , respectively , of the critical value .
figure 6 clearly shows the progressively stronger fragmentation efficiency as the initial @xmath76 ratio increases , showing that for an initial @xmath76 less than @xmath78 , fragmentation is as impeded as it is for the prolate cloud cores with @xmath76 less than @xmath79 . all things being equal ,
an oblate cloud is more likely to be able to fragment during collapse than a prolate cloud , given the greater amount of cloud mass at larger initial cloud radius , and thus higher specific angular momentum , so one expects that the critical @xmath76 ratio should be somewhat lower for oblate clouds than for prolate clouds , as seems to be the case for these models , though given the coarseness of the changes in @xmath76 ratio ( i.e. , by factors of two ) , this is not definitive .
it is also evident that the oblate clouds tend to produce more clumps when fragmentation occurs than prolate clouds , e.g. , oblate models po2r , po2a , po2c , and po2e produced a total of 31 clumps , whereas the corresponding prolate models pp2r , pp2a , pp2c , and pp2e produced a total of 11 clumps at the final time .
sink particles offer the promise of allowing 3d hd or mhd calculations to be continued much farther in time than would otherwise be the case , given the constraints placed on times steps by the courant condition and on spatial resolution by the need to resolve the jeans length . here
we briefly explore the reliability of enzo s sink particles to depict both fragmentation and the evolution of the sink particles that form , by running a set of models ( table 3 ) with variations in the merging mass ( sinkmergemass ) and the merging distance ( sinkmergedistance ) for sink particles , both described in section 2 .
the models are otherwise identical , except for changes by factors of 10 , in each direction , for these parameters , compared to the standard values used in model bopo2rsp ( table 3 ) and in the sink particle models to follow .
figure 7 shows the outcome of four of the models from table 3 , at comparable final times .
it is evident from figure 7 that while the details vary for each of these models , as expected given the chaotic orbital evolution of closely - packed , gravitationally interacting bodies , nevertheless even with significant changes in these two key parameters , the sink particle technique robustly predicts fragmentation into a large number of protostellar clumps .
in fact , table 3 shows that the final number of sink particles only varies from 5 to 8 , and the total amount of mass in the sink particles is always within the range of 1.48 to 1.63 @xmath0 , showing reasonably good overall agreement .
the masses of the sink particles in these five models are also similar , with less than a factor of 2 spread in the maximum particle mass , and a somewhat larger variation in the minimum particle mass .
these models indicate that the results of sink particle representations of fragmenting clouds are fairly robust , at least to within factors of @xmath3 2 .
table 4 lists the initial conditions and basic results for the mhd models starting with prolate cloud cores with sink particles .
figure 8 shows the results for four models that are identical except for having variations in the initial rotation rates .
the fastest rotating models , bopp2rs and bopp2as , each fragment into 3 sink particles ( figure 8(a , b ) ) , while the two slower rotating models ( bopp2cs and bopp2es ) collapse to form single sink particles ( figure 8(c , d ) ) . comparing to the corresponding models without sink particles in table 1 ( bopp2r , bopp2a , bopp2c , and bopp2e ) , all of which fragmented into multiple clumps , the sink particle models show a reduced tendency to form long - lasting fragments .
this is largely due to the fact that the sink particle models can be advanced considerably farther in time than the models without sink particles : table 4 shows that the former reach final times of 2.60 to 4.30 @xmath80 , compared to only 1.85 to 2.59 @xmath80 for the latter .
once fragmentation has occurred , continued orbital evolution is likely to only reduce the total number of clumps or sink particles , following orbital mergers or ejections .
in addition , by their very definition , the sink particles are designed to accrete nearby clumps and lower - mass particles , in order to speed the computation , which will result in fewer sink particles ( e.g. , table 3 shows that the total number of sink particles is reduced when the sinkmergedistance is increased to 333 au , and the opposite when sinkmergedistance is decreased to 3.3 au ) . finally , table 5 lists the initial conditions and basic results for the mhd models starting with oblate cloud cores with sink particles .
figure 9 shows the results for four models that are identical except for having variations in the initial rotation rates .
the fastest rotating models , bnpo2rs and bnpo2as , fragment into either 6 or 3 sink particles , respectively ( figure 9(a , b ) ) , while the two slower rotating models ( bnpo2cs and bnpo2es ) collapse to form single sink particles ( figure 9(c , d ) ) .
these outcomes are quite similar to those found for the initial prolate clouds with sink particles , shown in figure 8 . once again , compared to the corresponding oblate cloud models without sink particles ( table 2 ) , the inclusion of sink particles results in single protostars for the slower rotating clouds , as is to be expected , given the advanced times achieved with the sink particle technique , and the susceptibility of the slower rotating clouds to having their clumps be merged into a single protostar .
table 5 shows that by the final times , the sink particles are typically able to accrete more than half of the total oblate cloud mass of 2.73 @xmath0 . figure 10
compares the time evolution of the clump masses in the oblate cloud model bnpo2r with the sink particle masses in the corresponding model bnpo2rs .
while model bnpo2r shows a wide variation in clump masses once fragmentation begins after 1 @xmath80 , the sink particles in model bnpo2rs show a more monotonic increase in the minimum and maximum sink particle masses .
overall , the clump masses and sink particle masses are in basic agreement , at least at the level of factors of two .
two distinct outcomes characterize these magnetically supercritical clouds : collapse and formation of a single protostar with significant spiral arms , or else collapse and fragmentation into binary or multiple protostar systems , with multiple spiral arms . for prolate clouds ,
the critical magnetic field strength is about 12 @xmath1 g , or a mass to flux ratio of about 14 times the critical ratio ( table 1 ) , whereas for oblate clouds , the critical magnetic field strength is about 25 @xmath1 g , or a mass to flux ratio of about 11 times the critical ratio ( table 2 ) .
these results are comparable to the results found by bk13 , who studied the barotropic fragmentation of initially uniform density , spherical , solar - mass clouds with magnetic fields initially parallel to the cloud rotation axis , finding that a mass to flux ratio of about 13 times the critical ratio ( corresponding to a magnetic field strength of about 70 @xmath1 g ) separated clouds that collapsed and fragmented from those that collapsed to form single protostars ( see table 4 in bk13 ) .
these results suggest that the mass to flux ratio of a given initial cloud is a more robust predictor of the outcome of dynamic collapse than the initial magnetic field strength .
joos et al .
( 2013 ) also studied the collapse and fragmentation of initially centrally condensed cloud cores with an ideal mhd code , finding that fragmentation occurred for mass to flux ratios greater than 5 times the critical ratio , roughly a factor of 2 times smaller than found here .
however , their initial cloud masses were 5 @xmath0 , and contained significant turbulent velocity fields , which served to misalign the magnetic field direction compared to the overall cloud rotation axis .
joos et al .
( 2013 ) noted that fragmentation did not occur for this mass to flux ratio when the cloud mass was 1 @xmath0 , showing that their results depended strongly on the initial cloud mass . as a result of these factors , the results obtained here for centrally condensed clouds with masses of 1.73 and 2.73 @xmath0 appear to be in reasonable agreement with the results of joos et al .
( 2013 ) .
we now compare to the results of boss ( 2009 ) , who used a pseudo - mhd code with radiative transfer to study the 3d collapse of prolate and oblate spheroids identical to those studied here ( i.e. , compare tables 1 and 2 of boss ( 2009 ) with the present tables 1 and 2 , respectively ) . the boss ( 2009 ) models all started with an initial magnetic field strength @xmath81 g , i.e. , with magnetically sub - critical clouds , but used a prescription for ambipolar diffusion to reduce the reference magnetic field strength over a nominal ambipolar diffusion time scale of @xmath82 . as a result ,
by the time that collapse occurred , the nominal field strength @xmath83 had dropped to @xmath84 g for the 20 to 1 density ratio clouds also studied here , equivalent to starting with an initial mass to flux ratio of about 2 times the critical value , i.e. , super - critical .
boss ( 2009 ) found mixed results : the prolate clouds fragmented into quadruple , binary , or binary - bar configurations as the initial rotation rate dropped from @xmath60 to @xmath58 rad s@xmath24 , whereas the oblate clouds over that same span of rotation rates either did not collapse significantly , or collapsed to form a ring with no evidence for fragmentation .
given the differences between the boss ( 2009 ) models and those presented here , a more precise comparison does not appear to be possible , except to note in general terms that lowering the initial rotation rate tends to stifle fragmentation , and that magnetically super - critical clouds with mass to flux ratios significantly greater than the critical value are necessary for collapse and fragmentation into multiple protostar systems .
the present calculations rely on the barotropic approximation , which is a convenient assumption for searching a large parameter space of 3d mhd amr models , but which ultimately needs to be replaced by a proper treatment of radiative transfer , as demonstrated by boss et al .
radiative transfer has been included in cloud collapse models calculated with both amr ( krumholz et al . 2007 ; offner et al . 2009 ; commercon et al .
2010 ) and sph codes ( whitehouse and bate 2006 ; price & bate 2009 ; bate 2009 , 2011 , 2012 ) .
the amr calculations generally found that including radiative transfer led to less fragmentation in circumprimary disks than in models with the barotropic approximation , as the radiative heating from the newly formed primary protostar heated the disk enough to forestall gravitational instability .
the sph calculations found a similar trend , with fewer brown dwarfs formed when radiative transfer was included ( bate 2009 ) , a result in better agreement with observations .
however , bate ( 2011 ) found that radiative transfer resulted in protostellar disks with lifetimes as much as three times longer , the most massive of which might be susceptible to fragmentation .
bate ( 2012 ) studied the effects of radiation from newly formed stars during the collapse and early evolution of a 500 @xmath0 cloud that was initially highly jeans unstable .
bate ( 2012 ) found a similar result to bate ( 2009 ) for the collapse of a non - magnetic cloud with a mass ten times lower : fewer brown dwarfs .
bate s ( 2012 ) distribution of semi - major axes for the binary and multiple systems that resulted had a maximum for separations in the range of 1 to 100 au , similar to the observed distribution , which peaks at @xmath3 30 au ( raghavan et al .
2010 ) . however , bate s ( 2012 ) calculations resulted in more triple and higher order systems than binary systems , compared to the statistics for solar - type stars , where binary systems dominate by a factor of three or more ( raghavan et al . 2010 ) .
this could imply that magnetic fields , which were not included in bate ( 2012 ) , play an essential role in discouraging fragmentation enough to produce the correct distribution and statistics for solar - type stars .
alternatively , this discrepancy could to be due to the loss of some of these multiple systems through their expected chaotic post - formational orbital evolution in the middle of a protostellar cluster
. in any event , the present set of models serves to help define the extent to which magnetic fields can stifle fragmentation , depending on the precise initial conditions of a molecular cloud core .
these 3d ideal mhd models have shown that for the case of rotating , centrally - condensed , barotropic , magnetic cloud cores , mass to flux ratios of about 14 times the critical ratio separate clouds that collapse to form single protostars and those that collapse and fragment into multiple protostar systems . for roughly solar - mass clouds
, this critical value corresponds to a magnetic field strength of about 10 @xmath1 g , which falls in the midst of the observed values of @xmath85 and their lower limits ( figure 7 in crutcher 2012 ) .
hence it appears that magnetic molecular cloud cores span the range of @xmath85 values needed to produce both single and multiple protostar systems .
the calculations produce clumps with masses in the range of @xmath86 to 0.5 @xmath0 , clumps which will continue to accrete mass and interact gravitationally with each other .
it can be expected that the multiple systems will undergo dramatic subsequent orbital evolution , through a combination of mergers and ejections following close encounters , resulting ultimately in a small cluster of stable hierarchical multiple protostars , binary systems , and single protostars .
such evolution appears to be necessary in order produce the binary and multiple star statistics that hold for the solar - type stars in the solar neighborhood ( raghavan et al .
2010 ) .
we note that the use of sink particles resulted in fewer fragments at the final times , in large part because they allow the calculations to be followed significantly farther in time , allowing clumps to merge together .
however , given the approximate treatment of protostellar physics in the sink particles ( e.g. , the assumption that they do not accrete magnetic flux or angular momentum , compared to the clumps , which do contain significant magnetic flux and angular momentum ) , it is unclear if the sink particle results are to be trusted over the models without sink particles . in both cases , the inclusion of physical effects that have not been included in the present set of models ( e.g. , magnetic reconnection , non - ideal mhd , and radiative transport ) is needed to decide whether either treatment gives a realistic assessment of the protostellar fragmentation process .
evidently there is much remaining to be studied with 3d amr mhd codes .
we thank bo reipurth for discussions about binary and multiple star systems , the referee for a number of improvements to the manuscript , and nathan goldbaum , john wise , and bo zhao for advice about sink particles in enzo .
the computations were performed using the enzo code developed by the laboratory for computational astrophysics at the university of california san diego ( http://lca.ucsd.edu ) .
the calculations were performed on the carnegie xenia cluster , the purchase of which was partially supported by the national science foundation ( nsf ) under grant mri-9976645 .
this work was partially supported by nsf grant ast-1006305 .
we thank michael acierno and ben pandit for their invaluable assistance with the xenia cluster .
cccccccc bpp2r & @xmath61 & @xmath87 & 100.0 & 1.8 & 2.67 & single & z + bpp2a & @xmath88 & @xmath89 & 100.0 & 1.8 & 1.68 & single & z + bpp2c & @xmath59 & @xmath90 & 100.0 & 1.8 & 1.71 & single & z + bpp2e & @xmath91 & @xmath2 & 100.0 & 1.8 & 1.64 & single & x + blpp2r & @xmath61 & @xmath87 & 50.0 & 3.6 & 1.86 & single & z + blpp2a & @xmath88 & @xmath89 & 50.0 & 3.6 & 1.52 & single & z + blpp2c & @xmath59 & @xmath90 & 50.0 & 3.6 & 1.52 & single & z + blpp2e & @xmath91 & @xmath2 & 50.0 & 3.6 & 1.47 & single & y + bmpp2r & @xmath61 & @xmath87 & 25.0 & 7.2 & 2.34 & single & z + bmpp2a & @xmath88 & @xmath89 & 25.0 & 7.2 & 1.62 & single & z + bmpp2c & @xmath59 & @xmath90 & 25.0 & 7.2 & 1.61 & single & z + bmpp2e & @xmath91 & @xmath2 & 25.0 & 7.2 & 1.60 & single & z + bnpp2r & @xmath61 & @xmath87 & 12.5 & 14.4 & 2.76 & multiple & z + bnpp2a & @xmath88 & @xmath89 & 12.5 & 14.4 & 2.17 & multiple & z + bnpp2c & @xmath59 & @xmath90 & 12.5 & 14.4 & 1.92 & single & z + bnpp2e & @xmath91 & @xmath2 & 12.5 & 14.4 & 1.92 & single & z + bnpp2r2 & @xmath61 & @xmath87 & 12.5 & 14.4 & 3.23 & binary & z + bnpp2a2 & @xmath88 & @xmath89 & 12.5 & 14.4 & 1.74 & single & z + bnpp2c2 & @xmath59 & @xmath90 & 12.5 & 14.4 & 1.64 & single & z + bnpp2e2 & @xmath91 & @xmath2 & 12.5 & 14.4 & 1.61 & single & z + bopp2r & @xmath61 & @xmath87 & 6.3 & 28.4 & 2.59 & multiple & z + bopp2a & @xmath88 & @xmath89 & 6.3 & 28.4 & 1.91 & multiple & z + bopp2c & @xmath59 & @xmath90 & 6.3 & 28.4 & 1.85 & multiple & z + bopp2e & @xmath91 & @xmath2 & 6.3 & 28.4 & 1.88 & multiple & z + pp2r & @xmath61 & @xmath87 & 0.0 & @xmath77 & 2.11 & multiple & z + pp2a & @xmath88 & @xmath89 & 0.0 & @xmath77 & 1.68 & multiple & z + pp2c & @xmath59 & @xmath90 & 0.0 & @xmath77 & 1.66 & multiple & z + pp2e & @xmath91 & @xmath2 & 0.0 & @xmath77 & 1.66 & multiple & z + ccccccc bmpo2r & @xmath61 & @xmath87 & 25.0 & 11.4 & 2.03 & multiple + bmpo2a & @xmath88 & @xmath89 & 25.0 & 11.4 & 1.39 & single + bmpo2c & @xmath59 & @xmath90 & 25.0 & 11.4 & 1.36 & single + bmpo2e & @xmath91 & @xmath2 & 25.0 & 11.4 & 1.62 & single + bnpo2r & @xmath61 & @xmath87 & 12.5 & 22.8 & 2.38 & multiple + bnpo2a & @xmath88 & @xmath89 & 12.5 & 22.8 & 1.51 & multiple + bnpo2c & @xmath59 & @xmath90 & 12.5 & 22.8 & 1.50 & multiple + bnpo2e & @xmath91 & @xmath2 & 12.5 & 22.8 & 1.55 & multiple + bopo2r & @xmath61 & @xmath87 & 6.3 & 44.9 & 2.15 & multiple + bopo2a & @xmath88 & @xmath89 & 6.3 & 44.9 & 1.37 & multiple + bopo2c & @xmath59 & @xmath90 & 6.3 & 44.9 & 1.50 & multiple + bopo2e & @xmath91 & @xmath2 & 6.3 & 44.9 & 1.42 & multiple + po2r & @xmath61 & @xmath87 & 0.0 & @xmath77 & 2.31 & multiple + po2a & @xmath88 & @xmath89 & 0.0 & @xmath77 & 1.48 & multiple + po2c & @xmath59 & @xmath90 & 0.0 & @xmath77 & 1.69 & multiple + po2e & @xmath91 & @xmath2 & 0.0 & @xmath77 & 1.62 & multiple + bopo2rsp & 0.01 & 33 & 2.69 & 1.63 & 0.387 & 0.115 & 6 + bopo2rsq & 0.01 & 333 & 2.61 & 1.56 & 0.422 & 0.176 & 5 + bopo2rsr & 0.1 & 33 & 2.54 & 1.48 & 0.292 & 0.109 & 7 + bopo2rss & 0.01 & 3.3 & 2.57 & 1.68 & 0.281 & 0.145 & 8
+ bopo2rst & 0.001 & 33 & 2.52 & 1.53 & 0.460 & 0.0522 & 6 + cccccccc bnpp2rs & @xmath61 & @xmath87 & 12.5 & 14.4 & 5.82 & multiple & 0.470 + bnpp2as & @xmath88 & @xmath89 & 12.5 & 14.4 & 2.83 & multiple & 1.14 + bnpp2cs & @xmath59 & @xmath90 & 12.5 & 14.4 & 3.63 & single & 1.33 + bnpp2es & @xmath91 & @xmath2 & 12.5 & 14.4 & 2.59 & single & 1.10 + bopp2rs & @xmath61 & @xmath87 & 6.3 & 28.4 & 4.30 & multiple & 0.627 + bopp2as & @xmath88 & @xmath89 & 6.3 & 28.4 & 3.11 & multiple & 1.28 + bopp2cs & @xmath59 & @xmath90 & 6.3 & 28.4 & 2.60 & single & 1.14 + bopp2es & @xmath91 & @xmath2 & 6.3 & 28.4 & 3.63 & single & 1.41 + pp2rs & @xmath61 & @xmath87 & 0.0 & @xmath77 & 3.84 & multiple & 0.791 + pp2as & @xmath88 & @xmath89 & 0.0 & @xmath77 & 2.89 & single & 0.818 + pp2cs & @xmath59 & @xmath90 & 0.0 & @xmath77 & 2.79 & single & 1.36 + pp2es & @xmath91 & @xmath2 & 0.0 & @xmath77 & 2.90 & single & 1.40 + cccccccc bmpo2rs & @xmath61 & @xmath87 & 25.0 & 11.4 & 3.48 & multiple & 1.42 + bmpo2as & @xmath88 & @xmath89 & 25.0 & 11.4 & 1.43 & single & 0.309 + bmpo2cs & @xmath59 & @xmath90 & 25.0 & 11.4 & 1.50 & single & 0.955 + bmpo2es & @xmath91 & @xmath2 & 25.0 & 11.4 & 1.83 & single & 1.74 + bnpo2rs & @xmath61 & @xmath87 & 12.5 & 22.8 & 2.72 & multiple & 1.63 + bnpo2as & @xmath88 & @xmath89 & 12.5 & 22.8 & 1.84 & multiple & 1.65 + bnpo2cs & @xmath59 & @xmath90 & 12.5 & 22.8 & 1.47 & single & 0.887 + bnpo2es & @xmath91 & @xmath2 & 12.5 & 22.8 & 1.85 & single & 1.70 + bopo2rs & @xmath61 & @xmath87 & 6.3 & 44.9 & 2.58 & multiple & 1.55 + bopo2as & @xmath88 & @xmath89 & 6.3 & 44.9 & 2.00 & binary & 1.83 + bopo2cs & @xmath59 & @xmath90 & 6.3 & 44.9 & 1.62 & single & 1.39 + bopo2es & @xmath91 & @xmath2 & 6.3 & 44.9 & 2.59 & single & 2.21 + po2rs & @xmath61 & @xmath87 & 0.0 & @xmath77 & 1.72 & multiple & 0.748 + po2as & @xmath88 & @xmath89 & 0.0 & @xmath77 & 1.77 & binary & 1.78 + po2cs & @xmath59 & @xmath90 & 0.0 & @xmath77 & 2.96 & single & 2.64 + po2es & @xmath91 & @xmath2 & 0.0 & @xmath77 & 1.72 & single & 1.83 + | we present the results of a large suite of three - dimensional ( 3d ) models of the collapse of magnetic molecular cloud cores using the adaptive mesh refinement ( amr ) code enzo2.2 in the ideal magnetohydrodynamics ( mhd ) approximation .
the cloud cores are initially either prolate or oblate , centrally condensed clouds with masses of 1.73 or 2.73 @xmath0 , respectively .
the radial density profiles are gaussian , with central densities 20 times higher than boundary densities . a barotropic equation of state is used to represent the transition from low density , isothermal phases , to high density , optically thick phases .
the initial magnetic field strength ranges from 6.3 to 100 @xmath1 g , corresponding to clouds that are strongly to marginally supercritical , respectively , in terms of the mass to magnetic flux ratio .
the magnetic field is initially uniform and aligned with the clouds rotation axes , with initial ratios of rotational to gravitational energy ranging from @xmath2 to 0.1 .
two significantly different outcomes for collapse result : ( 1 ) formation of single protostars with spiral arms , and ( 2 ) fragmentation into multiple protostar systems .
the transition between these two outcomes depends primarily on the initial magnetic field strength , with fragmentation occurring for mass to flux ratios greater than about 14 times the critical ratio for prolate clouds .
oblate clouds typically fragment into several times more clumps than prolate clouds .
multiple , rather than binary , system formation is the general rule in either case , suggesting that binary stars are primarily the result of the orbital dissolution of multiple protostar systems . |
the nrg we employ is one appropriate to the study of continuum field theories @xcite which in turn is based upon wilson s numerical renormalization group first used to study quantum impurity problems @xcite .
the basic idea behind wilson s numerical approach is that one performs a series of numerical diagonalizations which are ordered such that `` important '' states in the considered hilbert space of the problem are taken into account first , while the effect of less important states are included only in subsequent diagonalizations .
the sequence of diagonalizations , the sequence of renormalizations , are done such that the numerical burden is the same at each step in the sequence .
the metric that determines the order in which states are taken into account , however , is arbitrary and can be chosen to be appropriate for the problem at hand . in wilson s case ,
namely the kondo problem where the impurity spin sits at the end of a half line lattice , states involving only the kondo spin together with nearby lattice sites are taken into account first . that such states are most significant is guaranteed by lattice hopping parameters that decrease in magnitude the further one gets away from the impurity .
in the case of a quantum critical ising model perturbed by a magnetic field ( considered in @xcite ) the hilbert space used to form the matrices in the sequence of numerical diagonalizations is that of the quantum critical ising model .
the states in this model are ordered in terms of energy ( relative to the unperturbed theory ) . here
low - energy states are the most important as the spin operator coupling to the magnetic field is highly relevant and so are taken into account first by the renormalization group .
however for models like the ll model in a trapping potential the complexity of the eigenstates ( as we have discussed in the main body of the text ) means that energy alone is not a sufficient metric for the nrg to distinguish important from less important states ; using such a limited metric would make the procedure drastically sub - optimal . to overcome this problem
we introduce a variational metric in the space of states similar to that used to compute the single particle spectrum of semiconducting carbon nanotubes @xcite .
although this latter problem could be represented as a perturbation of a conformal field theory ( cft ) , the cft was complex enough ( four bosons ) that the same issues arose .
this variational metric uses an iterative process which amounts to performing successively higher order computations in perturbation theory to determine which states are likely to significantly contribute to the low - energy spectrum of the fully perturbed theory . while the primary concern of the accompanying letter was the discussion of correlations ,
we feel it is important here to provide some details of the method so that readers can be reassured that we can describe the gas in the trap accurately , that is to say , that we can provide a reasonable description of the pre - quench state .
we will provide a more detailed description of these computations in a future publication @xcite . to demonstrate this we focus on the large @xmath95 limit where in the extreme tonks - girardeau limit ( @xmath29 ) the system reduces to one of free fermions ( at least for the computation of the energy ) in a trap and where @xmath96 corrections can be systematically computed by mapping the bose system to one of fermions interacting with an ultra - short ranged potential with strength proportional to @xmath96 @xcite .
we present the computation of the ground state energies in fig .
4 . in the left panel
we present the ground state energies of the gas ( with @xmath27 ) in traps of different strengths , @xmath34 .
we get good agreement between the nrg computation and the analytics ( better than 0.02% for the first three trap values and about 1% for the largest of the trap values , @xmath68 , studied ) . in the right panel
we show the computation of the ground state energies as a function of @xmath95 for values running from @xmath23 to @xmath30 for the same four values of the trap strength . in order to match analytics with numerics we needed to include corrections up to @xmath97 . computing the @xmath96 correction is straightforward .
the @xmath98 correction , when computed naively with second order perturbation theory , shows an ultraviolet divergence related to the ultra short ranged potential of the equivalent fermionic model .
this divergence can however be regulated with a point splitting procedure adopted from ref .
however these first two corrections are not enough to get good agreement for the @xmath30 values of the ground state energies .
we thus include the @xmath97 correction coming from the untrapped gas and which is readily computed from its integrability @xcite .
with these three analytic contributions in place , we see we get excellent agreement for the first three trap values while the strongest trap value ( @xmath68 ) continues to see a deviation on the order of @xmath99 for all values of @xmath95 .
finally we consider our ability to accurately compute the density profile of the gas in the trap .
this is a much more complicated quantity to compute than the ground state energies , since the matrix elements of the density operator must be employed . in fig .
5 we plot the density profile of the gas for two values of trapping strength and two values of @xmath95 .
again we work at larger values of @xmath95 where we can compare our numerics to an analytic computation ( @xmath29 plus first order @xmath96 corrections ) .
we see that we get good agreement in all cases .
( left ) and @xmath100 ( right ) .
top : the gas after release from a trap of strength @xmath101 and @xmath102 for @xmath23 .
bottom : the gas after release from the same traps but for @xmath87 .
the gge estimate is seen to be more accurate than the gce throughout the range of values considered.,scaledwidth=40.0%,scaledwidth=35.0% ] in fig .
6 we present data for system sizes @xmath103 and @xmath27 of the differences of the mdf between the gge and the de as well as the gce and the de .
we see that generically differences are present for small momenta .
however these differences seem to behave differently for the different ensembles . for de - gce these differences
remain approximately the same as one moves from @xmath103 to @xmath27 .
however for de - gge , these differences are notably reduced in increasing the system size , particularly for the @xmath30 case .
while we do not have sufficient data to perform a finite size scaling analysis , it appears that the difference de - gge is vanishing as system size grows , while the difference de - gce remains at some finite , though k - dependent , value . for @xmath87 ( top ) and @xmath23 ( bottom ) .
shown are the gases at ( @xmath24 ) , ( @xmath25 ) , and ( @xmath26 ) .
the @xmath27 data for higher momenta @xmath104 is not fully saturated .
the error bars ( given in insets only ) are calculated as for the mdf ( see text ) .
, title="fig:",scaledwidth=40.0% ] for @xmath87 ( top ) and @xmath23 ( bottom ) . shown
are the gases at ( @xmath24 ) , ( @xmath25 ) , and ( @xmath26 ) .
the @xmath27 data for higher momenta @xmath104 is not fully saturated .
the error bars ( given in insets only ) are calculated as for the mdf ( see text ) . , title="fig:",scaledwidth=40.0% ] while we focused on the mdf in the main body of the text , we also have computed the density - density correlation function ( static structure factor ( ssf ) ) @xmath105 ( with @xmath106 for definitions of the density operator see ref .
@xcite ) . in fig .
[ fig : ssf ] we plot the ssf in the de for @xmath30 and @xmath23 .
we see that the effects of the trap are restricted to small momenta and that they are enhanced as one reduces @xmath95 , moving away from the hardcore limit .
the role of scaling with system size is more complicated with the ssf than with the mdf . with the mdf we were able to argue ( at least for weak values of the trap ) that it remained invariant under the following scaling : @xmath107 .
this is not true for the ssf . at weak values of @xmath34
, we find that the ssf in the de is given by @xmath108 the simpler scaling for the mdf led us to emphasize this quantity in the main body of the text .
( left ) and @xmath100 ( right ) .
top : the gas after release from a trap of strength @xmath109 and @xmath102 for @xmath23 .
bottom : the gas after release from the same trap but for @xmath87 .
the gge estimate is seen to be more accurate than the gce throughout the range of values considered.,scaledwidth=40.0%,scaledwidth=35.0% ] in figs . 8 and 9 , we now contrast these results for the ssf in the de with those obtained in the gge and gce .
we plot the results vs. momentum expressed in units of @xmath110 .
the particular form of the ssf in at least the diagonal ensemble at small @xmath34 then suggests that in doubling @xmath42 and @xmath111 while halving @xmath34 , the value of the ssf will double ( see eqn .
[ srrde ] ) . for both displayed values of the interaction strength ,
the agreement between the de and gge is very good , and much better than between the de and gce .
this is true for the different trap strengths presented in both figs .
we note , however , that the disagreement between the de and gge is larger for @xmath30 than for @xmath23 ( fig . 8) . in terms of a finite size analysis
, we note that for the data in fig .
8 , the de - gge curves maintain , roughly speaking , the same shape between the @xmath103 and @xmath27 . because of how the ssf is scaling with system size and our choice of units for momenta , this means the difference between these two ensembles is decreasing as system size grows .
however the same can not be said for the de - gce curves .
for the larger system size , the de - gce curves are notably more upturned at small momenta suggesting that in the infinite volume limit , the two ensembles will yield different results for the ssf .
this effect is , however , much less pronounced for the data in fig .
9 where a stronger trap is used . here
the de - gge curve appears flatter for the larger system size data ( n = l=56 ) , while , the de - gce curve appears much the same for the two different system sizes .
a more definable trend may be elusive here because of the larger uncertainties associated with the larger trap values .
[ [ assessing - the - convergence - of - the - computation - of - the - mdf - and - ssf - in - the - de ] ] assessing the convergence of the computation of the mdf and ssf in the de ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in computing the mdf in the de for @xmath27 and @xmath30 , we truncated the expression for the ground state to 7000 states ( for @xmath112 , @xmath113 states were used ; for @xmath114 , @xmath115 ) .
the sum of the coefficients , @xmath116 , over these 7000 states was @xmath117 ( for @xmath118 and @xmath114 , respectively @xmath119 and @xmath120 ) .
the mdf itself can be checked using the sum rule @xmath121 .
the sum rule saturation for @xmath27 was @xmath122 , for @xmath123 , @xmath124 , and for @xmath103 , @xmath125 . for @xmath27 and @xmath23
, we kept @xmath126 states , leading to the truncated sum @xmath127 equalling @xmath128 .
the integrated mdf sum rule was saturated to @xmath129 . for @xmath118 ,
@xmath113 states were used , yielding saturations of @xmath119 and @xmath120 respectively . for @xmath103 ,
@xmath115 states were used , giving @xmath119 and @xmath130 . in computing the ssf in the de for @xmath30 and @xmath27
, we again truncated the expression for the ground state to the same 7000 states .
the f - sum rule @xmath134 then provides a convenient check on the end result at any fixed momentum .
this was saturated to @xmath135 for @xmath136 and @xmath137 by using abacus for excited states ; the lack of saturation around @xmath137 is visible in the top of fig .
[ fig : ssf ] . for the case of
@xmath23 we kept @xmath138 states in the ground state expansion in the basis of bethe states , yielding a truncated sum @xmath139 equal to @xmath140 .
this leads to a f - sum rule saturation for these three same momenta to the same value of @xmath140 .
for the other system sizes , we obtain the following saturations . for @xmath123 and @xmath30 , we used 100 states saturating @xmath119 of the wavefunction norm , and we obtained as the f - sum saturations at the three wavevectors , @xmath119 , @xmath141 , and @xmath142 . for @xmath123 and @xmath23 , we used 100 states saturating @xmath143 of the wavefunction norm , with f - sum saturations , @xmath144 , @xmath144 , and @xmath144 . for @xmath103 and @xmath30
, we used 500 states saturating @xmath120 of the wavefunction norm with f - sum saturations , @xmath120 , @xmath145 , and @xmath146 . and finally for @xmath103 and @xmath23
, we used 500 states saturating @xmath144 of the wavefunction norm , with f - sum saturations , @xmath147 , @xmath147 , and @xmath147 .
the error bars for the mdf and the ssf in the diagonal ensemble ( de ) ( as presented in figs . 1 and 3 of the main text , and in figs .
6 - 9 of the supplementary material ) were obtained according to the following scheme .
the nrg gives a ground state in the form @xmath148 where @xmath9 are exact eigenstates of the untrapped gas .
this sum , formally , should be over _ all _ eigenstates in the system ( i.e. @xmath149 ) . in running the nrg to determine the coefficients @xmath14 , we limit ourselves to a _ finite _ number of states ( approximately @xmath150 for @xmath27 ) .
the coefficients arrived at by the nrg always sum to 1 , i.e. @xmath151 . in computing the ssf or mdf in the de
, we further truncate this sum in order to make the computation of the correlation function numerically manageable , for example keeping only the first @xmath152 states ( @xmath153 states for @xmath27 ) .
the question is then what uncertainties these truncations introduce . to get a feel for this for the case @xmath27 , we compute the de at two different truncation levels , the first using all @xmath153 states . for this truncation the coefficients @xmath116 sum to @xmath154 in all cases .
but we also look at a second , more severe , truncation where @xmath155 . for both truncated wavefunctions , we renormalize the wavefunction so that it has unit norm .
we then ascribe the uncertainty to our computation of the de to the difference between these two results . for the cases @xmath28
, we feel the convergence of the sum @xmath156 to @xmath157 is so rapid that the results for the mdf and the ssf in the de are completely converged and the error due to truncation is negligible ( or at least smaller than the symbol size used in the plots ) .
we first note that in computing the ssf ( as well as the mdf ) in the gge and gce , that while the weights , @xmath158 and @xmath159 , in these ensembles are computed without fixing the particle number , when we compute the trace over states we only perform a partial trace involving those states with the same particle number , @xmath42 .
thus the results presented in figs .
6 - 9 of the supplementary material and 3 of the main text are computed , strictly speaking , in fixed - density sub - ensembles . to obtain error bars for the gge and gce , we compute the ssf and mdf in two different sub - ensembles and take the differences between these computations to obtain an ( again heuristic ) feel for the uncertainty . for the @xmath27 case , the two ( sub)-ensembles are the same collections of states used for determining uncertainties in the de
, one sub - ensemble has a sum @xmath160 while the other satisfies @xmath161 . for the @xmath103 case
, the first sub - ensemble again satisfies @xmath160 while the second sub - ensemble is ( differently from the @xmath27 case ) , is taken to be one half of the states of the first . for this case , keeping instead only states that gave a saturation of @xmath162 led to too few states in the ensemble to compute , even approximately , the ssf and mdf in the gge and gce .
9 t. kinoshita et al . ,
nature * 440 * , 900 ( 2006 ) .
s. hofferberth et al . ,
nature * 449 * , 324 ( 2007 ) .
n. gedik et al . , science * 316 * , 425 ( 2007 ) .
m. rigol et al .
nature * 452 * , 854 ( 2008 ) ; m. rigol et al .
* 98 * , 050405 ( 2007 ) .
e. h. lieb and w. liniger , phys . rev . *
130 * , 1605 ( 1963 ) .
t. barthel and u. schollwck , phys .
* 100 * , 100601 ( 2008 ) .
m. a. cazalilla , phys .
lett . * 97 * 156403 ( 2006 ) .
a. iucci , m. a. cazalilla , new j. phys . * 12 * , 055019 ( 2010 ) .
p. calabrese et al . , phys .
lett . * 106 * , 227203 ( 2011 ) .
d. rossini et al . , phys .
lett . * 102 * , 127204 ( 2009 ) .
d. fioretto and g. mussardo , new j. phys .
* 12 * , 055015 ( 2010 ) .
r. m. konik and y. adamov , phys .
lett . * 98 * , 147205 ( 2007 ) ; ibid .
lett . * 102 * , 097203 ( 2009 ) .
g. vidal , phys .
lett . * 93 * , 040502 ( 2004 ) .
s. r. white and a. e. feiguin , phys .
* 93 * , 076401 ( 2004 ) .
a. j. daley et al . , j. stat .
mech . : th .
p04005 ( 2004 ) . c. kollath
* 98 * , 180601 ( 2007 ) .
m. cramer et al . , phys .
* 101 * , 063001 ( 2008 ) .
n. a. slavnov , teor .
79 , 232 ( 1989 ) ; _ ibid .
_ , 82 , 389 ( 1990 ) .
caux , j. math . phys .
* 9 * , 095214 ( 2009 ) . ; j .- s .
caux , p. calabrese and n. slavnov , j. stat .
p01008 ( 2007 ) ; j .- s .
caux and p. calabrese , phys .
a * 74 * , 031605 ( 2006 ) .
see the supplementary material following the acknowledgements .
a. lenard , j. math .
* 5 * , 930 ( 1964 ) .
a. shashi , m. panfil , j .- s .
caux , a. imambekov , phys .
b * 85 * , 155136 ( 2012 ) . m. olshanii and v. dunjko , phys
lett . * 91 * , 090401 ( 2003 ) .
d. m. gangardt and g. v. shlyapnikov , phys .
lett . * 90 * , 010401 ( 2003 ) .
b. davies and v. e. korepin , arxiv : 1109.6604 . the generalized tba and the key role played in it by @xmath71 in determining the entire gge density matrix
is also discussed in j. mossel and j .- s .
caux , j. phys .
* 45 * , 255001 ( 2012 ) . c. n. yang and c. p. yang , j. math
. phys . * 10 * , 1115 ( 1969 ) .
l. santos et al . , phys .
lett . * 107 * , 040601 ( 2011 ) r. m. konik phys .
lett . * 106 * , 136805 ( 2011 ) .
k. wilson , rev .
phys . * 47 * , 773 ( 1975 ) .
r. m. konik and j .- s .
caux , to be submitted to phys . rev .
a. j .- s .
caux and p. calabrese , phys .
a * 74 * , 031605 ( 2006 ) .
t. cheon and t. shigehara , phys .
82 * , 2536 ( 1999 ) .
d. sen , int .
a * 14 * , 1789 ( 1999 ) . | using a numerical renormalization group based on exploiting an underlying exactly solvable nonrelativistic theory , we study the out - of - equilibrium dynamics of a 1d bose gas ( as described by the lieb - liniger model ) released from a parabolic trap .
our method allows us to track the post - quench dynamics of the gas all the way to infinite time .
we also exhibit a general construction , applicable to all integrable models , of the thermodynamic ensemble that has been suggested to govern this dynamics , the generalized gibbs ensemble . we compare the predictions of equilibration from this ensemble against the long time
dynamics observed using our method .
understanding non - equilibrium quantum quench behavior in low - dimensional systems is a difficult theoretical challenge . because one is initializing the system in a state that is not an eigenstate ,
this behavior is determined not merely by the system s ground state ( or a small number of excited states ) , but rather by some coherent sum of a large number of eigenstates .
if one wants to explore the emergence of a resulting steady state , the time evolution of this coherent sum must then be tracked over long periods of time .
this problem confronts theorists who wish to understand dynamics in perturbed quantum gases @xcite , ultrafast phenomena in superconductors @xcite , and questions of thermalization in integrable systems @xcite .
this last set of questions arise because of the surprising experimental finding that a perturbed one - dimensional bose gas retains memory of its initial non - equilibrium state over long periods of time @xcite and does not appear to relax to a state of thermodynamic equilibrium . to understand this ,
it was proposed @xcite that equilibriation does occur but not as described by a grand canonical ensemble ( gce ) .
instead the ensemble describing equilibriation needs to take into account the additional , non - trivial conserved quantities that , at least according to the theoretical minimal model of the gas ( the lieb - liniger ( ll ) model @xcite ) , are present in the system .
this new ensemble has been dubbed the generalized gibbs ensemble ( gge ) .
the gge takes as the density matrix @xmath0 where the @xmath1 form an independent , complete sequence of conserved quantities in the system and @xmath2 correspond to a set of generalized ( inverse ) temperatures .
computation of this density matrix is non - trivial and has only been successfully accomplished in certain special limits .
most of these limits are in models where interactions ( though not necessarily correlation functions ) correspond to a free model ( the hard core limit of the interacting bose gas @xcite , quadratic hamiltonians @xcite , luttinger liquids @xcite , the sine - gordon model at the free - fermion point and in the semi - classical limit @xcite , and the quantum ising model in the absence of a longitudinal field @xcite ) .
a notable exception was the study of fioretto and mussardo @xcite where it was possible to study quenches in general interacting integrable models but with the restriction to a very special set of quench protocols .
it is against this backdrop that we present a general methodology able to study non - equilibrium behavior and quench dynamics of low - dimensional interacting models , both integrable and non - integrable .
this method is predicated on a numerical renormalization group ( nrg ) able to study models which can be represented as perturbed integrable and conformal field theories ( cft ) @xcite : @xmath3 the ll model in a trapping potential takes this form .
we believe that this methodology is a valuable addition to other general methodologies used to study dynamics in low - dimensional systems such as the time - dependent density matrix renormalization group @xcite . at least for a subset of quenches , where we quench into an integrable system ( say by turning off the trapping potential in a ll system ) , we can track the dynamics for all times .
concomitant with the introduction of this tool to study quench dynamics , we present a general methodology to compute the density matrix of the gge using information arising from the application of the nrg .
we show how one can write down a simple set of equations governing the gge and how the entire infinite set of generalized temperatures , @xmath4 can be readily determined .
the specific example we consider is the ll model perturbed by a one - body parabolic trap @xmath5 , @xmath6 ( we will work in units where @xmath7 ) . in running the nrg
, we use the basis of eigenstates of the ll model and their matrix elements with respect to the trapping potential .
both the description of the states and the computation of matrix elements in the ll model are much more complicated than the examples of relativistic field theories where the nrg has been applied previously . the states in the ll model consist of n strongly interacting particles and not few - particle excitations above the true vacuum state , while the matrix elements do not see a chiral factorization as in a relativistic gapless theory but are n - dimensional determinants @xcite .
to tackle this , we took recourse to a highly optimized set of routines known as abacus @xcite which solves and evaluates all equations needed to characterize both the necessary eigenstates and their matrix elements .
this package has been shown to be able to successfully compute dynamical response functions for the ll model @xcite .
we first use the nrg to extract the ground state of the ll model in a trap @xcite .
the nrg produces the ground state of the gas , @xmath8 , as a linear combination of exact eigenstates , @xmath9 , of the ll model : @xmath10 in order to accurately describe the ground state in the nrg procedure we typically consider on the order of @xmath11 states .
we then consider a sudden release of the trap , that is we will study the gas where we quench into an integrable model .
for these types of quenches our methodology gives us the ability to study the evolution of the gas for arbitrary times .
each state , @xmath9 , appearing in the ground state is characterized by a set of n ( one for each particle ) rapidities ( quasi - momenta ) @xmath12 .
these rapidities are solutions to the bethe equations , @xmath13 and can be readily obtained to arbitrary accuracy . with the nrg we can compute the coefficients @xmath14 with reasonably high accuracy @xcite .
time evolution under the post - quench hamiltonian ( the unperturbed ll model ) is extremely simple .
if @xmath15 is the energy of state @xmath9 , the time evolution is described by @xmath10 because each state s energy , @xmath15 , is given in terms of the @xmath16 s as @xmath17 , we can compute the phases appearing in the above sum to arbitrary accuracy for arbitrary time . to characterize the evolution of the gas in the long time limit we compute the momentum distribution function ( mdf ) @xmath18 in the diagonal ensemble ( de ) .
an observable @xmath19 in this ensemble is simply given by @xmath20 to compute this correlation function we insert a resolution of the identity between @xmath21 and @xmath22 and use a specially - designed version of abacus for excited states to compute all of the necessary matrix elements @xcite .
( top ) and @xmath23 ( bottom ) .
shown are the gases at ( @xmath24 ) , ( @xmath25 ) , and ( @xmath26 ) .
error bars are given for the @xmath27 data alone and are estimated from the speed of convergence of the nrg ( see @xcite ) we believe the @xmath28 data is completely converged .
the mdf of the untrapped gas ( @xmath27 ) is shown for comparison as is the analytic expression available for the tonks - girardeau gas @xmath29 from ref .
@xcite.,title="fig:",scaledwidth=40.0% ] + ( top ) and @xmath23 ( bottom ) . shown
are the gases at ( @xmath24 ) , ( @xmath25 ) , and ( @xmath26 ) .
error bars are given for the @xmath27 data alone and are estimated from the speed of convergence of the nrg ( see @xcite ) we believe the @xmath28 data is completely converged .
the mdf of the untrapped gas ( @xmath27 ) is shown for comparison as is the analytic expression available for the tonks - girardeau gas @xmath29 from ref .
@xcite.,title="fig:",scaledwidth=40.0% ] in fig .
[ fig : mdf ] we plot the mdf in the de of the gas post - release for two values of c ( @xmath30 and @xmath23 ) and for a variety of system sizes , with @xmath31 fixed and keeping @xmath32 . for comparison
we also plot the mdf of the gas in its ground state .
we see , as expected , that the mdf of the gas is perturbed from that of the ground state at low momenta but remains unchanged from the ground state mdf at higher momenta .
the relative insensitivity to different values of @xmath33 is consistent with a perturbative ( in @xmath34 ) computation of the mdf in the de at @xmath29 which shows @xmath35 . here
@xmath36 is the mdf of the ground state , the constant @xmath37 @xcite , and @xmath38 is the velocity of the gas .
the scaling with @xmath39 , and @xmath34 indicated by this expression implies that variations in @xmath40 between different system sizes in fig .
1 are due to finite size corrections which are small ( on the order of the symbol size ) .
as an important check of our results , the high momenta tails of the mdf s at @xmath23 behave as the predicted @xmath41 @xcite .
while the diagonal ensemble tells us what the final steady state of the gas is after its release , a question of primary interest is whether the steady state can be associated with some ensemble .
it has been postulated @xcite that for a quench into an integrable system the correct ensemble to use is the gge ensemble in eqn .
[ eq : rhogge ] .
the @xmath1 s are here non - trivial polynomials in the field operators ( and their derivatives ) @xcite . the action of the @xmath1 s on the states , @xmath9 is straightforward . with each state , @xmath9 , characterized by a set of @xmath42 rapidities , @xmath43 ,
the action of the @xmath1 upon @xmath9 is @xmath44 that is to say , @xmath1 acts on the state like an i - th power sum .
this shows that the @xmath1 s are both a complete and independent set of charges inasmuch as the polynomials form a complete and independent basis in the space of single variable functions . to compute @xmath45
the most straightforward path is to compute @xmath46 at @xmath47 and insist that the set of @xmath2 s is such that @xmath48 gives the same answer . in the case of the hard core limit
this is readily doable as the @xmath1 s can be written in terms of a more amenable basis , the momentum occupation numbers : @xmath49 where @xmath50 tells you whether there is a particle with rapidity of the form @xmath51 for @xmath52 .
in this basis of charges , @xmath53 simplifies to @xmath54 , i.e. for such expectation values the ensemble factorizes , and @xmath55 is readily computed .
this simplification , however , does not exist away from the hard core limit and we are instead left with a complicated non - linear minimization problem which on the face of it does not obviously have a solution .
we now show that it does and that the @xmath2 s can be computed readily .
we do so through a ( generalized ) thermodynamic bethe ansatz @xcite . because the action of the charges @xmath1 on the states , @xmath9 , are given simply in terms of the rapidities , @xmath43 , identifying the state , to ask that @xmath56 amounts to asking whether there is a set of @xmath57 s , @xmath58 , such that @xmath59 there is in fact such a set .
we can moreover determine its rapidity distribution , which we will call @xmath60 , directly from @xmath8 .
to each state , @xmath61 , we associate a distribution , @xmath62 , governing the @xmath57 s of that particular state : @xmath63 then @xmath60 is the weighted sum of the @xmath64 s : @xmath65 in particular @xmath66 . and @xmath67 for both the gge and gce ensembles for a gas with @xmath27 , @xmath23 , and a prequench trap strength , @xmath68 .
for the gce ensemble the effective temperature is @xmath69 .
the quantities plotted are symmetric about @xmath70.,scaledwidth=35.0%,scaledwidth=28.0% ] @xmath71 contains , implicitly , all the information to characterize the action of @xmath45 on a eigenstate of the ll model @xcite
. a distribution of @xmath57 s must be consistent with the bethe equations ( eqn .
[ beqn ] ) . in the continuum limit , these equations can be rewritten as @xcite @xmath72 where @xmath73 is the density of holes in the @xmath57-distribution and @xmath74 .
now the gge is derived by the same principles as the grand canonical ensemble : namely entropy is maximized subject to the constraints of fixed conserved charges ( energy for the grand canonical ensemble , all the charges , @xmath1 , for the gge ) .
thus associated with gge is a generalized free energy @xmath75 where @xmath76 is a generalized energy .
it corresponds to the action of @xmath45 on a state @xmath77 : @xmath78 in particular knowing @xmath79 then allows us to compute general expectation values in the gge .
while @xmath79 differs from its form in the grand canonical ensemble , @xmath80 is the standard entropy @xcite of a system with a given distribution of particles , @xmath71 , and holes , @xmath81 : @xmath82.\end{aligned}\ ] ] we now show that we can express @xmath79 in terms of @xmath71 that we derived from @xmath8 . if we minimize the generalized free energy we arrive at a constraint between the particle and hole distributions and @xmath79 : @xmath83 where @xmath84 . thus to determine @xmath79 we take our knowledge of @xmath60 obtained from @xmath8 , use eqn .
( [ den_eqn ] ) to determine @xmath81 which then gives us @xmath85 . from eqn .
( [ en_eqn ] ) , we then can fix @xmath79 . ) in the gce and gge for the gas after release from a trap of strength @xmath86 for @xmath87 ( top ) and @xmath23 ( bottom ) .
we again show the mdf of the untrapped gas for comparison ( blue stars).,title="fig:",scaledwidth=40.0% ] + ) in the gce and gge for the gas after release from a trap of strength @xmath86 for @xmath87 ( top ) and @xmath23 ( bottom ) .
we again show the mdf of the untrapped gas for comparison ( blue stars).,title="fig:",scaledwidth=40.0% ] following this procedure we plot in fig .
[ fig : rho ] @xmath71 and @xmath79 for the gas in the hard core limit . for comparison we plot what these quantities would be if instead of a generalized gibbs ensemble
, the thermodynamics was governed by the grand canonical ensemble .
( in this case we use the standard thermodynamic bethe ansatz equations @xcite to determine what @xmath88 and @xmath89 need to be , i.e. what the effective temperature needs to be , if they are to reproduce the correct density and average energy of the system @xmath90 . )
we see that both @xmath71 and @xmath79 have considerably more structure than that of their grand canonical counterparts .
we now use this ability to compute @xmath91 , to compute various expectation values of observables in the gge . in fig .
[ fig : mdfge ] we plot the mdf as computed in the de and in both the gge and gce .
the error estimate is computed similarly as in fig .
[ fig : mdf ] ( see @xcite for details ) . for the data at hand
, we see that for low momenta the two ensemble averages , gge and gce , disagree with the de .
however the gge provides a considerably better match to the de than does the ordinary thermal ensemble gce . from the finite size comparison ( see fig . 3 of @xcite , it can be argued ( although not conclusively ) that at small but finite @xmath92 , this difference will vanish with increasing system size . the disagreement between ensembles in the data is not entirely surprising .
the logic of the gge is such that it is expected to describe correlations that are local in space ( and that involve a distance scale significantly smaller than the system size ) .
we thus do not expect the correlations at @xmath93 to be particularly well described by the gge .
however there is the possibility that the differences between ensembles will remain at finite @xmath94 even in the infinite volume limit . in recent work @xcite
the entropy associated with the de was shown to be considerably smaller than that of the gge implying that the de is more tightly constrained than the gge , i.e. the gge seems to be missing correlations .
it would be interesting to understand if this missing entropy is solely associated with non - local correlations . in conclusion
, we demonstrated how an nrg based on exploiting the integrability of the ll model can be used to study the time - dependent evolution after a quantum quench where a 1d gas is released from a parabolic trap .
we have also demonstrated how to use the information arising from the nrg to construct the corresponding gge which has been suggested as a possibility for governing the post - quench dynamics . while we have focused on the ll model , this methodology is applicable to any non - relativistic integrable theory of which the heisenberg and xxz spin chains are two prominent examples .
acknowledgements : this research was supported by the us doe ( de - ac02 - 98ch10886 ) , the new york center for computational sciences at stony brook
university / brookhaven national laboratory , the foundation for fundamental research on matter , and the netherlands organisation for scientific research .
we thank g. brandino and j. mossel for useful discussions . |
while the standard model(sm ) is in full agreement with all experimental data@xcite , it is generally believed that new physics must exist at a scale not far beyond the reach of existing accelerators .
associated with this scale may be a host of new and exotic particles .
a new neutral gauge boson , @xmath1 , is the most well - studied of all exotic particles and is the hallmark signature for extensions of the sm gauge group .
current direct searches for the existence of such particles at the tevatron@xcite suggest that their masses must be in excess of 500 - 700 gev depending upon their couplings to the sm fermions and their kinematically accessible decay modes .
if such a particle is found at future colliders the next step will be to ascertain its couplings to all of the conventional fermions . in this way
, we may hope to identify whether this new particle corresponds to any one of the many @xmath1 s proposed in the literature or is something else entirely . at hadron colliders ,
a rather long list of observables has been proposed over the years to probe these couplings each with its own limitations@xcite .
it has been shown under idealized conditions , at least within the context of @xmath5-inspired models , that the lhc(@xmath6 tev , @xmath7 ) will be able to extract useful information on all of the @xmath1 couplings for @xmath8 below @xmath9 tev .
it is _ not _ clear , however , how much of this program can be carried out using realistic detectors at the lhc@xcite and how well it generalizes to other extended gauge models since detailed simulation studies have yet to be performed . at the nlc , when @xmath10 ( the most likely scenario for a first generation machine given the tevatron bounds ) a @xmath1 can only manifest itself indirectly as deviations in , , cross sections and asymmetries from their sm expectations
this is analogous to the observation of the sm @xmath11 at pep / petra / tristan energies through deviations from the expectations of qed .
fortunately the list of useful precision measurements that can be performed at the nlc is reasonably long and the expected large beam polarization(@xmath12 ) plays an important role essentially doubling the number of useful observables . in the past
, analyses of the ability of the nlc to extract @xmath1 coupling information in this situation have taken for granted that the value of @xmath8 is already known from elsewhere , , the lhc@xcite .
( in fact , one might argue that if a 1 tev @xmath1 is discovered at the lhc , a future lepton linear collider designed to sit on this @xmath1 must be built and will thus quite easily determine all of the @xmath1 couplings in analogy to slc and lep . ) here we address the more complex issue of whether it is possible for the nlc to obtain information on couplings of the @xmath1 if the mass were for some reason _ a priori _ unknown . in this case
we would not only want to determine couplings but the @xmath1 mass as well .
we will limit our discussion below to the @xmath13 channel and ignore the additional information available through @xmath14 collisions@xcite .
if the @xmath1 mass were unknown it would appear that the traditional nlc @xmath1 coupling analyses would become problematic .
given a set of data at a fixed value of @xmath2 which shows deviations from the sm , one would not be able to _ simultaneously _ extract the value of @xmath8 as well as the corresponding couplings .
the reason is clear : to leading order in @xmath15 , rescaling all of the couplings and the value of @xmath8 by an overall common factor would leave the observed deviations from the sm invariant . in this approximation
, the @xmath1 exchange appears only as a simple contact interaction .
thus as long as @xmath10 , the only potential solution to this problem lies in obtaining data on deviations from the sm at _ several _ , distinct @xmath2 values and combining them into a single fit .
it is clear from the beginning that all of the set of @xmath2 values chosen for this analysis can not lie too far below the @xmath1 mass otherwise we would always remain in the contact interaction limit .
it must be , for at least one of the @xmath2 choices , that sub - leading terms of relative order @xmath15 are of numerical importance .
this suggests that the maximum value of @xmath2 within this set should only be about a factor of 2 - 3 lower than the @xmath1 mass . here
we report on the first analysis of this kind , focussing on observables involving only leptons and/or @xmath0-quarks . in performing such an analysis
, we need to know how many @xmath2 values are needed .
we need to know how we distribute the integrated luminosity(@xmath16 ) to optimize the results .
similarly , we must address whether such an analysis can be performed while maintaining model - independence . in this _ initial _ study we begin to address these and some related issues .
it is clear that that there is a lot more to be done before we have the answers to all these questions .
in order to proceed with this benchmark study , we will make a number of simplifying assumptions and parameter choices . these can be modified at a later stage to see how they influence our results .
( the basic analysis follows that discussed in @xcite . ) in this analysis we consider the following ten observables : the total production cross sections for leptons and @xmath0-quarks , @xmath17 , the corresponding forward - backward asymmetries , @xmath18 , the left - right asymmetry obtained from flipping the initial electron beam polarization , @xmath19 , and the polarized forward - backward asymmetry , @xmath20 . for @xmath21 final states
, we include the average @xmath22 polarization , @xmath23 , as well as the forward - backward asymmetry in the @xmath22 polarization , @xmath24 .
other inputs and assumptions are summarized as follows : anomalous trilinear gauge couplings fun = 8.fun and games with extended technicolor models e,@xmath25,@xmath22 universality isr with @xmath26 + @xmath27 , @xmath28 @xmath29 + @xmath30 , @xmath31 @xmath32 + @xmath33 ,
@xmath34 neglect @xmath35-channel exchange in @xmath36 of special note on this list of assumptions are : ( @xmath37 ) a @xmath0-tagging efficiency(@xmath38 ) of @xmath39 for a purity(@xmath40 ) of 100@xmath41 , ( @xmath42 ) the efficiency for identifying all leptons is assumed to be 100@xmath41 , although only @xmath39 of @xmath22 decays are assumed to be polarization analyzed , ( @xmath43 ) a @xmath44 angle cut has been applied to all final state fermions to mimic the anticipated detector acceptance , ( @xmath45 ) a strong energy cut to remove events with an excess of initial state radiation(isr ) has been made this is critical since events with lower effective values of @xmath2 substantially dilute our sensitivity , ( @xmath46 ) it has been assumed that both the beam polarization(@xmath12 ) and machine luminosity(@xmath16 ) are both well measured . it is important to note that we have _ not _ included the @xmath35-channel contributions to @xmath47 in these calculations .
in addition to the above , final state qed as well as qcd corrections are included , the @xmath0-quark and @xmath22 masses have been neglected , and the possibility of any sizeable @xmath48 mixing has also been neglected ; this is an excellent approximation for the @xmath1 mass range of interest to us given that we are not interested in the @xmath49 mode . since our results will generally be statistics limited , the role played by the systematic uncertainties associated with the parameter choices above will generally be rather minimal , especially in the lepton case .
larger systematics should possibly be associated with the @xmath0-quark final states@xcite but they have been ignored here for simplicity . to insure model - independence , the values of the @xmath1 couplings , , @xmath50 , as well as @xmath8 ,
are chosen _ randomly _ and _ anonymously _ using a random number generator from rather large ranges representative of a number of extended gauge models .
monte carlo data representing the above observables is then generated for several different values of @xmath2 . at this point ,
the values of the mass and couplings are not ` known ' _ a priori _ , but will later be compared with what is extracted from the monte carlo generated event sample . following this approach
there is no particular relationship between any of the couplings and there is no dependence upon or relationship to any particular @xmath1 model .
( we chose to normalize our couplings so that for the sm @xmath11 , @xmath51 . ) performing this analysis for a wide range of possible mass and coupling choices then shows the power as well as the potential limitations of this technique . to get an understanding for how this procedure works in general we will make three representative case studies for the @xmath1 mass and couplings , labelled here by i , ii and iii .
there is nothing special about these three choices and several other parameter sets have been analyzed in comparable detail to show that the results that we display below are rather typical . to begin our analysis ,
let us try choosing three distinct @xmath2 values .
specifically , we generate monte carlo ` data ' at @xmath520.5 , 0.75 and 1 tev with associated integrated luminosities of 70 , 100 , and 150 @xmath53 , respectively . these luminosities are only slightly larger than the typical one year values as conventionally quoted@xcite and assumes a reasonable time evolution of the collider s center of mass energy . subsequently , we determine the 5-dimensional @xmath54 cl allowed region for the mass and couplings from a simultaneous fit to all of the leptonic and @xmath0-quark observables following the input assumptions listed above .
this 5-dimensional region is then projected into a series of 2-dimensional plots which we now examine in detail .
1 - 3 show the results of our analysis for these three case studies compared with the expectations of a number of well - known @xmath1 models@xcite . to be specific
we have considered ( _ i _ ) the @xmath5 effective rank-5 model(er5 m ) , which predicts a @xmath1 whose couplings depend on a single parameter @xmath55 ; ( _ ii _ ) the sequential standard model(ssm ) wherein the new @xmath56 and @xmath1 are just heavy versions of the sm particles ( of course , this is not a true model in the strict sense but is commonly used as a guide by experimenters ) ; ( _ iii _ ) the un - unified model(uum ) , based on the group @xmath57 , which has a single free parameter @xmath58 ; ( _ iv _ ) the left - right symmetric model(lrm ) , based on the group @xmath59 , which also has a free parameter ( @xmath60 ) of order unity which is just the ratio of the gauge couplings and , lastly , ( _ v _ ) the alternative left - right model(alrm ) , based on the same extended group as the lrm but now arising from @xmath5 , wherein the fermion assignments are modified in comparison to the lrm due to an ambiguity in how they are embedded in the * 27 * representation . by examining these figures ,
several things are immediately apparent the most obvious being that two distinct allowed regions are obtained from the fit in all three cases .
this ambiguity is two - fold and _ not _ four - fold in that there is a unique choice of @xmath0-couplings for a fixed choice of leptonic couplings .
( of course as we might hope , the input values are seen to lie nicely inside one of them . )
this two - fold ambiguity occurs due to our inability to make an absolute determination of the overall sign of _ one _ of the @xmath1 couplings , , @xmath61 .
if the sign of @xmath61 were known , only a single allowed region would appear in figs . 1 - 3 for both leptons and @xmath0-quarks and a unique coupling determination would thus be obtained .
note that this _ same _ sign ambiguity arises in slc / lep data for the sm @xmath11 and is only removed through the examination of low - energy neutrino scattering .
secondly , we see that the leptonic couplings are always somewhat better determined than are those of the @xmath0-quark , which is due to the fact that the leptonic observables involve only leptonic couplings , while those for @xmath0-quarks involve both types .
in addition , there is more statistical power available in the lepton channels due to the assumption of universality and the fact that the leptonic results employ two additional observables related to the final state @xmath22 polarization .
if the @xmath0-quark observables had significant systematic errors , these allowed regions would become larger still .
thirdly , we see in particular from figs .
1a - b and 3a - b the importance in obtaining coupling information for a number of different fermion species .
if only the fig .
1a(3b ) results were available , one might draw the hasty conclusion that an @xmath5-type @xmath1 had been found .
1b(3a ) clearly shows us that this is not the case .
evidently none of the @xmath1 s associated with cases i - iii correspond to _ any _ well - known model .
fourthly , we note that changing the beam polarization from @xmath62 to @xmath63 does not appreciably alter our results .
lastly , as promised , the @xmath1 mass is determined in all three cases , although with somewhat smaller uncertainties in case ii .
it is important for the reader to realize that there is nothing special about any of these three particular cases .
it is clear from this set of random choices for masses and couplings that this procedure should be viable for @xmath1 masses up to about 2 tev for the set of integrated luminosities that we have chosen unless both of the leptonic @xmath1 couplings are accidentally small resulting in a reduced sensitivity to the existence of the @xmath1 .
it is straightforward to extend this analysis to fermion final states other than leptons and the @xmath0-quark .
the extrapolation to charm is the most obvious in that apart from identification efficiencies and potentially larger systematic errors there is little difference in performing the five - dimensional fit with either @xmath0 or @xmath64 since the fermion couplings were randomly chosen .
( of course we might imagine , however , now doing a more ambitious _
seven_-dimensional fit with all of the couplings being allowed to float . )
the extension to @xmath35-quarks would be less straightforward due to their large mass and rapid decay to @xmath65 . in principle , however , the same set of observables could be constructed for top as was used in the @xmath0-quark analysis above .
of course , the clever reader must now be asking the question ` why did we start off using 3 different values of @xmath2 why not 2 or 5 ? '
this is a very important issue which we can only begin to address here .
let us return to the mass and couplings of case i and generate monte carlo ` data ' for only two values of @xmath2=0.5 and 1 tev with luminosities of @xmath16= 100 and 220 @xmath53 , respectively , thus keeping the total @xmath16 the _ same _ as in the discussion above .
repeating our analysis we then arrive at the ` 2-point ' fit as shown in fig .
4a ; unlike fig .
1a , the allowed region in the leptonic coupling plane does not close and extends outward to ever larger values of @xmath66 ; we find that a similar result occurs for the @xmath0-quark couplings which are even more poorly determined .
the corresponding @xmath1 mass contour is also found not to close , again extending outwards to ever larger @xmath8 values .
we realize immediately that this is just what happens when data at only a single @xmath2 is available . for our fixed @xmath16 ,
distributed as we have now done , we see that there is not a sufficient lever arm to simultaneously disentangle the @xmath1 mass and couplings .
of course the reverse situation can also be just as bad .
we now generate monte carlo ` data ' for the case i mass and couplings in 100 gev steps in @xmath2 over the 0.5 to 1 tev interval with the same total @xmath16 as above but now distributed as 30 , 30 , 50 , 50 , 60 , and 100 @xmath53 , respectively .
we then arrive at the ` 6-point ' fit shown in fig .
4b which suffers a problem similar to that presented in fig .
what has happened now is that we have spread the fixed @xmath16 too thinly over too many points for the analysis to work .
these same results are found to hold for all three cases .
this brief study indicates that a proper balance is required to simultaneously achieve the desired statistics as well as an effective lever arm to obtain the @xmath1 mass and couplings .
it is important to remember that we have _ not _ demonstrated that the ` 2-point ' fit will never work .
we note only that it fails with our specific fixed luminosity distribution for the masses and couplings associated with cases i - iii .
it is possible that for ` lucky ' combinations of masses and couplings a 2-point fit will suffice or it may work if substantially more luminosity is achievable .
it is certainly true that all cases where at least three values of @xmath2 are used will allow simultaneous mass and coupling extraction provided the integrated luminosity is available .
clearly , more work is required to further address these issues .
how do these results change if @xmath8 _ were _ known or if our input assumptions were modified ? in this case we use as additional input in our analysis the value of @xmath8 chosen by the monte carlo and perform four - dimensional fits to the same set of ` data ' .
let us return to case i and concentrate on the allowed coupling regions corresponding to a choice of negative values of @xmath67 ; these are expanded to the solid curves shown in figs . 5a and 5c .
( there will also be a corresponding region where @xmath67 are positive , which we ignore for the moment . ) the large dashed curve in fig .
5a corresponds to a reduction of the polarization to @xmath63 with the same relative error as before .
while the allowed region expands the degradation is not severe .
if the @xmath1 mass were known , the ` large ' ellipses shrink to the small ovals in fig .
5a ; these are expanded in fig .
this is clearly a radical reduction in the size of the allowed region .
we see that when the mass is known , varying the polarization or its uncertainty over a reasonable range has very little influence on the resulting size of the allowed regions . from fig .
5c we see that while knowing the @xmath1 mass significantly reduces the size of the allowed region for the @xmath0 couplings , the impact is far less than in the leptonic case for the reasons discussed above .
6a - d show that case i is not special in that similar results are seen to hold as well for cases ii and iii . for both of these cases , as in case
i , there is an enormous reduction in the size of the allowed region for the leptonic couplings of the @xmath1 but the corresponding allowed region for the @xmath0-quark shrinks by about only a factor of two . in case iii
, there is hardly _ any _ reduction in the size of the allowed @xmath0-quark coupling region .
just how large a role does large beam polarization play in obtaining our results ? this becomes a critical issue , especially at higher energies , if our lepton collider is actually a muon collider where we may need to trade off luminosity for high beam polarization@xcite .
as an extreme case , we repeat the previous analysis of cases i - iii without including the observables associated with beam polarization in the fits ; luminosities remain the same as before .
the results of this approach are shown in figs .
7a - d . in case
i , shown in fig .
7a , the loss of the polarization dependent observables causes the allowed region in the leptonic coupling plane to open up and no closed region is found . correspondingly , the @xmath0-quark couplings to the @xmath1 as well as the @xmath1 mass are constrained but are no longer localized .
somewhat better results are obtained in cases ii and iii , shown in figs . 7b and 7c . for case
ii , the size of the allowed region essentially doubles in the @xmath68 plane and triples in the @xmath69 plane , as shown in fig .
the @xmath70 constraints are found to relax in a corresponding manner . in case ii , while we are hurt by the lack of polarized beams we are still able to carry out the basic program of coupling extraction and @xmath1 mass determination unlike case i. in case iii , fig .
7c shows that the leptonic couplings are reasonably constrained without beam polarization but now neither the @xmath1 mass nor the @xmath0-quark couplings were found to be constrained . from these considerations
we may conclude that beam polarization is of critical importance to this analysis unless the couplings lie in a ` lucky ' range .
it is clear that for _ arbitrary _ values , we will not be able to simultaneously obtain mass and coupling determinations without large beam polarization .
we note , however , that this conclusion can soften dramatically if the @xmath1 mass is already known . what happens for larger @xmath1 masses or when data at larger values of @xmath2 becomes available ?
( as stated above , the ` reach ' in our coupling determinations was @xmath71 tev using the ` data ' at 500 , 750 and 1000 gev . )
let us assume that the ` data ' from the above three center of mass energies is already existent , with the luminosities as given .
we now imagine that the nlc increases its center of mass energy to @xmath2= 1.5 tev and collects an additional 200 @xmath53 of integrated luminosity , which corresponds to 1 - 2 design years . clearly for @xmath1 masses near or below 1.5 tev our problems
are solved since an on - shell @xmath1 can now be produced .
thus we shall concern ourselves only with @xmath1 masses in excess of 2 tev , inaccessible in the lower energy study above . figs .
8a - d show the result of extending our previous procedure now using 4 different @xmath2 values , for two distinct choices ( iv and v ) of the @xmath1 mass and couplings .
these ` 4-point ' results are a combined fit to the data at all four center of mass energies .
we show the results for both the general case where @xmath8 is unconstrained as well as when it is already determined by other data .
( as before , only one of the allowed pair of ellipses resulting from the overall sign ambiguity is shown for simplicity . )
note that the @xmath1 input masses we have chosen are well in excess of 2 tev where the lhc may provide only very minimal information on the fermion couplings@xcite . clearly by using the additional data from a run at @xmath2=1.5 tev this technique can be extended to perform coupling extraction for @xmath1 masses in excess of 2.5 tev .
the maximum ` reach ' for the type of coupling analysis we have performed is not yet determined .
it seems likely , based on these initial studies , that the extraction of interesting coupling information for @xmath1 masses in excess of 3 tev may be possible for a reasonable range of coupling parameters .
in this paper we have shown that it is possible for the nlc to extract information on the @xmath1 couplings to leptons and @xmath0-quarks even when the @xmath1 mass is not _ a priori _ known and , in fact , determine the @xmath1 mass .
this has been demonstrated in a model - independent manner by randomly and anonymously choosing the mass and couplings of the @xmath1 and demonstrated the power of precision measurements at future linear colliders .
the critical step for the success of the analysis is to combine the data available from measurements performed at several different center of mass energies . for a reasonable distribution of the luminosities the specific results we have obtained suggest , but do not prove , that data sets obtained at at least 3 different energies are necessary for the procedure to be successful .
the mass ` reach ' for this approach is approximately twice the highest center of mass energy available . several question
remain about the optimization of our approach ; these will be addressed in future work . .25 in
the author would like to thank j.l .
hewett , s. godfrey , s. riemann , k. maeshima and h. kagan for discussions related to this work .
# 1 # 2 # 3 mod .
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 nucl .
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys . rep .
* # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
* # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 phys .
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 rev . mod .
# 1 * , # 2 ( # 3 ) # 1 # 2 # 3 z. phys .
* # 1 * , # 2 ( # 3 ) # 1 # 2 # 3 int . j. mod .
# 1 * , # 2 ( # 3 ) for a complete set of references and a recent review of the physics of new gauge bosons see , m. cvetic and s. godfrey , carleton university report ocip / c-95 - 2 , 1995 , a part of the dpf long - range planning study to be published in _ electroweak symmetry breaking and physics beyond the standard model _ , eds
. t. barklow , s. dawson , h. haber , and j. seigrist ( world scientific 1996 ) .
hewett and t.g rizzo in a4 4551 1989 and in _ proceedings of the dpf summer study on high energy physics in the 1990 s _ , snowmass , co , july 1988 , ed . by s. jensen(world scientific , singapore 1989 ) , p.235 . | we examine of the capability of the next linear collider to determine the mass as well as the couplings to leptons and @xmath0-quarks of a new neutral gauge boson , @xmath1 , below direct production threshold . by using simulated data collected at several different values of @xmath2
, we demonstrate how this can be done in a model - independent manner via an anonymously case approach .
the importance of beam polarization to the success of this program is discussed .
the procedure is shown to be easily extended to the case of top and charm quark couplings .
psfig.sty -0.5 in 6.5 in 8.5 in 4.0=0.5 in -0.5 in # 1_.25ex # 1_-.1em#1 # 1_-.2em#1 # 1_-.1em.25ex # 1_.25ex # 1_.4ex # 1^#1 # 1^#1 / plus0pt minus0pt # 1#20.7ex # 1@xmath3 # 1@xmath4 # 1 # 1 # 1[#1 ] * an exploration of below threshold @xmath1 mass and coupling determinations at the nlc * thomas g. rizzo + + 0.45 in submitted to physical review * d*. |
understanding the properties of non - equilibrium systems has been a central effort of the scientific community for many years . of specific interest are non - equilibrium processes that take place at the nano - meter scale and at which energy is converted from one form to another , for instance photovoltaic ( pv ) energy conversion , photochemistry and photosynthesis . in these cases , the interaction between electrons and photons under non - equilibrium conditions plays an essential role .
theoretical modeling of such processes is a challenging task , since the interacting nature of the system and its many - body characteristics , the multitude of constituents , the presence of external environments , and the non - equilibrium conditions must all be taken into consideration . even harder to address theoretically are situations in which the system has two independent fluxes , originating from separate non - equilibrium drivings , and are both far away from the linear response regime ,
a situation which is designated as _ strong non - equilibrium_. a paradigmatic example for such a system are photo - voltaic cells , where the two fluxes are the heat flux , originating from the huge temperature difference between the sun and the earth , and the particle current originating from the voltage difference between the electrodes . in recent years
a new and exciting class of photo - voltaic cells has emerged , namely molecular photo - cells , where the energy conversion process takes place at the single molecule level @xcite . here
we propose a formalism to study non - equilibrium transport in molecular junctions , and use it to investigate a model for the molecular photo - cell , a single molecular donor - acceptor complex attached to electrodes and subject to external illumination .
this model was recently suggested @xcite to be the minimal model to describe pv energy conversion in ideal , single - molecule heterojunction organic pv cells . in ref .
@xcite pv conversion efficiency was analyzed using the ( essentially classical ) rate equations for the electronic degrees of freedom .
the dynamics and non - equilibrium properties of the phonons and photons were ignored , being considered only within a ( non - self - consistent ) mean - field approximation and assumed to have equilibrium distributions . here
we show the non - equilibrium properties of the phonons and photons have a strong impact on the pv conversion properties in realistic parameter range and can not be neglected .
the formalism we present here allows us to treat electrons , photons and phonons fully quantum mechanically and on an equal footing ( without resorting to a mean - field approximation ) and to take into account the action of the environments producing a strong non - equilibrium situation .
we use the many - body lindblad quantum master equation @xcite to describe the environments , which consist of metallic electrodes in touch with a molecular complex , a phonon bath ( at ambient temperature ) and a photon bath ( at the solar temperature ) .
non - equilibrium is induced by two sources , namely the temperature difference between the incoming photons ( originating from the sun ) and the device , and the voltage bias between the electrodes . using this formalism
, we calculate the non - equilibrium densities of electrons , photons and phonons , the electric current and power output , and the thermodynamic efficiency at maximal power .
we find that under certain conditions , the distribution functions for the phonons and photons can be very different from the equilibrium distributions , and therefore approximating the system as close to equilibrium is not a valid approximation .
we then study the signature of non - equilibrium on the energy conversion efficiency of the system .
in addition to be able to include strong non - equilibrium effects and to account for all constituents on equal footing , our formalism allows one to introduce effects of dephasing and decoherence in a simple and physically transparent way .
we study how the loss of quantum coherence affects the efficiency , and show that classical electron transfer ( as opposed to coherent electron propagation ) _ enhances _ the efficiency .
the system under consideration is composed of a molecular junction , in which a donor - acceptor ( d - a ) complex is placed between two metallic electrodes ( fig .
[ fig1 ] ) @xcite .
this is an idealization of an envisioned future single - layer molecular photovoltaic cells , where a self - assembled layer of d - a pairs is placed on a conducting substrate , and is covered by a top transparent electrode . for the donor ( d ) , we consider the highest occupied molecular orbital ( homo ) with energy @xmath1 and lowest unoccupied molecular orbital ( lumo ) with energy @xmath2 .
for the acceptor ( a ) , we consider only the lumo with energy @xmath3 ( the energy of the a s homo is much lower and is unaffected by any dynamics in the photocell @xcite ) .
recent advances in the experimental ability to measure photo - conductivity and pv conversion in single - molecule junctions @xcite make our theoretical model experimentally relevant .
0.5truecm we begin by examining the electron , photon and phonon densities at zero bias voltage .
note that the system is still out of equilibrium due to the temperature difference between the photon and phonon baths . in fig .
[ fig2](a ) the photon density @xmath4 , the phonon density @xmath5 , and the electronic occupation of the d - lumo , @xmath6 , are plotted as a function of @xmath7 coupling @xmath8 .
we set the orbital energies to be @xmath9 ev , @xmath10ev , and @xmath11ev @xcite .
the @xmath12 interaction is fixed at @xmath13 ev ( dashed vertical line in fig .
[ fig2](a ) ) .
three distinct regimes are observed : ( i ) at small @xmath7 coupling , @xmath14 , the system is close to equilibrium , and the photon occupation is defined by the solar temperature @xmath15 ( dotted line ) .
the phonons are excited according to the ambient temperature ( which is very small compared to @xmath16 , and consequently the phonon occupation is very small ) , and the occupation of the d - lumo level is also very small .
( ii ) as the @xmath7 and @xmath12 interactions become comparable , energy is transferred from the photons to the phonons , mediated by excitation of electrons from the d - homo to the d - lumo level . as a result , the d - lumo and phonon occupations increase , while the photon occupation decreases .
alternatively , this situation can be described in terms of heating ( although the notion of temperature is not applicable out of equilibrium , it is still useful to think in terms of an effective temperature ) : the junction is locally ( and efficiently ) heated by the photons .
this heat is transferred to the phonons , resulting in an elevated effective phonon temperature and , consequently , enhanced phonon occupation .
( iii ) when @xmath7 coupling is large , @xmath17 , there is no longer efficient transfer of heat to the phonons .
however , the d - lumo occupation continues to rise , due to energy pumping from the photons to the d - lumo .
0.0truecm , phonon density @xmath5 and the electronic occupation of the donor lumo as a function of @xmath7 coupling @xmath8 .
dashed line indicates the value of @xmath18 .
dotted horizontal line marks the equilibrium occupation of solar photons.(b ) efficiency at maximum power @xmath19 as a function of the @xmath7 coupling ( same parameters as in ( a ) ) .
inset : typical current @xmath20 and power output @xmath21 vs. bias voltage @xmath22 .
, title="fig : " ] an important measure of the operational efficiency of the molecular pv cell is the efficiency at maximal power , @xmath19 , defined as the ratio between the cell s maximal power output @xmath23 and the corresponding input power @xmath24 supplied by the photons @xcite .
the output power is given by @xmath25 , where @xmath20 is the particle current through the system , and the input power is calculated in a similar way @xcite from @xmath26 , which is related to the photon - induced part of the particle current .
as shown in the inset of fig .
[ fig2](b ) , a typical @xmath27 curve has a maximal power . in fig .
[ fig2](b ) , @xmath19 at the maximal power is plotted as a function of the @xmath7 coupling ( we use the same parameters as in fig .
[ fig2](a ) ) .
for very small @xmath7 coupling ( @xmath280.002 ev ) , the efficiency is very small , and grows linearly with @xmath8 . in this regime ,
the time it takes for an electron to absorb a photon is larger than the time the photons spent in the cell ( defined by @xmath29 ( which translates to @xmath30 ev ) , and so the photon absorption is very small leading to poor efficiency .
this regime is followed by a plateau regime , where any energy transfered from the photons to the electrons is quickly dissipated by phonons and is not converted into electrical power .
there is thus little change in the efficiency , as long as the rate of photon absorption is smaller than the eletron - photon - relaxation time .
only when the @xmath7 coupling reaches the @xmath12 coupling @xmath13 ev the efficiency begins to increase : in this regime , energy is transferred to the electrons by the photons faster than can be dissipated by the phonons , and as a result an increasing amount of this energy is transferred into electronic power , resulting in a rise of efficiency .
it is important to note that the results described in fig .
[ fig2 ] , especially in the region where the electron - phonon and electron - photon couplings are of the same order , _ can not be obtained by assuming equilibrium distributions for the phonons and photons _ , and this situation is described here for the first time . since in future realistic devices the strength of the electron - photon and electron - phonon interactions are unknown , a situation where they are of similar magnitude may occur , in which case the system dynamics can not be described as close to equilibrium , and the full non - equilibrium dynamics need to be taken into account . to further demonstrate the power of this method , we next discuss the effect of coulomb interactions on the efficiency . in the hamiltonian of eq .
[ hamiltonian ] , the a - lumo energy , @xmath31 , already includes the coulomb repulsion energy on the acceptor @xcite . in excitonic systems ,
the coulomb interaction is typically considered through the `` exciton binding energy '' , defined by the coulomb interaction term @xmath32 between an electron at the d - lumo and a hole in the d - homo . while in methods such as non - equilibrium green s function adding coulomb interaction requires substantial effort , the present method does not require either additional technical complexity or additional computational power to account for any coulomb interaction effects . in fig .
[ fig2_c ] we show the efficiency at maximum power @xmath33 as a function of the exciton coulomb energy @xmath34 ( which can be estimated from , e.g. density - functional calculations ) .
we set @xmath35ev and @xmath36ev .
we find an almost linear decrease in the efficiency , with a reduction of @xmath37 for @xmath38ev . 0.0truecm as a function of the exciton coulomb energy @xmath34 , title="fig : " ]
the next question we wish to address is the extent to which the quantum nature of the system affects the pv conversion efficiency , a question which is beyond the reach of the formalism presented in refs .
the formalism we present here allows us to access , in addition to fully quantum - coherent processes described above , also incoherent processes .
the most important incoherent processes are electron transfer from the d - lumo to the a - lumo , described classically in ref .
these are addressed here by adding an additional pair of @xmath39operators that accounts for incoherent transitions , namely @xmath40 , where @xmath41 .
thus , the pair of parameters @xmath42 and @xmath43 describe the strength of the coherent and incoherent donor - acceptor electron transfer processes , respectively . in what was shown in fig .
[ fig2 ] , the d- and a- lumo levels were connected by quantum - mechanical bonding .
in contrast , ref .
@xcite accounted for the electronic transfer between the d- and a- lumo levels by an incoherent ( or classical ) transfer process .
it is thus of interest to interpolate between the fully quantum case ( @xmath44 ) , through the mixed quantum - classical case ( @xmath45 ) , to the fully classical case ( @xmath46 ) .
to do so , we define a variable @xmath47 such that @xmath48 , and define @xmath49 ev and @xmath50 ( as in ref .
we now parametrize @xmath42 and @xmath43 with @xmath47 , @xmath51 ( @xmath52 is the heaviside unit step - function ) .
this parametrization is shown on the right inset of fig .
[ fig3 ] , and is constructed such that for @xmath53 the system is fully coherent , for @xmath54 the system is mixed ( both quantum and classical processes ) , and for @xmath55 the system is fully incoherent , so the range @xmath56 interpolates between all three cases . in fig .
[ fig3 ] we plot the efficiency at maximum power @xmath19 as a function of the position of the a - lumo , @xmath57 and the parameter @xmath47 .
we set @xmath58 as in ref .
we find that the quantum coherence or classical decoherence ( parametrized by @xmath47 ) has a profound effect on the efficiency of the molecular pv - cell in two important aspects .
first , the optimal position of the a - lumo energy differs according to the nature of the transition under consideration : quantum ( coherent ) , both quantum and classical , or classical d - a transitions ( solid lines in fig . [ fig3 ] ) . for the last case ( @xmath59 ) , we find that @xmath31 is optimal at @xmath60 ev , verifying the result of ref . @xcite . when quantum correlations are added ( @xmath61 ) ,
two peaks emerge at @xmath62 ev , and a lower peak emerges at @xmath63 ev . for a system with only quantum transitions ( @xmath64 , enlarged in the back inset in fig .
[ fig3 ] ) , the lower peak vanishes , and the optimal lumo positions are at @xmath65 ev and @xmath66 ev .
thus , in the design of optimal molecular pv cells , it is important to take into account the quantum nature .
second , as can be clearly seen in fig .
[ fig3 ] , the addition of classical d - a transitions _ increases _ the efficiency substantially by more than an order of magnitude .
this finding is surprising , since one would expect that incoherent ( and dissipative ) transitions would lead to a decrease in the efficiency .
to understand the origin of this effect , we performed time - dependent calculations ( not shown ) and found that for the coherent case , an electron that is excited to the d - lumo coherently oscillates between the d- and a - lumo , while for the incoherent case the electron decays from the d- to the a - lumo exponentially ( and its return rate is exponentially small ) .
this implies that in the coherent case the electron spends much more time in the d - lumo than in the incoherent case , before transferring to the right electrode .
since the electron can decay back to the d - homo ( emitting a phonon ) only directly from the d - lumo , the longer it spends on the d - lumo , the higher the probability for non - radiative decay back to the d - homo , leading to a decrease in efficiency .
this phenomena is similar to dephasing - assisted transport conjectured to occur in biological systems , @xcite but here is the first time it is discussed and demonstrated in the context of molecular pv cells . since in realistic single - molecule
both coherent and incoherent effects may be important , they must be included in a theoretical description of the system .
0.0truecm as a function of the position of the a - lumo @xmath57 and the parameter @xmath47 which describes the interpolation from coherent to incoherent electron d - a processes ( see text ) .
solid lines mark the coherent ( @xmath67 ) , mixed ( @xmath68 ) and incoherent ( @xmath69 ) donor - acceptor electron transfer processes .
dashed lines are guides to the eye , showing the position of the maximal efficiency for the different cases .
right inset : parametrization of @xmath42 and @xmath43 with @xmath47 .
back inset : efficiency as a function of the position of the a - lumo @xmath57 at @xmath53 ( fully coherent system).,title="fig : " ] to illustrate this connection between dynamics and efficiency , we address the relaxation dynamics of the donor - acceptor system . considering only the d - a lumos and the right electrode ( without the photons and phonons ) , we construct the lindbladian @xmath70-matrix of eq .
[ lindbladian ] ( at zero bias ) , by constructing a vector form for the lindblad equation of eq .
[ lindbladian ] , @xmath71 .
the @xmath70-matrix has a zero eigenvalue , which defines the steady state .
the ( real part of the ) rest of the eigenvalues define the relaxation rates towards the steady state .
the minimal rate @xmath72 ( i.e. , eigenvalues of @xmath70 with smallest real part which , we numerically check , is non - zero ) represents the longest relaxation time for the system to reach the steady state from any general state . in fig .
[ fig5 ] , we plot the decay rate @xmath72 of the model which only contains d - a lumos and the right electrode ( without the photons and phonons ) as a function of the position of the a - lumo @xmath57 for the three cases of fully coherent ( @xmath67 ) , mixed ( @xmath73 ) and incoherent ( @xmath69 ) donor - acceptor electron transfer processes . as can be seen
, the decay rate is much smaller for the fully coherent case , indicating a longer relaxation time .
this is in line with the observation above that slower relaxation dynamics lead to lower efficiency .
in addition , we point out that the relaxation times ( which are much simpler to calculate than the full efficiency ) serve as an indicator for the efficiency , even though they do not capture the fine details required for an optimal design of the system ( see fig .
[ fig3](a ) and ( b ) ) .
0.0truecm ( on a log scale ) as a function of the position of the acceptor lumo , @xmath57 , for the three cases of fully coherent ( @xmath67 ) , mixed ( @xmath73 ) and incoherent ( @xmath69 ) donor - acceptor electron transfer processes.,title="fig : " ] to further examine possible effects of coherence on the efficiency of the molecular pv cell , we study a system where the donor has two degenerate d - lumo levels which have been introduced experimentally @xcite .
here , we study a simplified system ( schematically depicted on the right side of fig .
[ fig6 ] ) , in which the photons and the phonons excite electrons with equal amplitudes from the d - homo to the two d - lumo levels . each of the levels is coupled to the a - lumo with the same hopping amplitude @xmath74 , and they are coupled to each other with a complex hopping amplitude @xmath75 . in fig .
[ fig6 ] , the efficiency at maximum power @xmath19 is plotted as a function of the inter - lumo coupling @xmath76 ( solid line ) and the phase @xmath77 ( dashed line ) for @xmath78 , i.e. , no incoherent d - a transfer .
we find that while @xmath76 has little effect on the efficiency , the phase @xmath77 has a significant effect ( increasing the efficiency by up to @xmath79 ) .
surprisingly , we also find that this quantum interference effect persists even when incoherent d - a transfer is included ( @xmath58 as in fig .
[ fig3 ] ) , and an substantial increase of @xmath80 is observed by varying @xmath77 ( dotted line ) .
-0.5truecm as a function of hopping amplitude @xmath76 between the donor lumo levels and the acceptor lumo ( solid line ) , and hopping matrix element phase @xmath77 ( dashed line ) , indicating the effect of quantum coherence on the efficiency for fully coherent d - a transfer ( @xmath78 ) .
the dotted line is the same for a mixed coherent - incoherent transfer ( @xmath81 ) .
, title="fig : " ] 0.5truecm _ discussion_. in summary , we have proposed a novel formalism to study non - equilibrium quantum transport in molecular junctions , and applied it to investigate a minimal model of pv energy conversion in ideal , single - molecule pv cells .
the results shown above indicate that quantum coherence effects are important in determining the non - equilibrium energy conversion performance of molecular pv cells .
the formalism presented here sets the stage for a fully coherent quantum mechanical calculation of energy conversion in more realistic models for molecular pv cells and can be directly linked to quantum chemistry methods ( such as density - functional theory ) .
the progress in the experimental ability to measure photo - conductivity and pv conversion in single - molecule junctions @xcite allows one to envision real pv devices composed of a single molecular junction or a molecular monolayer , making our theoretical model experimentally relevant . furthermore , our method can include both coherent and incoherent effects , making it a useful tool in the study of other energy conversion processes such as photosynthesis , where both classical and quantum processes take place @xcite , or other chemical and photo - chemical processes @xcite .
the full hamiltonian of the molecular pv cell , including the molecular orbitals , the photons and the phonons , may be written as @xmath82 where @xmath83 is the hamiltonian for the molecular complex , @xmath84 is the photon ( phonon ) hamiltonian , and @xmath85 describes the electron - photon ( phonon ) interaction ( we set @xmath86 hereafter ) , _ m & = & _ x_x c^_xc_x - t_d - a(c^_d,2c_a+h.c ) , _ pht & = & _ 0 a^a , _ phn= _
0 b^b , _ m - pht & = & _ e - pht a^c^_d,1c_d,2+h.c . , _ m - phn & = & _ e - phn b^c^_d,1c_d,2+h.c . .
[ hamiltonian]here @xmath87 creates ( annihilates ) an electron in the d - homo ( @xmath88 ) , d - lumo state ( @xmath89 ) or a - lumo state ( @xmath90 ) , with the corresponding level energies @xmath91 , @xmath92 creates ( annihilates ) a photon with energy @xmath93 , and @xmath94 creates ( annihilates ) a phonon with the same energy . in principle
one should consider many photon ( and phonon ) modes , however the strongest effect on the dynamics comes from the resonant photons ( with energy same as the homo - lumo gap ) .
the electron - photon hamiltonian @xmath95 describes ( within the rotating wave approximation ) the process ( and its reverse process ) in which an electron in the d - lumo state relaxes to the d - homo state and emits a photon , with the electron - photon ( @xmath7 ) coupling @xmath8 .
the electron - phonon hamiltonian @xmath96 is similar to @xmath95,but with phonons instead of photons .
for simplicity we consider spinless electrons , since spin does not play a significant role in the energy conversion process . to study the dynamics of the system , we use the lindblad equation to model the system and the environments @xcite .
the essence of the lindblad approach is that instead of describing the environment by encoding it into a self - energy ( as is done in the non - equilibrium green s function approach @xcite ) the environment is characterized by its action on the system .
this action is mapped onto so - called lindblad @xmath97-operators , which describe incoherent transitions of the system elements due to the presence of an environment .
the lindblad equation was recently employed to address various aspects of electron transport @xcite , yet in these studies the interaction with an environment was limited to electrons only , and the non - equilibrium dynamics of other constituents ( i.e. phonons or photons ) was not considered .
we assume that the left electrode is coupled only to the d - homo and that the right electrode is coupled only to the a - lumo @xcite , as in fig .
the corresponding @xmath97-operators are then @xcite : _
, + & = & c^_d,1 , _ , -= c_d,1 , _ , + & = & c^_a , _ , -= c_a , [ vops1 ] where @xmath98 are electron transfer rates to the left and right electrodes , @xmath99 is the ambient temperature ( we take @xmath100 k ) , @xmath101 is the left - electrode chemical potential , @xmath102 is the right electrode chemical potential , @xmath22 is the voltage bias , @xmath103 are the fermi - dirac distributions of the left and right electrodes , and @xmath104 . following ref .
@xcite we set @xmath105 ( which corresponds to an energy scale of @xmath106 ev , much smaller than the other electronic energy scales , validating the use of the lindblad equations ) . for the bosons ( photons and phonons ) ,
similar @xmath97-operators that relate to the bose - einstein statistics of the boson baths are constructed , _
pht,+&= & a^ , _
pht,-= a , _ phn,+&= & b^ , _
pht,+= b , [ vops2]where @xmath107 are photon and phonon relaxation rates ( set to @xmath108 ) , @xmath109 k is the solar temperature , @xmath110 is the bose - einstein distribution , and @xmath111 .
once the @xmath97-operators are set , the dynamics are determined by the propagation of the density matrix via the lindblad equation , & = & -[,]+ _ ( -\ { ^ , } + ^ ) , [ lindbladian ] where @xmath112 $ ] is the commutator and @xmath113 is the anti - commutator .
we numerically study the steady - state density matrix @xmath114 with truncated bosonic space containing @xmath115 excitations and find that all the physical properties converge at @xmath116 .
therefore , we demonstrate our results with @xmath116 ( for comparison , in ref .
@xcite the photons and phonons were treated in a non - self - consistent mean - field approximation ) .
the expression for the current is obtained from the formal continuity equation @xmath117 , where @xmath118 .
the resulting expression for the current is @xmath119 , where @xmath120 represents the steady state average .
equivalently , the current @xmath121 can be written as the sum of photon - induced and phonon - induced current , @xmath122 .
the authors thank i. visoly - fisher for valuable discussions .
sa and yd acknowledge support from a bgu start - up grant . bz acknowledges the fondecyt grant no .
_ competing financial interests _ : the authors declare no competing financial interests
. 10 url # 1`#1`urlprefix[2]#2 [ 2][]#2 .
_ _ * * , ( ) .
http://pubs.acs.org/doi/abs/10.1021/ic0508371 . , . &
_ _ * * , ( ) . & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) .
_ _ ( , ) . .
_ _ * * , ( ) . , & .
_ _ * * , ( ) .
http://scitation.aip.org/content/aip/journal/jmp/17/5/10.1063/1.522979 .
_ et al . _ . _
_ * * , ( ) . , & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . , & .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , , & .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) .
_ _ * * , ( ) . & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . , & .
_ _ * * , ( ) . & .
_ _ * * , ( ) .
http://link.aps.org/doi/10.1103/physreve.86.061917 .
_ * * , ( ) . , & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . , & .
_ _ * * , ( ) . & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . , & .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) . , , , & .
_ _ * * , ( ) . , , , & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . .
_ _ * * , ( ) .
_ _ * * , ( ) .
_ et al . _ . _
* * , ( ) .
& _ _ ( , ) . _ _ ( , ) . & . _
_ * * , ( ) .
_ _ * * , ( ) . & .
_ _ * * , ( ) . & .
_ _ * * , ( ) . & .
_ _ * * , ( ) . & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) .
y.d . conceived the research and drafted the paper .
s.a . performed the numerical and analytical work together with b.z .
all authors discussed the results and contributed to the final version of the manuscript . | the molecular photo - cell is a single molecular donor - acceptor complex attached to electrodes and subject to external illumination . besides the obvious relevance to molecular photo - voltaics ,
the molecular photo - cell is of interest being a paradigmatic example for a system that inherently operates in out - of - equilibrium conditions and typically far from the linear response regime .
moreover , this system includes electrons , phonons and photons , and environments which induce coherent and incoherent processes , making it a challenging system to address theoretically . here , using an open quantum systems approach , we analyze the non - equilibrium transport properties and energy conversion performance of a molecular photo - cell , including for the first time both coherent and incoherent processes and treating electrons , photons , and phonons on an equal footing .
we find that both the non - equilibrium conditions and decoherence play a crucial role in determining the performance of the photovoltaic conversion and the optimal energy configuration of the molecular system .
+ + + @xmath0 correspondence to [email protected] |
in the context of astronomical masers , an enormous outburst of the h@xmath0o maser at 22 ghz ( @xmath4=1.3 cm ) in the orion kl ( kleinmann - low object ) region is one of the most enigmatic phenomena . the first h@xmath0o maser burst in orion kl was discovered in 1979 and the active phase continued until 1985 with several flare - up events @xcite .
based on the vlbi ( very long baseline interferometer ) observations , the bursting maser feature was revealed to have east - west elongation which was interpreted as a circumstellar disk @xcite .
followed by the active phase for about 6 years , the quiescent phase lasted from 1985 to 1997 . in this phase ,
a jet - like maser feature with the northwest - southeast elongation was identified @xcite .
the second burst was detected at the end of 1997 and monitoring observations were carried out with single - dish radio telescopes and vlbi @xcite .
the structure in the second burst phase was consistent with that in the adjacent quiescent phase @xcite .
the maximum flux density flared up to an order of 10@xmath5 jy in both burst phases . the line - of - sight velocity with respect to the local standard of rest ( lsr ) of the bursting maser was @xmath6 8 km s@xmath1 and was unchanged in both phases , suggesting a common origin .
two different scenarios have been proposed for this outburst ; one is related to the existence of a circumstellar disk ( e.g. * ? ? ?
* ; * ? ? ?
* ) and another is the jet / outflow driven by young stellar object ( yso ) in the orion nebula @xcite .
however , no evidence of either circumstellar disk or the jet / outflow system associated with the bursting maser has been identified , and hence , the origin of this maser burst is still unclear .
the third - time burst of the h@xmath0o maser in orion kl has started since february 2011 @xcite .
the lsr velocity of the bursting maser is again 7.8 km s@xmath1 .
the burst appears to be possibly periodic with an interval of about 13 years ( 1985 , 1998 , and 2011 ) , which could support the hypothesis of the common origin . in order to explore the mechanism of this maser burst
, we have started astrometric observations of the bursting h@xmath0o maser in orion kl with vera ( vlbi exploration of radio astrometry ) , a japanese vlbi network developed for astrometry . here
we report the result of our absolute position measurement of the bursting maser features with a submilli - arcsecond ( mas ) accuracy .
it allows us to measure the proper motions of the bursting features .
further detailed studies on the structure and time variation of the bursting maser features will be undertaken .
observations of the h@xmath0o maser ( @xmath7-@xmath8 , 22235.080 mhz ) in orion kl were carried out on 2011 march 09 , may 01 , and may 17 with vera .
the weather conditions were good at all of the four stations ( see figure 1 of * ? ? ? * ) , and the typical system noise temperature was 100 - 200 k. the maximum baseline length was 2270 km and the uniform - weighted synthesized beam size ( fwmh ) was 1.7 [email protected] mas with a position angle of 143 degrees on average .
we employed the dual beam observation mode , in which orion kl and an icrf source j054138.0 - 054149 @xcite were observed simultaneously .
the data were recorded onto magnetic tapes at a rate of 1024 mbps , providing a total bandwidth of 256 mhz in which one if channel and the rest of 15 if channels with 16 mhz bandwidth each were assigned to orion kl and icrf j054138.0 - 054149 , respectively . for the h@xmath0o maser lines , a spectral resolution was set to be 15.625 khz , corresponding to a velocity resolution of 0.21 km s@xmath1 .
a bright continuum source , icrf j053056.4 + 133155 , was observed every 80 minutes for bandpass and delay calibration .
calibration and imaging were performed using the nrao astronomical image processing system ( aips ) software package .
in addition to the present observations , we reanalyzed the archive data of the previous monitoring observations of orion kl @xcite conducted on 2006 july 15 in order to compare the results of the new observations
. details of the observation and data analysis are described in @xcite .
the h@xmath0o maser burst is detected at the peak lsr velocity channel of 7.58 km s@xmath1 in the first epoch of observation on 2011 march 09 , as shown in figure [ fig - sp ] .
the total flux density is ( [email protected])@xmath10 jy . the peak velocity is apparently shifted toward lower velocity channel compared with that in the previous burst event , 7.64 km s@xmath1 @xcite . the correlated flux , @xmath11 jy , recovers 50% of the total flux density . due to the lack of short baselines in vera ,
extended emission component would be resolved out .
two months later from the first epoch , another new velocity component at 6.95 km s@xmath1 appears in 2011 may 01 while the flux density of the 7.58 km s@xmath1 component gradually decreases .
the total flux density of the 6.95 km s@xmath1 component becomes comparable with that of the 7.58 km s@xmath1 component on 2011 may 17 .
we have successfully obtained phase - referenced images of the bursting h@xmath0o maser features .
the position , peak intensity , and size of each feature are listed in table [ tab - gauss ] .
note that due to the loss of coherence caused by the atmospheric zenith delay residual ( * ? ? ? * and references therein ) in the phase - referencing calibration , the peak flux density is degraded by a factor of 0.5 - 0.9 compared with the results obtained from the conventional self - calibration method ( columns 6 and 7 in table [ tab - gauss ] ) .
figure [ fig - prmap ] shows the phase - referenced images of the bursting maser features at the lsr velocity channels of 6.95 km s@xmath1 and 7.58 km s@xmath1 . in the first epoch in march 2011 , the 7.58 km s@xmath1 feature shows single - peaked structure while it splits into double peaks in may 2011 .
in addition , another spatially distinct feature at the lsr velocity of 6.95 km s@xmath1 appears at 12 mas north as also seen in the spectra ( figure [ fig - sp ] ) .
therefore , the current maser burst is turned out to occur at two spatially different features .
the maser features elongate along the northwest - southeast direction .
results of the two - dimensional gaussian fitting on the phase - referenced images are also summarized in table [ tab - gauss ] .
the size of 1 mas corresponds to the linear size of 0.4 au when adopting the distance to orion kl to be 420 pc @xcite .
the structures of the bursting maser features are sometimes resolved along their major axes ( i.e. those with the major axes of larger than 2 mas ) .
the 7.58 km s@xmath1 feature shows a double - peaked structure with a separation angle of 2.7 mas on 2011 may 01 and 17 as shown in figures [ fig - prmap](b ) and ( c ) .
the size of the maser feature is comparable to or slightly smaller than that found in the previous burst in 1998 ( 1.8 [email protected] mas-5.3 [email protected] mas ; * ? ? ?
the brightness temperature reaches 2.3@xmath210@xmath12 k. the velocity width of the total power spectrum of the bursting maser is [email protected] km s@xmath1 ( fwhm ) for the first epoch .
when we fit the velocity width of each feature at different velocity and position derived from the gaussian fitting of the self - calibrated images ( table [ tab - gauss ] ) , we obtain the average velocity width of [email protected] km s@xmath1 .
these velocity widths are broader than that in the previous burst , 0.40 - 0.48 km s@xmath1 @xcite . in the previous active phases ,
high - resolution observations were conducted with the vla and vlbi @xcite . with the vla ,
the positional accuracy was @xmath650 - 100 mas even in the most extended a - configuration ( e.g. * ? ? ? * ) .
although vlbi observations achieved much higher relative position accuracy of 0.05 - 0.1 mas , the absolute position information was usually lost by the self - calibration analysis .
in contrast , we have carried out phase - referencing vlbi astrometry , in which one can measure the position of the target maser source with respect to the extragalactic position reference source .
for instance , the accuracy of absolute position measurement with vera is reported to be 0.36 mas and 0.74 mas in right ascension and declination directions , respectively , for the previous h@xmath0o maser observations in orion kl @xcite .
the main source of the astrometric errors in the vera observations is the variation of the structures of maser features as well as the atmospheric zenith delay residual due to the troposphere and the positional accuracy of the reference source @xcite .
the observed maser positions are summarized in table [ tab - gauss ] .
this is the first time ever to measure the absolute positions of the bursting maser features in orion kl with the submilli - arcsecond accuracy .
the current position of the bursting maser is shifted by + 70 mas and -200 mas in right ascension and declination , respectively , with respect to that of the previous burst in 1983 ( e.g. * ? ? ?
* ) , probably because of much worse astrometric accuracy of previous observations than that of our present result .
on the other hand , the position of the bursting maser feature measured in 1998 @xcite is different by 2 in declination .
this is due to the lower positional accuracy of an order of @xmath61 in the fringe - rate mapping method in particular for declination which is typical for the interferometric observations of target sources close to the equator such as orion kl .
we have measured the absolute proper motions of three bursting maser features as shown in table [ tab - proper ] and figure [ fig - proper ] .
taking into account the annual parallax of orion kl of 2.39 mas @xcite , we obtain the absolute proper motions of @xmath64 - 9 mas yr@xmath1 or 8 - 18 km s@xmath1 toward southwest direction . at first glance , they seem inconsistent with the apparent position movement in the phase - referenced images due to the relatively large parallax approximately toward the eastward direction . although the present results contain large uncertainties due to the short monitoring period ( @xmath63 months ) , the absolute proper motions could be attributed to the outflow in the orion kl region as discussed below .
we tentatively identify the possible counterpart of the bursting maser in the quiescent phase data on 2006 july 15 at the peak lsr velocity channel of 7.37 km s@xmath1 , according to the similarity in the line - of - sight velocity , position ( within 100 mas from that in march 2011 ) and the structure ( nw - se elongation ) .
the position is also listed in table [ tab - gauss ] .
the peak intensity of the phase - referenced image is only [email protected] jy beam@xmath1 on 2006 july 15 ( see table [ tab - gauss ] ) .
the peak intensity derived from the conventional self - calibrated map is [email protected] jy beam@xmath1 .
the intensity of the bursting feature is three orders of magnitude greater than that of this feature in the quiescent phase .
judging from the similarities in peak lsr velocity , position , and structure among the detected features in 2011 , 2006 , and earlier , they could be attributed to the common origin .
however , it is unlikely that the bulk of the bursting gas clump would be identical throughout every burst and quiescent phases because the time span of observations ( @xmath135 yrs ) is longer than the typical life time of the h@xmath0o masers in orion kl ( @xmath61 yr ; see figure 1 in * ? ? ?
in fact , we can see significant time variation of velocity and spatial structure of the bursting maser features even within the present monitoring observations for only two months .
if the gas clump of the bursting maser feature would be identical to that of the possible counterpart detected in 2006 , one of the features should show an absolute proper motion of ( + 8 mas yr@xmath1 , -17 mas yr@xmath1 ) , which corresponds to the position movement by ( + 34 mas , -80 mas ) in right ascension and declination , respectively , during the 5-years separation .
we did not detect such a large proper motion in our astrometric data from march to may 2011 .
we also find two other maser features at the same velocity range @xmath68 km s@xmath1 in the quiescent phase in 2006 .
these positions are shifted by ( -1.0,-3.1 ) and ( + 1.7 , -5.4 ) from the bursting maser features .
if either of them was identical to the bursting feature , the proper motion would be unrealistically large , 1400 - 2400 km s@xmath1 , according to the position difference during the 5-year separation in our vera observations .
therefore , we conclude that no maser feature detected in 2006 is identical to those in the bursting phase in 2011 . in this section , we discuss a possible powering source of the bursting maser features .
one of the most plausible interpretations is related to the outflow from yso(s ) in the orion kl region as proposed by @xcite .
it has long been established that the h@xmath0o masers in orion kl are excited by interaction with outflow and ambient dense gas @xcite .
the most possible powering source is the radio source i which drives northeast - southwest low - velocity outflow @xcite .
the bursting maser features are located at 8 or 3400 au southwest from source i , and is coincident with the interacting region between the outflow and a dense ambient gas , which is called the orion compact ridge ( e.g. * ? ? ?
although the elongation of the bursting maser features ( pa=130 - 150 degrees ) are perpendicular to the direction to source i , the maser features could trace the shocked layer between the low - velocity outflow and the compact ridge .
the h@xmath0o maser burst may occur at the different part of the shocked layer when the episodic or possibly 13 year - periodic outflow from source i interacts with the compact ridge .
the magnitude of the proper motions of the bursting maser features are 16 - 20 km s@xmath1 pointing toward the west to southwest direction when we subtract the absolute proper motion of source i ( 6.3 mas yr@xmath1 , -4.2 mas yr@xmath1 ) recently reported by @xcite .
interestingly , they are roughly consistent with the low - velocity outflow from source i @xcite although this should be confirmed by the further vlbi monitoring observations .
nevertheless , it is still unclear why only the 8 km s@xmath1 features show such anomalous outburst .
it should be noted that the current burst event can not be explained by the local phenomenon around each feature with the scale of a few milliarcseconds ( e.g. * ? ? ?
* ; * ? ? ?
* ) because the maser burst is occurring at two distinct features separated by 12 mas and at different velocities .
one of the possibilities is that powerful outflows from source i may interact with the dense core associated with a pre - existing yso in the compact ridge region as suggested by @xcite .
what makes the bursting features always at about 8 km s@xmath1 would be that it is the rest velocity of the compact ridge ( e.g. * ? ? ?
the low - velocity outflow hitting at any fortuitous place in the compact ridge may amplify the maser emission most efficiently at the rest velocity , leading to such features . according to the simbad database ,
there are number of closer infrared , radio and x - ray sources around the bursting maser features than source i. the nearest yso candidate is the radio continuum source r , which is 0.5 southwest from the bursting maser features . a millimeter and infrared counterpart , hc438 , is also associated with the same source @xcite .
submillimeter continuum emission peaks are detected within their beam widths @xcite .
although the mass of the submillimeter core is estimated to be @xmath64 m@xmath14 @xcite , the nature of this source , whether it is a self - luminous yso or not , is still unclear .
thus , we can not definitely identify one of these sources as a powering source of the bursting maser features only based on its proximity at this moment .
we have successfully measured the absolute positions of the bursting h@xmath0o maser features in orion kl with the vlbi astrometry technique using vera .
this is the first time ever to identify the bursting maser features with better than the milli - arcsecond accuracy .
the burst event is found to be occurring at two spatially different maser features .
these features are located in the interacting region of the outflow from the radio source i and the orion compact ridge . although the spatial and velocity structure of the bursting masers are variable with time , the position and structure of the bursting masers are similar to those in the previous burst phases , implying a common origin .
we suggest that the outflow from source i or another yso in the compact ridge region could be a possible powering source of the bursting maser features .
if the current burst phase persists for one year as in the case of previous bursts , we will be able to determine the proper motions of the bursting maser features more accurately through our vlbi astrometry with vera
. they will be crucial to reveal dynamical structure of the masing region .
for instance , we will provide novel information on the vlbi images and proper motions to construct three - dimensional dynamical model of the outflow .
if the bursting masers are associated with a circumstellar disk , we will be able to derive the mass of the central protostar by directly measuring the rotating motion .
in addition , high spatial resolution observations with alma and evla of the thermal molecular lines and continuum emissions from centimeter to submillimeter wavelengths are still indispensable to identify the powering source of the bursting maser along with the velocity structure of possible outflow from it .
all of these results are essential to shed light on the emission mechanism and origin of the burst of the maser features in orion kl , as well as the driving source(s ) of complex outflows in this region .
we thank the staff of the vera stations to support our project , shuji deguchi , satoshi yamamoto , and anonymous referees for useful comments .
this work is supported in part by the graduate university for advanced studies ( sokendai ) , and has made use of the simbad database , operated at cds , strasbourg , france .
eisner , j. a. , plambeck , r. l. , carpenter , j. m. , corder , s. a. , qi , c. , & wilner , d. 2008 , apj , 683 , 304 garay , g. , moran , j. m. , & haschick , a. d. 1989 , 338 , 244 genzel , r. , reid , m. j. , moran , j. m. , & downes , d. 1981 , apj , 244 , 884 goddi , c. , humphreys , e. m. l. , greenhill , l. j. , chandler , c. j. , & matthews , l. d. 2011 , apj , 728 , 15 greenhill , l. j. , gwinn , c. r. , schwartz , c. , moran , j. m. , & diamond , p. j. 1998 , nature , 396 , 650 liu , s. -y . , girart , j. m. , remijan , a. , & snyder , l. e. 2002 , apj , 576 , 255 ma , c. et al .
1998 , aj , 116 , 516 matveyenko , l. i. , graham , d. a. , & diamond , p. j. 1988 , sov .
lett . , 14 , 468 matveyenko , l. i. , diamond , p. j. , & graham , d. a. 1998 , astron .
lett . , 24 , 623 matveyenko , l. i. , zhakharin , k. m. , diamond , p. j. , & graham , d. a. 2004 , astron . lett . , 30 , 100 hirota , t. et al .
2007 , pajs , 59 , 897 kim , m. k. et al .
2008 , pasj , 60 , 991 omodaka , t. et al . 1999 ,
pasj , 51 , 333 petrov , l. , hirota , t. , honma , m. , shibata , k. m. jike , t. , kobayashi , h. 2007 , aj , 133 , 2487 plambeck , r. l. et al .
2009 , apj , 704 , l25 shimoikura , t. , kobayashi , h. , omodaka , t. , diamond , p. j. , matveyenko , l. i. , & fujisawa , k. 2005 , apj , 634 , 459 tang , y. w. , ho , p. t. p. , koch , p. m. , & rao , r. 2010 , apj , 717 , 1262 tolmachev , a. 2011 , atel , 3177 zapata , l. a. , schmid - burgk , j. , & menten , k. m. 2011 , a&a , 529 , a24 lccrrrrrrr july 15 2006 & & 7.37 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath15 & ( [email protected])@xmath210@xmath15 & [email protected] & [email protected] & [email protected] + march 09 2011 & 1 & 6.95 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + & 2 & 7.58 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + may 01 2011 & 1 & 6.95 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + & 2 & 7.58 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + & 3 & 7.58 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + may 17 2011 & 1 & 6.95 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + & 2 & 7.58 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + & 3 & 7.58 & [email protected] & [email protected] & ( [email protected])@xmath210@xmath3 & ( [email protected])@xmath210@xmath3 & [email protected] & [email protected] & [email protected] + o maser in orion kl .
the dashed lines indicate the peak lsr velocity channels at the burst phase , 6.95 km s@xmath1 and 7.58 km s@xmath1 .
( a ) scalar - averaged cross power spectra at the burst phase on 2011 march 09 ( upper ) and at the quiescent phase on 2006 july 15 ( lower ) .
the spectra are averaged over all of the time range and all of the baselines .
( b ) total - power spectrum ( i.e. spectrum observed with each single - dish antenna average over the whole of time range and all of the four antennas of vera ) and scalar - averaged cross - power spectrum at the bursting phase on 2011 march 09 .
( c ) same as ( b ) for the second epoch on 2011 may 01 .
( d ) same as ( b ) for the third epoch on 2011 may 17 .
, width=566 ] are as small as 0.03 mas , which can not be shown in the figure .
the reference position is the same as that in figure [ fig - prmap ] .
( a ) position offset in right ascension as a function of time .
( b ) same as ( a ) for declination .
declination offset of feature 1 is shifted by -10 mas .
( c ) movement of the bursting feature 1 on the sky .
crosses represent the position and the fwhm size of the maser features listed in table [ tab - gauss ] .
( d ) same as ( c ) but for the features 2 and 3 .
, width=566 ] | in february 2011 , a burst event of the h@xmath0o maser in orion kl ( kleinmann - low object ) has started after 13-year silence .
this is the third time to detect such phenomena in orion kl , followed by those in 1979 - 1985 and 1998 .
we have carried out astrometric observations of the bursting h@xmath0o maser features in orion kl with vera ( vlbi exploration of radio astrometry ) , a japanese vlbi network dedicated for astrometry .
the total flux of the bursting feature at the lsr velocity of 7.58 km s@xmath1 reaches 4.4@xmath210@xmath3 jy in march 2011 .
the intensity of the bursting feature is three orders of magnitudes larger than that of the same velocity feature in the quiescent phase in 2006 .
two months later , another new feature appears at the lsr velocity of 6.95 km s@xmath1 in may 2011 , separated by 12 mas north of the 7.58 km s@xmath1 feature .
thus , the current burst occurs at two spatially different features .
the bursting masers are elongated along the northwest - southeast direction as reported in the previous burst in 1998 .
we determine the absolute positions of the bursting features for the first time ever with a submilli - arcsecond ( mas ) accuracy .
their positions are coincident with the shocked molecular gas called the orion compact ridge .
we tentatively detect the absolute proper motions of the bursting features toward southwest direction .
it is most likely that the outflow from the radio source i or another young stellar object interacting with the compact ridge is a possible origin of the h@xmath0o maser burst . |
the normal and conormal cycles of a reasonably well behaved subset @xmath1 of an oriented euclidean space @xmath2 of dimension @xmath3 are currents that encode rather subtle topological and geometric features of the set .
the normal cycle @xmath4 is a legendrian cycle contained in the unit sphere bundle @xmath5 associated to the tangent bundle @xmath6 , while the conormal cycle is a lagrangian cycle @xmath7 in the cotangent bundle @xmath8 .
the two objects completely determine each other in a canonical fashion .
their precise definitions are rather sophisticated in general , but they can be easily described in many concrete examples .
for example , if @xmath1 is a submanifold of @xmath2 , then @xmath4 can be identified with the integration current defined by the total space of the unit sphere bundle associated to the normal bundle of the embedding @xmath9 , while @xmath7 can be identified with the current of integration defined by the total space of the conormal bundle of the embedding @xmath9 .
the normal cycle @xmath4 is intimately related to h. weyl s celebrated tube formula @xcite .
more precisely , there exist canonical forms ( see @xcite ) @xmath10 such that , for any compact submanifold @xmath9 , the integrals @xmath11 can be expressed as integrals over @xmath1 of universal polynomials in the curvature of the induced metric on @xmath1 .
for example , if @xmath12 , then @xmath13 is the @xmath14-dimensional volume of @xmath1 , and @xmath15 coincides ( up to a universal multiplicative constant ) with the integral over @xmath1 of the scalar curvature .
the quantity @xmath16 is the euler characteristic of @xmath1 which , according to the gauss - bonnet theorem , can be expressed as the integral of a universal polynomial in the curvature of of @xmath1 .
the quantities @xmath17 are known as _ curvature measures_. they are the key ingredients in the tube formula that states that for any sufficiently small @xmath18 the volume of a tube of radius @xmath19 around a compact submanifold @xmath9 of dimension @xmath14 is ( see @xcite ) @xmath20 where @xmath21 denotes the volume of the unit @xmath22-dimensional ball .
if @xmath1 happens to be a bounded domain in @xmath2 with sufficiently regular boundary @xmath23 , then we have a unit outer normal vector field @xmath24 and the normal cycle @xmath4 the integration current defined by the graph of the above map ; see example [ ex : reg ] . in this case the integrals @xmath25 can be expressed as integrals over @xmath23 of universal polynomials in the second fundamental form of the hypersurface @xmath23 , and they are involved in a tube formula similar to ( [ eq : tube ] ) , @xcite .
if additionally @xmath1 happens to be convex , then the curvature measures @xmath17 coincide with the _ quermassintegrale _ constructed by h. minkowski , @xcite . in the groundbreaking work
@xcite h. federer has explained how to associate curvature measures to subsets of @xmath2 of positive reach .
this class of subsets contain as subclasses the smooth submanifolds of @xmath2 , the bounded domains with smooth boundary and the convex bodies in @xmath2 , and in these cases federer s curvature measures specialize to the curvature measures described above .
as explained in @xcite , the _ quermasseintegrale _ can be extended in a canonical fashion to finitely additive measures ( valuations ) defined on the collection of polyconvex subsets of @xmath2 , i.e. , sets that are finite unions of convex bodies .
in particular , the quantity @xmath17 is well defined for any compact @xmath26 subset of @xmath2 . for most @xmath26 sets the gauss map ( [ eq : g1 ] ) is not defined and the above definition of @xmath4 is meaningless . in the mid 1980s , j. cheeger ,
s. mller and r .schrader @xcite have associated to an arbitrary compact @xmath26 subset @xmath27 a legendrian cycle @xmath28 contained in @xmath5 such that @xmath29 their elementary construction is very intuitive and is based on elementary morse theory on @xmath26-spaces . moreover
, the correspondence @xmath30 from the collection of compact @xmath26 subsets of @xmath2 to the abelian group of legendrian cycles in @xmath5 is a finitely additive measure , i.e. , @xmath31 for any @xmath26 sets @xmath32 .
the paper @xcite includes the first formal definition of the normal cycle . roughly speaking , the normal cycle @xmath4 is designed to be an ingenious catalogue of the morse theoretic behavior of the restrictions to @xmath1 of `` typical '' linear functions on @xmath2 . a few years after @xcite , m. kashiwara and p. schapira
@xcite have shown how to associate a normal cycle to any bounded subanalytic subset of @xmath2 .
although the morse theoretic point of view is still in the background , their approach is sheaf theoretic and geared towards topological applications .
their proof is quite sophisticated as it relies on highly nontrivial results about the derived categories of sheaves .
j. schrmann @xcite has proposed a simpler sheaf theoretic construction of the normal cycle , but this too requires a good familiarity with stratified spaces the basic operations in the derived category of sheaves . almost immediately following the work of kashiwara and schapira , j. fu @xcite gave another construction of the normal cycle of a subanalytic set using methods of geometric measure theory .
his proof is technically very demanding , and the complete details are spread over several papers . very recently , a. berning @xcite has proposed a very ingenious and elegant elementary construction of the normal cycle of a subanalytic set using the recent advances in @xmath0-minimal topology and basic facts about currents . unfortunately there is a flaw in a key existence result , ( * ? ? ?
* lemma 6.4 ) ; see remark [ rem : glitch](a ) for more detail;s .
the present paper grew out of our attempts to fix that flaw .
the main goal of this paper is to describe a very short complete proof of the existence of the normal cycle of a subanalytic set by relying on techniques and ideas from @xmath0-minimal topology .
we rely on several fundamental facts of geometric measure theory that we survey in appendix [ s : c ] .
the construction has a morse theoretic flavor , and it is based on two key principles . * _ an uniqueness result closely related to the uniqueness results of j. fu ( * ? ? ?
_ loosely speaking , this uniqueness result states that there exists a unique legendrian cycle in @xmath33 that catalogs in a certain explicit fashion ( see remark [ rem : normc](a ) ) the morse theoretic properties of the restrictions to @xmath1 of generic linear functions on @xmath2 . when it exists , this unique cycle is the normal cycle @xmath4 of @xmath1 . * _ an approximation process pioneered by j. fu @xcite . _
more precisely , we show that for any compact subanalytic set @xmath1 we can find a family of bounded domains @xmath34 with @xmath35-boundaries such that @xmath36 and the normal cycles @xmath37 converge in the sense of currents to a current satisfying the requirements of the uniqueness theorem .
thus , the limit cycle must be the normal cycle of @xmath1 .
the resulting correspondence @xmath38 , @xmath1 bounded subanalytic set , satisfies the inclusion - exclusion principle ( [ eq : inc - exc ] ) and for @xmath26 sets it coincides with the normal cycle constructed in @xcite .
the concrete implementation of our strategy was possible only due to the recent advances in @xmath0-minimal topology .
here is a more technical description of our main results .
let @xmath2 be an oriented real euclidean vector space of dimension @xmath3 .
denote by @xmath39 its dual , and by @xmath40 the unit sphere in @xmath39 .
we identify the cotangent bundle @xmath8 with the product @xmath41 .
we have two canonical projections @xmath42 let @xmath43 denote the canonical pairing @xmath44 the euclidean metric @xmath45 on @xmath2 defines isometries ( the classical lowering / raising the indices operations ) @xmath46 @xmath47 let @xmath48 denote the canonical @xmath49-form on the cotangent bundle . more explicitly , if @xmath50 are euclidean coordinates on @xmath2 , and @xmath51 denote the induced euclidean coordinates on @xmath39 , then @xmath52 we denote by @xmath53 the associated symplectic form @xmath54 in this paper we will work extensively with subanalytic objects .
our subanalytic sets are the sets in the @xmath0-minimal structure @xmath55 as defined appendix [ s : b ] to which we refer for more details .
we will work with special classes of currents .
for the reader s convenience we have gathered in appendix [ s : c ] the basic notations and facts involving currents that we use throughout this paper . for any closed subanalytic subset @xmath56
we denote by @xmath57 the abelian group of subanalytic , @xmath58-dimensional currents with support on @xmath1 ; see appendix [ s : c ] for details .
more precisely , @xmath57 is the abelian subgroup of @xmath59 spanned by the currents of integration over oriented @xmath58-dimensional subanalytic submanifolds contained in @xmath1 .
if @xmath60 , and @xmath61 , we denote by @xmath62 the @xmath22-slice of @xmath63 over @xmath64 , @xmath65 as explained in appendix [ s : c ] , the slice @xmath62 exists for all @xmath64 outside a codimension @xmath49 subanalytic subset of @xmath40 and it is supported on the fiber @xmath66 .
more precisely , @xmath62 is well defined if the fiber @xmath67 has the expected dimension , @xmath68 .
if @xmath63 is the current of integration along an oriented @xmath58-dimensional manifold , then for generic @xmath64 the slice @xmath62 is the current of integration along the fiber @xmath69 equipped with a canonical orientation . in general
, the slice gives a precise meaning as a current to the intersection of @xmath63 with the fiber @xmath70 , provided that this intersection has the `` correct '' dimension . if @xmath27 is a compact subanalytic set , @xmath61 , and @xmath71 we set @xmath72 where @xmath73 denotes the euler characteristic of a topological space .
if @xmath74 we set @xmath75 .
the integer @xmath76 can be interpreted as a morse index of the function @xmath77 at @xmath78 ; see @xcite or appendix [ s : a ] . for generic @xmath79 , we have @xmath75 , for all but finitely many points @xmath80 . the first goal of this paper
is to give a very short proof of the following uniqueness result closely related to the uniqueness result of j. fu , ( * ? ? ?
let @xmath1 be a compact subanalytic subset of @xmath2 .
then there exists at most one subanalytic current @xmath81 satisfying the following conditions . 1 . the current @xmath82 is a cycle , i.e. , @xmath83 .
2 . the current @xmath82 has compact support .
the current @xmath82 is legendrian , i.e. , @xmath84 4 . for any smooth function @xmath85
we have @xmath86 [ th : fu1 ]
\(a ) using ( * ? ? ?
4.3.2.(1 ) ) we deduce that the equality ( iv ) is equivalent with the condition @xmath87 where @xmath88 denotes the canonical @xmath89-dimensional current determined by the point @xmath90 .
the points @xmath78 for which @xmath91 should be viewed as critical points of the function @xmath77 ; see @xcite or appendix [ s : a ] . thus , the slice @xmath92 records both the collection of critical points of @xmath93 and their morse indices .
\(b ) our sign conventions are different from the ones used in @xcite , but they coincide with the conventions in @xcite .
[ rem : normc ] when a cycle as in theorem [ th : fu1 ] exists , it is called the _ normal cycle _ of @xmath1 , and we will denote it by @xmath4 . in the remarkable paper @xcite ,
j. fu proved the following result .
every compact subanalytic set @xmath27 has a normal cycle @xmath4.[th : fu2 ] the _ conormal cycle _ @xmath94 of @xmath1 constructed by kashiwara and schapira , @xcite can be obtained from normal cycle @xmath4 using a coning procedure described explicitly in ( [ eq : cone ] ) .
the equality ( [ tag : ast ] ) is then a special of the micro - local index theorem ( * ? ? ?
9.5.6 ) .
the second goal of this paper is to show that theorem [ th : fu2 ] is a consequence of theorem [ th : fu1 ] .
the proof takes full advantage of the subanalytic context of the problem which prohibits many of the possible pathologies in geometric measure theory .
in particular , the geometric cores of the arguments are much more transparent in this context .
\(a ) a sheaf - theoretic approach to the existence of normal cycles based on a conceptually similar approximation method can be found in @xcite .
the concept of limit of currents is replaced by the concept of specialization , while the uniqueness theorem is replaced by the injectivity of a certain morphism in borel - moore homology , ( * ? ? ?
* eq . ( 5.19 ) ) .
this injectivity is ultimately based on a special property of verdier stratifications , ( * ? ? ?
8.3.23 ) .
\(b ) we want to point out a key technical difference between the approach in this paper and the approach in @xcite . the uniqueness theorem ( * ? ? ?
3.2 ) is formulated in terms of an integral cycle @xmath95 defined for _ any _ legendrian cycle @xmath96 and for almost all @xmath97 ; see ( * ? ? ?
the construction of the cycle @xmath98 is quite involved , but it has a nice payoff because it shows that @xmath98 depends continuously on @xmath99 .
this fact is very useful in approximation problems .
our approach skips the construction of @xmath98 , but we have to pay a small price since the proofs of our convergence results are a bit more involved .
\(c ) we believe that the arguments used in this paper can yield an existence result for normal cycles of compact sets in an @xmath0-minimal category .
we chose to work in the subanalytic category only because of a lack of adequate references for a theory of slicing of @xmath0-minimal currents .
hardt s subanalytic work @xcite ought to extend with minor changes to an @xmath0-minimal context .
[ rem : long ] here is a brief outline of the paper .
we prove theorem [ th : fu1 ] in section [ s : 3 ] relying on an @xmath0-minimal implementation of the strategy in @xcite that affords considerable simplifications .
we construct the normal cycle in section [ s : 4 ] via an approximation method .
section [ s : 5 ] describes how a combination of facts proved in this paper and in @xcite leads to an alternate proof of the existence part of ( * ? ? ?
6.2 ) . in section
[ s : app ] we give an alternate proof to the main convergence theorem in @xcite .
we have included two appendices [ s : b ] , [ s : c ] that survey basic facts about subanalytic sets and currents used throughout the paper . in appendix
[ s : a ] we present short @xmath0-minimal proofs of some basic facts of singular morse theory that we use in the paper and in our opinion are not widely known . in particular , we have included a very short proof of kashiwara s non - characteristic deformation lemma in an @xmath0-minimal setting .
we will prove theorem [ th : fu1 ] following a strategy that is inspired from @xcite .
we first give a direct and very short proof of a subanalytic version of the general uniqueness theorem ( * ? ? ?
1.1 ) . then , arguing as in the proof of ( * ? ? ?
4.1 ) , we show that this theorem implies theorem [ th : fu1 ] .
suppose @xmath100 is a subanalytic current @xmath3-dimensional current satisfying the following conditions . 1 .
the current @xmath63 is a cycle , @xmath101 , 2 .
the current @xmath63 is lagrangian , i.e. , @xmath102 .
the current @xmath63 is conical , i.e. , @xmath103 where @xmath104 is the rescaling @xmath105 .
if @xmath106 denotes the support of @xmath63 , then the induced map @xmath107 is proper .
the set @xmath108 is a conical subanalytic subset of @xmath39 of dimension @xmath109 .
then @xmath110 .
[ th : fu11 ] we argue by contradiction .
let @xmath111 so that @xmath112 .
if @xmath113 , we deduce that @xmath114 .
the _ cycle _ @xmath63 is subanalytic , and of dimension @xmath115 so that @xmath116 $ ] for some integer @xmath58 . on the other hand , conditions ( iii ) and ( iv )
imply that @xmath117 is compact , so that @xmath118 , and therefore @xmath110 by the constancy theorem .
suppose @xmath119 .
then we can find a @xmath14-dimensional subspace @xmath120 so that if @xmath121 denotes the orthogonal projection onto @xmath122 , then @xmath123 is an @xmath14-dimensional tame subset of @xmath122 .
moreover , most of the fibers of the induced map @xmath124 are zero - dimensional . set @xmath125 ; see figure [ fig : 3 ] . from the properties of subanalytic sets , or more generally , sets in an @xmath0-minimal category , @xcite or @xcite
, we deduce that there exists a subanalytic subset @xmath126 whose complement has dimension @xmath127 such that following hold .
* @xmath128 is a @xmath129-manifold . *
the induced map @xmath130 is a locally ( definably ) trivial fibration with @xmath131-dimensional fibers .
set @xmath132 , @xmath133 . *
if @xmath134 and near @xmath135 the set @xmath136 is a @xmath129 manifold , then the differential of @xmath137 at @xmath135 is a surjection @xmath138 . * the set @xmath139 is a @xmath129-manifold and the induced map @xmath140 is a submersion .
* for any @xmath133 the set @xmath141 is finite .
is bounded from above by a constant independent of @xmath142 . ] in particular , for any @xmath133 the fiber @xmath143 is contained in the finite union of planes @xmath144 . * for any @xmath133 the slice @xmath145 is well defined .
it is an @xmath131-cycle with support @xmath146 .
there exists a subanalytic set @xmath147 of dimension @xmath148 such any @xmath149 in @xmath150 belongs both to the @xmath129-locus of @xmath136 , and to the @xmath129-locus of the fiber @xmath151 that contains @xmath135 .
consider an arbitrary point @xmath152 and then choose a vector @xmath153 tangent at @xmath135 to the fiber @xmath143 , @xmath154 .
the condition ( c5 ) shows that @xmath143 is contained in the finite union of planes @xmath144 .
this implies that @xmath155 . using the fact that @xmath63 is a lagrangian current , i.e. , @xmath102
, we deduce that for any @xmath156 we have @xmath157 if we denote by @xmath158 the covector dual to @xmath159 we deduce from the above that @xmath160 is perpendicular to @xmath161 .
this is an @xmath14-dimensional subspace of @xmath39 . at the point
@xmath149 the linear map @xmath162 must be a surjection .
thus @xmath163 .
we deduce that the tangent plane to @xmath143 at @xmath164 coincides with the plane @xmath165 , @xmath166 as we already know , @xmath143 is contained in the finite union of planes @xmath144 .
the above remarks show that for any @xmath167 , and any @xmath129-point @xmath135 of the component of @xmath143 contained in @xmath168 the tangent space @xmath169 coincides with @xmath165 .
in other words the gauss map of the @xmath129-locus of @xmath143 has finite range @xmath170 .
this shows that the support of the slice @xmath145 is contained in a finite number of @xmath131-dimensional planes .
the slice @xmath171 is a @xmath131-dimensional cycle with compact support .
the constancy theorem shows that it must be trivial .
this implies that @xmath172 .
this contradicts ( c2 ) and thus completes the proof of theorem [ th : fu11 ] .
if we denote by @xmath173 the euclidean volume form on @xmath39 , we see that for a _ subanalytic _
current @xmath174 the condition ( v ) of theorem [ th : fu11 ] is equivalent to the condition @xmath175 employed in ( * ? ? ?
indeed , clearly ( v ) @xmath176 ( [ tag : v ] ) . the implication ( [ tag : v ] )
@xmath176 ( v ) follows from sard s theorem and the fact that outside a subanalytic subset of dimension @xmath177 the support @xmath106 can be identified with a real analytic manifold .
[ rem : v ] * proof of theorem [ th : fu1 ] .
* suppose @xmath178 are two subanalytic cycles satisfying the condition ( i),(ii ) , ( iii ) , ( iv ) of the theorem .
then the subanalytic cycles @xmath179 , @xmath180 , have compact support .
since the reduced homology of @xmath2 is trivial we deduce from @xcite that there exist subanalytic currents @xmath181 such that @xmath182 the constancy theorem ( theorem [ th : const ] ) shows that the currents @xmath183 are uniquely determined by the above equality .
let @xmath184 denote the zero section of @xmath8 , i.e. , @xmath185 , @xmath186 .
consider the rescaling map @xmath187 and , as in ( * ? ? ?
4.8 ) , we form the currents @xmath188 as explained in ( * ? ? ?
4.8 ) , the current @xmath189 satisfies the assumptions ( i)(iv ) of theorem [ th : fu11 ] and also the condition ( [ tag : v ] ) .
using the remark [ rem : v ] and theorem [ th : fu11 ] we conclude that @xmath110 .
let us observe that the condition ( v ) in theorem [ th : fu11 ] is equivalent to the condition that the slices @xmath62 are trivial for almost all @xmath64 in @xmath39.[rem : ber - unique ]
we want to show that theorem [ th : fu1 ] @xmath176 theorem [ th : fu2 ] .
we start by describing a simple well known class of compact subanalytic sets that admit normal cycles .
suppose @xmath1 is a compact subanalytic domain in @xmath2 with @xmath129-boundary .
consider the oriented gauss map @xmath190 that associates to each @xmath191 the unit covector @xmath192 which is dual to the unit _
outer _ normal vector at @xmath78 .
we get an embedding @xmath193 whose image coincides with the graph of the gauss map .
denote by @xmath194 $ ] the integration current defined by @xmath23 equipped with the induced boundary orientation .
then the cycle @xmath195 $ ] supported by the graph of the gauss map is the normal cycle of @xmath1 .
indeed , it is obviously a subanalytic cycle , and @xmath196 $ ] is compact since @xmath1 is compact .
the legendrian condition is simply a rephrasing of the fact that for any @xmath191 the covector @xmath192 is conormal to @xmath197 . to verify ( iv ) we first observe that @xmath198 for @xmath191 denote by @xmath199 the second fundamental form of @xmath23 at @xmath78 .
the equality ( [ eq : morse - sl ] ) is a consequence of the following facts .
fix a regular value @xmath64 of @xmath200 , and a point @xmath201 . then , * the local degree of @xmath200 at @xmath78 is equal to the sign of the determinant of @xmath202 ; * @xmath203 , where @xmath204 is the number of positive eigenvalues of @xmath199 .
for a proof of ( f1 ) we refer to @xcite . to prove ( f2 ) we can assume that @xmath205 , and near @xmath78 the hypersurface @xmath23 is the graph of a quadratic form @xmath206 while the interior of @xmath1 is , locally , the region below the graph
. then @xmath207 and ( f2 ) now follows from standard facts of morse theory .
[ ex : reg ] suppose now that @xmath1 is a compact subanalytic set .
fix an integer @xmath208 . then there exists a subanalytic @xmath209-function @xmath210 such that @xmath211 ; see ( * ? ? ?
c.11 ) . set @xmath212 so that @xmath213 is @xmath209 , nonnegative , subanalytic and @xmath214 .
the following result should be obvious .
fix @xmath215 sufficiently large so that @xmath1 is contained in the open ball @xmath216
. then there exists @xmath217 such that , for any @xmath218 the level set @xmath219 does not intersect the sphere @xmath220 .
[ lemm : proper ] denote by @xmath221 the restriction of @xmath213 to the ball @xmath216 in the above lemma .
the set @xmath222 of critical values of @xmath221 is a subanalytic @xmath89-dimensional subanalytic subset of @xmath223 and thus it consists of a finite number of points .
fix @xmath224 so that the interval @xmath225 consists only of regular values of @xmath221 .
then for any @xmath226 the set @xmath227 is a compact subanalytic domain with @xmath129-boundary .
therefore it has a normal cycle @xmath228 .
the collection @xmath229 is an increasing subanalytic family of compact subanalytic sets such that @xmath230 the collection @xmath231 is a definable collection of compact , subanalytic manifolds of class @xmath209 .
we deduce that their volumes are bounded from above .
this shows that the family of currents @xmath232 is bounded in the mass norm .
the compactness theorem ( * ? ? ?
4.2.17 ) implies that there exists a subsequence @xmath233 such that the currents @xmath234 converge in the flat metric to a integral legendrian cycle @xmath235 . to prove that @xmath99 is a _ subanalytic _
current it suffices to show that its support is contained in a subanalytic set of dimension @xmath177 . to see this we consider the subanalytic set @xmath236
then @xmath237 and @xmath238 is a subanalytic set of dimension @xmath148 containing @xmath239 .
we want to show that @xmath99 satisfies ( [ eq : morse - sl ] ) .
since @xmath240 it suffices to show that @xmath241 we have @xmath242 to do this we will need to use some of the topological facts and terminology presented in appendix [ s : a ] . observe that @xmath243 is a @xmath209-manifold of dimension @xmath244 , and its projection onto @xmath2 is @xmath245 .
this shows that the definable set @xmath246 has dimension @xmath3 .
hence , there exists a subanalytic set @xmath247 such that @xmath248 and for any @xmath249 the set @xmath250 has dimension @xmath49 .
this implies that for any @xmath251 there exists @xmath252 such that , and any @xmath253 we have @xmath254 thus , for @xmath251 and @xmath255 the slice @xmath256 is well defined .
let us point out that the set of homological critical points of @xmath257 is contained in the projection on @xmath2 of @xmath258 .
in particular , if @xmath251 , and @xmath259 , this set of homological critical points is finite . as explained in appendix [
s : a ] , there exists a subanalytic subset @xmath260 such that @xmath261 and for any @xmath262 the set of homological critical points of @xmath77 is finite . set @xmath263 so that @xmath264 .
we have the following fundamental equality @xmath265 the choice @xmath266 guarantees that for any @xmath259 the above sum consists of finitely many terms .
let us first show that ( [ eq : limit ] ) implies ( [ eq : morse - sl1 ] ) .
we will achieve this using the lebesgue dominated convergence theorem so it suffices to show that there exists a constant @xmath267 so that for any @xmath268 there exists a subanalytic set @xmath269 such that @xmath270 and @xmath271 for every @xmath226 we denote by @xmath269 the set @xmath272 we deduce that @xmath273 .
note that for any @xmath274 the set @xmath275 coincides with @xmath258 and @xmath276 the set @xmath277 is definable , and for any @xmath278 the set @xmath258 is finite .
we have thus obtained a definable collection of finite sets @xmath279 and we conclude that there exists an integer @xmath280 such that @xmath281 let @xmath282 . then , for any @xmath283 we have @xmath284 this proves ( [ eq : bound ] ) .
let us prove ( [ eq : limit ] ) . fix @xmath285 and @xmath286 .
define @xmath287 using the terminology in appendix [ s : a ] , we see that @xmath288 is the set of numerically critical points of the function @xmath289 on @xmath290 .
consider the @xmath291-definable set @xmath292 and denote by @xmath293 the natural projection @xmath294 .
then there exists @xmath295 such that over the interval @xmath296 the map @xmath293 is a locally trivial fibration .
its fibers will consists of the same number @xmath297 of points .
thus , we can find @xmath291-definable continuous maps @xmath298 such that @xmath299 set @xmath300 .
the above limits exist since the functions @xmath301 are definable and @xmath1 is compact .
for @xmath71 we define @xmath302 the equality ( [ eq : limit ] ) is then a consequence of the following result .
@xmath303 in particular , if @xmath304 , then the above limit is zero .
[ lemma : limit ] set @xmath305 . for @xmath18 and @xmath306
we set @xmath307 choose @xmath18 and @xmath308 sufficiently small such that the following hold . *
the topological type of @xmath309 is independent of @xmath310 $ ] . *
the set @xmath311 is contractible for any @xmath312 , and the set @xmath313 contains no homological critical points of @xmath77 . *
@xmath314 and @xmath315 @xmath316 , @xmath317 , @xmath318 .
fix @xmath319 so that the interval @xmath320 contains no critical values of @xmath321 .
we deduce from ( [ eq : top - drop ] ) and ( [ eq : vary ] ) @xmath322 the only ( numerically ) critical points of @xmath289 on @xmath323 are @xmath324 .
choose @xmath325 such that @xmath326 using again ( [ eq : top - drop ] ) and ( [ eq : vary ] ) for the function @xmath289 on on @xmath323 , @xmath327 , we deduce @xmath328 the equality ( [ eq : limit1 ] ) is obtained by observing that @xmath329 the equality ( [ eq : limit ] ) is immediate .
set @xmath330 .
then @xmath331 @xmath332 the equality ( [ eq : limit1 ] ) implies that the last term goes to zero as @xmath333 .
this proves ( [ eq : limit ] ) and ( [ eq : morse - sl1 ] ) . from theorem
[ th : fu1 ] we deduce that @xmath99 must be the normal cycle of @xmath1 and that @xmath334 converges in the flat metric to @xmath335 as @xmath333 .
let us explain how the results proved so far lead to an alternate approach to the existence statement in ( * ? ? ?
we need to introduce some terminology .
a function @xmath336 is called constructible if its range is finite and for any @xmath337 the level set @xmath338 is subanalytic .
if we let @xmath339 denote the characteristic function of a set @xmath340 , then for any constructible function @xmath341 we can write @xmath342 we denote by @xmath343 the abelian group of constructible functions and by @xmath344 the abelian group of constructible functions with compact support . the triangulability theorem ( * ? ? ?
8.2.9 ) implies that this group is generated by the characteristic functions of compact subanalytic sets .
observe that for any compact subanalytic sets @xmath345 and any @xmath346 we have @xmath347 for almost all @xmath61 . in view of the uniqueness theorem
this implies that @xmath348 from groemer s extension theorem ( * ? ? ?
2.2.2 ) we deduce that the correspondence @xmath349 extends to a group morphism @xmath350 the normal cycle @xmath351 of a compactly supported constructible function @xmath341 is a compactly supported , subanalytic legendrian cycle . in the remainder of this section we want to compare this construction of the normal cycle of a constructible function to the one proposed in @xcite .
we will need to use the @xmath0-minimal euler characteristic function @xmath352 as defined in @xmath0-minimal topology ; see appendix [ s : b ] and @xcite .
we will denote by @xmath353 the usual topological euler characteristic .
these two notions coincide on compact subanalytic sets , but they could be quite different on non - compact ones .
for example , if @xmath354 is an _ open _ @xmath58-dimensional ball , then @xmath355 more generally , if @xmath1 is locally compact , then @xmath356 can be identified with the euler characteristic of the borel - moore homology of @xmath1 .
each of these two notions of euler characteristic has its own advantages .
the @xmath0-minimal euler characteristic @xmath352 is _ fully additive _ additive , i.e. , for subanalytic sets @xmath1 and @xmath357 we have @xmath358 on the other hand , it is not a homotopy invariant as its topological cousin @xmath353 . the additivity condition ( [ eq : add ] ) and the groemer extension theorem implies that @xmath352 defines a linear map @xmath359 , called the _ integral with respect to the euler characteristic _ , and denoted by @xmath360 .
suppose @xmath1 is a compact subanalytic set , and @xmath61 is such that the induced function @xmath77 is a nice in the sense of appendix [ s : a ] to which we refer for notations . using ( [ eq : top - drop ] ) we deduce that for every @xmath361 , and every sufficiently small @xmath268 we have @xmath362 @xmath363 @xmath364 for @xmath268 sufficiently small the induced map @xmath365 is a locally trivial fibration .
we denote its fiber by @xmath366 . since the @xmath0-minimal euler characteristic of an open interval is @xmath367 we deduce @xmath368
we conclude that @xmath369 following @xcite , we associate to any constructible function @xmath370 and any @xmath371 the integral @xmath89-dimensional current @xmath372 given by @xmath373 where @xmath374 is the integer @xmath375 let us point out that when @xmath341 is the characteristic function of a bounded subanalytic set @xmath1 , then @xmath376 if we denote by @xmath377 the abelian group of integral @xmath89-dimensional currents on @xmath223 , then we can organize the above construction as a _ jump _ map @xmath378 this map is homogeneous and , as proved is the support map as defined in @xcite , then @xmath379 .
the `` culprit '' for this discrepancy is the outer - normal convention used in example [ ex : reg ] . ] in @xcite , it is also lipschitz with respect to the flat metric on @xmath377 .
it is also constructible in the sense that the function @xmath380 is constructible .
let us observe that if @xmath381 , where @xmath1 compact subanalytic set with normal cycle @xmath4 , then the equality ( [ eq : jump ] ) implies that for almost any @xmath79 we have @xmath382 where @xmath383 is the bilinear map @xmath384 .
we denote by @xmath385 the abelian group of constructible , homogeneous , lipschitz continuous maps @xmath386 , so that the jump construction gives a morphism of abelian groups @xmath387 according to ( * ? ? ?
3.1 ) , this morphism is an _ isomorphism _ of abelian groups .
the injectivity of @xmath388 is a consequence of the injectivity of the `` motivic '' radon transform of p. schapira @xcite .
the surjectivity is more subtle , and we refer to @xcite for details .
denote by @xmath389 the abelian group of subanalytic , conical lagragian cycles @xmath390 , such that the restriction to @xmath391 of the projection @xmath392 is proper . to any compact subanalytic set @xmath1
we denote by @xmath393 the conormal cycle constructed as in ( [ eq : cone ] ) , @xmath394 where @xmath395 is the zero section , and @xmath396 is the multiplication map @xmath397 the resulting correspondence @xmath398 satisfies the inclusion - exclusion identity , i.e. , @xmath399 for any compact subanalytic sets @xmath400 . invoking the groemer extension theorem again
we obtain a group morphism @xmath401 . given @xmath402 , the slice @xmath403 is defined for almost every @xmath404 .
in @xcite it is shown that the almost everywhere defined function @xmath405 is the restriction of a function @xmath406 .
we have thus obtained a morphism of abelian groups @xmath407 we have the following fundamental result , @xcite .
the morphism @xmath408 is injective .
more precisely , if @xmath409 and @xmath410 then @xmath110 .
[ th : ber - unique ] for the reader s convenience we decided to include a proof of this result .
what follows is a slightly different incarnation of the strategy employed in @xcite .
we will show that ( [ tag : vp ] ) implies that @xmath411 for almost any @xmath64 .
we then conclude using remark [ rem : ber - unique ] and theorem [ th : fu11 ] .
first of all , let us observe that since @xmath63 is conical and @xmath412 , then @xmath413 , where we recall that @xmath414 is the canonical @xmath49-form and @xmath415 is the canonical symplectic form .
for any subset @xmath416 we set @xmath417 since @xmath63 is conical and subanalytic we can find a definable triangulation @xmath418 of @xmath40 such that for any top dimensional ( open ) face @xmath408 of there exists * a finite collection @xmath419 of subanalytic , @xmath420-maps @xmath421 and * a multiplicity map @xmath422 with the following properties 1 .
for any @xmath423 , the map @xmath424 is injective .
2 . any @xmath425 is homogeneous of degree @xmath89 .
3 . for any @xmath426
the slice @xmath62 is well defined and it is described by @xmath427 then @xmath428 we fix a top dimensional face @xmath408 and we want to prove that @xmath429 , @xmath430 .
we argue by contradiction so we assume that there exists a function @xmath431 such that @xmath432 .
then @xmath433 and we deduce from ( [ eq : zero ] ) that for any @xmath423 the set @xmath434 is non - empty . the collection @xmath435 is a definable collection of subsets of the finite set @xmath436 . from the definable selection theorem
we deduce that there exists a definable map @xmath437 such that @xmath438 , @xmath439 .
since @xmath440 is definable , there exits a definable set @xmath441 with the following properties .
* @xmath442 .
* @xmath443 is closed in @xmath444 . * for any connected component @xmath445 of @xmath446 the resulting map @xmath447 .
is constant .
we will refer to the connected components of @xmath446 as _
let @xmath448 , and denote by @xmath449 the chamber containing @xmath450 .
let @xmath451 be the constant value of @xmath452 .
set @xmath453 and denote by @xmath454 the dual covector . since @xmath449 is open we deduce that @xmath455 if @xmath456 is sufficiently small .
we deduce @xmath457 derivating the above equality at @xmath458 we deduce @xmath459 the @xmath420-paths @xmath460 , @xmath461 are contained in the @xmath420-locus of the support of the current @xmath63 . since @xmath413
we deduce @xmath462 using this in ( [ eq : punch ] ) , and observing that @xmath463 , we conclude that @xmath464 .
hence @xmath465 .
this contradicts the injectivity assumption ( i ) .
theorem [ th : fu1 ] and ( [ eq : jump1 ] ) imply that we have a commutative diagram @xmath466{e , t}{{\boldsymbol{s}}}\node[2]{{\euscript{l}}_{{\boldsymbol{v}}}}\arrow{sw , b}{{{\sigma}}}\\
\node[2]{{\boldsymbol{i}}_0({{\mathbb r } } ) } \end{diagram}\ ] ] since @xmath388 is an isomorphism , we deduce that the morphism @xmath408 must also be surjective .
\(a ) the surjectivity of @xmath408 is also claimed in @xcite .
however , the proof is incorrect due to a glitch in the proof of ( * ? ? ?
* lemma 6.4 ) . that lemma is an existence result having to do with a certain cycle @xmath63 with support contained in the closure of a cell @xmath467 , and such that @xmath468 .
loosely , the lemma states that the relative cycle determined by @xmath63 in the homology of the pair @xmath469 is trivial .
the bordism proving the vanishing of this homology class is a cone on @xmath63 constructed using a certain homotopy . that homotopy is defined only on @xmath467 , not on @xmath470 , so the homotopy formula stated as in ( * ? ? ?
4.3 ) can not be applied .
\(b ) from the above discussion it follows that the morphism @xmath471 is also bijective .
its inverse can be described in terms of the _ local euler obstruction _ , @xcite , @xcite .
[ rem : glitch ]
we have now developed enough technology to provide an alternate proof to the convergence theorem in @xcite .
suppose that @xmath472 and @xmath1 are compact subanalytic subsets of @xmath2 satisfying the following conditions . 1 .
there exists @xmath215 such that @xmath473 2 .
the sequence of currents @xmath474 is bounded in the mass norm .
3 . for any @xmath475 and almost any @xmath79
we have @xmath476 then the sequence of currents @xmath477 converges in the flat metric to @xmath478 .
[ th : approx ] it suffices to show that any subsequence of @xmath479 contains a sub - subsequence that converges in the flat metric to @xmath99 . to keep the notations at bay
, we will denote by @xmath479 the various intervening subsequences of @xmath479 . the conditions ( a ) and ( b ) imply via the compactness theorem for integral currents that @xmath479 contains a subsequence convergent in the flat metric to a subanalytic cycle @xmath480 .
to prove that @xmath481 we will invoke theorem [ th : ber - unique ] so that we have to show that @xmath482 denote by @xmath483 the flat norm .
using the slicing lemma ( * ? ? ?
* lemma 8.1.16 ) we deduce that @xmath484 thus , to prove ( [ eq : morse - sl2 ] ) is suffices to show that @xmath485 converges weakly to @xmath486 for almost any @xmath61 .
we think of the integral currents @xmath485 and @xmath486 as ( signed ) borel measures on the real axis concentrated on finite sets .
if @xmath487 denotes the characteristic function of the interval @xmath488 , @xmath489 , then we deduce from ( [ eq : vary ] ) that for almost any @xmath490 and any @xmath489 we have @xmath491 using ( c ) and ( [ eq : vary ] ) we deduce @xmath492 since @xmath493 we conclude @xmath494 let us show that the above equality implies that @xmath495 converges weakly to @xmath486 . for @xmath496 we define @xmath497,\;\ ; h_k(\xi)=\sum_{x\in x } |i_{x_k}(\xi , x)|.\ ] ] observe that if the slice @xmath498 is defined , then @xmath499 .
if @xmath500 , then @xmath501 hence @xmath502 we set @xmath503 and we deduce from the above inequality that for almost any @xmath79 we have @xmath504 .
let @xmath505 .
denote by @xmath506 the lipschitz constant of @xmath507 and fix a very small @xmath268 .
define @xmath508 using ( [ eq : conv - half ] ) we deduce @xmath509 choose @xmath510 such that @xmath511 now observe that @xmath512 .
we conclude @xmath513 hence , @xmath514 we deduce that @xmath515 i.e. , @xmath485 converges weakly to @xmath486 for almost any @xmath61 .
since the subject of tame geometry is not very familiar to many geometers we devote this section to a brief introduction to this topic .
unavoidably , we will have to omit many interesting details and contributions , but we refer to @xcite for more systematic presentations . for every set @xmath1 we will denote by @xmath516 the collection of all subsets of @xmath1 an _ _ @xmath223-structure _ _ , such as extensions of @xmath223 by `` infinitesimals '' .
this can come in handy even if one is interested only in the field @xmath223 . ]
is a collection @xmath517 , @xmath518 , with the following properties .
@xmath519 : : @xmath520 contains all the real algebraic subvarieties of @xmath521 , i.e. , the zero sets of finite collections of polynomial in @xmath3 real variables . @xmath522 : : for every linear map @xmath523 , the half - plane @xmath524 belongs to @xmath520 .
@xmath525 : : for every @xmath526 , the family @xmath520 is closed under boolean operations , @xmath527 , @xmath528 and complement .
@xmath529 : : if @xmath530 , and @xmath531 , then @xmath532 .
@xmath533 : : if @xmath530 , and @xmath534 is an affine map , then @xmath535 .
denote by @xmath536 the collection of real semialgebraic sets .
thus , @xmath537 if and only if @xmath538 is a finite union of sets , each of which is described by finitely many polynomial equalities and inequalities .
the celebrated tarski - seidenberg theorem states that @xmath539 is a structure .
let @xmath540 be an @xmath223-structure .
then a set that belongs to one of the @xmath520-s is called @xmath540-_definable_. if @xmath541 are @xmath540-definable , then a function @xmath542 is called @xmath540-_definable _ if its graph @xmath543 is @xmath540-definable .
given a collection @xmath544 , @xmath545 , we can form a new structure @xmath546 , which is the smallest structure containing @xmath540 and the sets in @xmath547 .
we say that @xmath546 is obtained from @xmath540 by _ adjoining the collection @xmath548_. an @xmath223-structure is called _
( order minimal ) or _ tame _ if it satisfies the property * t * : : any set @xmath549 is a _ finite _ union of open intervals @xmath550 , @xmath551 , and singletons @xmath552 .
\(a ) ( tarski - seidenberg ) the collection @xmath536 of real semialgebraic sets is a tame structure .
\(b ) ( a. gabrielov , r. hardt , h. hironaka , @xcite ) a _ restricted _ real analytic function is a function @xmath553 with the property that there exists a real analytic function @xmath554 defined in an open neighborhood @xmath555 of the cube @xmath556^n$ ] such that @xmath557 we denote by @xmath558 the structure obtained from @xmath539 by adjoining the graphs of all the restricted real analytic functions .
then @xmath558 is a tame structure , and the @xmath558-definable sets are called _ ( globally ) subanalytic sets_. the definable sets and function of a tame structure have rather remarkable _ tame _ behavior which prohibits many pathologies .
it is perhaps instructive to give an example of function which is not definable in any tame structure .
for example , the function @xmath559 is not definable in a tame structure because the intersection of its graph with the horizontal axis is the countable set @xmath560 which violates the tameness condition @xmath561
. we will list below some of the nice properties of the sets and function definable in a fixed tame structure @xmath540 .
their proofs can be found in @xcite
. we will interchangeably refer to sets or functions definable in a given tame structure @xmath540 as _ definable _ , _ constructible _ or _ tame_. ( _ piecewise smoothness of tame functions .
_ ) suppose @xmath538 is a definable set , @xmath22 is a positive integer , and @xmath562 is a definable function
. then @xmath538 can be partitioned into finitely many definable sets @xmath563 , such that each @xmath564 is a @xmath209-manifold , and each of the restrictions @xmath565 is a @xmath209-function .
( _ triangulability .
_ ) for every compact definable set @xmath538 , and any finite collection of definable subsets @xmath566 , there exists a compact simplicial complex @xmath567 , and a definable homeomorphism @xmath568 such that all the sets @xmath569 are unions of relative interiors of faces of @xmath567 .
( _ dimension .
_ ) the dimension of a definable set @xmath570 is the supremum over all the nonnegative integers @xmath571 such that there exists a @xmath420 submanifold of @xmath521 of dimension @xmath571 contained in @xmath538 .
then @xmath572 , and @xmath573 .
( _ definable selection . _ ) any tame map
@xmath542 ( not necessarily continuous ) admits a tame section , i.e. , a tame map @xmath574 such that @xmath575 , @xmath576 .
( _ local triviality of tame maps _ ) if @xmath542 is a tame continuous map , then there exists a tame triangulation of @xmath577 such that over the relative interior of any face the map @xmath341 is a locally trivial fibration .
( _ the @xmath0-minimal euler characteristic _ ) there exists a function @xmath578 uniquely characterized by the following conditions .
* @xmath579 , @xmath580 . * if @xmath581 is compact , then @xmath356 is the usual euler characteristic of @xmath1 .
( _ the scissor principle _ )
suppose @xmath538 and @xmath577 are two tame sets
. then the following are equivalent * the sets @xmath538 and @xmath577 have the same @xmath0-minimal euler characteristic and dimension .
* there exists a tame bijection @xmath542 .
( the map @xmath341 _ need not be continuous_. ) ( _ crofton formula _
, @xcite , ( * ? ? ?
2.10.15 , 3.2.26 ) . )
suppose @xmath582 is an euclidean space , and denote by @xmath583 the grassmannian of affine subspaces of codimension @xmath58 in @xmath582 .
fix an invariant measure @xmath584 on @xmath583 .
is unique up to a multiplicative constant . ]
denote by @xmath585 the @xmath58-dimensional hausdorff measure .
then there exists a constant @xmath267 , depending only on @xmath584 , such that for every compact , @xmath58-dimensional tame subset @xmath586 we have @xmath587 ( _ finite volume .
_ ) any compact @xmath58-dimensional tame set has finite @xmath58-dimensional hausdorff measure .
( _ uniform volume bounds .
_ ) if @xmath588 is a proper , continuous definable map such that all the fibers have dimensions @xmath589 , then there exists @xmath267 such that @xmath590
in this appendix we gather without proofs a few facts about the subanalytic currents introduced by r. hardt in @xcite .
our terminology concerning currents closely follows that of federer @xcite ( see also the more accessible @xcite ) .
however , we changed some notations to better resemble notations used in algebraic topology .
suppose @xmath1 is a @xmath129 , oriented riemann manifold of dimension @xmath3 .
we denote by @xmath591 the space of @xmath58-dimensional currents in @xmath1 , i.e. , the topological dual space of the space @xmath592 of smooth , compactly supported @xmath58-forms on @xmath1
. we will denote by @xmath593 the natural pairing .
the boundary of a current @xmath594 is the @xmath595-current defined via the stokes formula @xmath596 for every @xmath597 , @xmath598 , @xmath599 define @xmath600 by @xmath601 we have @xmath602 @xmath603 which yields the _ homotopy formula _
@xmath604 we have the following important result @xcite .
suppose @xmath605 is a @xmath58-dimensional _ cycle _ whose support is contained in a @xmath58-dimensional affine subspace @xmath606 .
then there exists an orientation @xmath607 on @xmath122 and an integer @xmath297 such that @xmath608 $ ] where @xmath609 $ ] is the current of integration along the oriented affine plane @xmath122 .
in particular , if @xmath391 is compact , then @xmath110.[th : const ] we say that a set @xmath610 is _ locally subanalytic _ if for any @xmath611 we can find an open ball @xmath577 centered at @xmath22 such that @xmath612 is globally subanalytic .
there is a rather subtle distinction between globally subanalytic and locally subanalytic sets .
for example , the graph of the function @xmath613 is a locally subanalytic subset of @xmath614 , but it is not a globally subanalytic set . note that a compact , locally subanalytic set is globally subanalytic .
if @xmath610 is an orientable , locally subanalytic , @xmath420 submanifold of @xmath521 of dimension @xmath58 , then any orientation @xmath615 on @xmath63 determines a @xmath58-dimensional current @xmath616 $ ] via the equality @xmath617{\rangle}:=\int_s \alpha,\;\;\forall
\alpha\in \omega^k_{cpt}({{\mathbb r}}^n).\ ] ] the integral in the right - hand side is well defined because any bounded , @xmath58-dimensional globally subanalytic set has finite @xmath58-dimensional hausdorff measure . for any open , locally subanalytic subset @xmath618
we denote by @xmath616\cap u$ ] the current @xmath619 $ ] . for any locally subanalytic subset @xmath620
we denote by @xmath57 the abelian subgroup of @xmath621 generated by currents of the form @xmath616 $ ] , as above , where @xmath622 .
the above operation @xmath616\cap u$ ] , @xmath555 open subanalytic extends to a morphism of abelian groups @xmath623 we will refer to the elements of @xmath57 as _ subanalytic ( integral ) @xmath58-chains _ in @xmath1 .
given compact subanalytic sets @xmath624 we set @xmath625 and @xmath626 we set @xmath627 r. hardt has proved in @xcite that the assignment @xmath628 satisfies the eilenberg - steenrod homology axioms with @xmath629-coefficients .
this implies that @xmath630 is naturally isomorphic with the integral homology of the pair . to describe the intersection theory of subanalytic chains we need to recall a fundamental result of r. hardt , ( *
* theorem 4.3 ) .
suppose @xmath631 are two oriented real euclidean spaces of dimensions @xmath632 and respectively @xmath633 , @xmath634 is a real analytic map , and @xmath635 a subanalytic current of codimension @xmath636 .
if @xmath637 is a regular value of @xmath341 , then the fiber @xmath638 is a submanifold equipped with a natural coorientation and thus defines a subanalytic current @xmath639 $ ] in @xmath640 of codimension @xmath633 , i.e. , @xmath639]\in { \euscript{c}}_{d_0-d_1}(e_0)$ ] .
we would like to define the intersection of @xmath642 and @xmath639 $ ] as a subanalytic current @xmath643 .
it turns out that this is possibly quite often , even in cases when @xmath637 is not a regular value .
let @xmath640 , @xmath644 , @xmath642 and @xmath341 be as above , denote by @xmath645 the euclidean volume form on @xmath644 , by @xmath646 the volume of the unit ball in @xmath644 , and set @xmath647 for every @xmath268 and @xmath648 we define @xmath649 by @xmath650 then for every @xmath651 , the currents @xmath652 converge weakly as @xmath268 to a subanalytic current @xmath653 called the _
@xmath341-slice _ of @xmath642 over @xmath637 .
moreover , the map @xmath654 is continuous in the locally flat topology.[th : slice ]
throughout this appendix we fix an @xmath0-minimal category of sets and we will refer to the sets and maps in this category as _ tame _ or _ definable_. for a topological space @xmath655 we denote by @xmath656 the ( ech ) cohomology with real coefficients , and we define its topological euler characteristic to be the integer @xmath657 whenever the sum in the right - hand side is well defined .
suppose @xmath1 is a locally closed tame subset of @xmath2 , and @xmath63 is a closed tame subset of @xmath1 .
we define the local cohomology of @xmath1 along @xmath63 ( with real coefficients ) to be @xmath658 we can now define the _ local cohomology sheaves _
@xmath659 to be the sheaves associated to the presheaves @xmath660 .
if @xmath71 and @xmath661 denotes the open ball of radius @xmath662 centered at @xmath78 then for every @xmath663 we have morphisms @xmath664 , and then the stalk of @xmath665 at @xmath78 is the inductive limit @xmath666 .
observe that since @xmath1 is locally conical we have @xmath667 we set @xmath668 we have a grothendieck spectral sequence converging to @xmath669 whose @xmath670 term is @xmath671 .
if it happens that the local cohomology sheaves are supported by finite sets then @xmath672 so that the spectral sequence degenerates at the @xmath670-terms . in this case we have @xmath673 in particular @xmath674 suppose now that @xmath1 is a compact connected tame subset of @xmath2 and @xmath675 is a definable , continuous function .
we will write @xmath676 a real number @xmath636 is said to be a _ regular _ value of @xmath341 if there exists @xmath268 such that the induced map @xmath677 is a locally , definably trivial fibration .
a real number @xmath636 is said to be a
_ critical value _ of @xmath341 if it is not a critical value . the local trivialization theorem in ( * ? ? ? * chap .
9 ) implies that the set of critical values of @xmath341 is finite .
fix a real number @xmath636 , and consider the sheaves of local cohomology @xmath678 associated to the closed subset set @xmath679 .
note that if @xmath680 , and @xmath636 is a regular value , then @xmath681 .
any number outside the range of @xmath341 is automatically a regular value .
a point @xmath71 is called a _ homological _
critical point of @xmath341 if @xmath682 note that in this case @xmath636 must be a critical value of @xmath341 .
we denote by @xmath683 the set of homological critical points of @xmath341 .
we say that @xmath341 is _ nice _ if @xmath683 is finite .
we define the _ index _ of @xmath341 at a point @xmath684 to be the integer @xmath685 where @xmath686 and @xmath687 denotes the _ open _ ball in @xmath2 of radius @xmath19 , centered at @xmath78 .
note that @xmath688 if @xmath689 .
due to the local conical structure of @xmath1 we have @xmath690 a point @xmath71 is called a _ numerically critical point _ of @xmath341 if @xmath691 .
we denote by @xmath692 the set of numerically critical points of @xmath341 .
observe that @xmath693 .
one can ask the following natural question . given
a compact tame set @xmath1 , do there exist nice continuous tame functions @xmath675 ?
the answer is , yes , plenty of them .
more precisely one can show ( see @xcite ) that there exists a subset @xmath694 , such that @xmath695 and for any @xmath696 , the induced function @xmath697 has only a finite number of homological critical points .
\(a ) note that if @xmath1 is a compact @xmath129-submanifold of @xmath2 , @xmath341 is a morse function on @xmath1 , and @xmath78 is a critical point of @xmath341 with morse index @xmath698 , then @xmath699 . in this case @xmath700 .
\(b ) if @xmath1 is a compact , convex , subanalytic subset of @xmath2 and @xmath701 is a linear map , then a point@xmath71 is critical for the restriction of @xmath64 to @xmath1 if and only if @xmath78 is a minimum point for @xmath64 .
in this case we have @xmath702 .
suppose that @xmath1 is a compact connected tame subset of @xmath2 and @xmath675 is a nice continuous tame function that contains no homological critical points on the level set @xmath703 .
then the inclusion induced morphism @xmath704 is an isomorphism for all @xmath268 sufficiently small .
[ lemma : kashi ] we first prove that the morphism @xmath705 .
is an isomorphism .
indeed , it suffices to show that the local cohomology of @xmath706 along @xmath707 is trivial .
this follows from ( [ eq : local - coho ] ) by observing that the local cohomology sheaves @xmath708 are trivial .
next observe that for some @xmath709 the induced map @xmath710 is a locally trivial fibration .
this implies that for any @xmath711 the induced morphism @xmath712 is an isomorphism .
we conclude by observing that @xmath713 .
suppose is a compact connected tame set subset of @xmath2 and @xmath675 is a nice , continuous tame function .
fix @xmath714 .
we have @xmath715 . from the equality @xmath716 we deduce @xmath717 using ( [ eq : vanish ] ) and ( [ eq : local - chi ] ) we deduce that for any sufficiently small @xmath268 we have @xmath718 suppose now that @xmath719 , @xmath319 and the interval @xmath720 contains no critical values of @xmath341 .
then , @xmath721 iterating ( [ eq : top - drop ] ) and ( [ eq : = ] ) we deduce that for any @xmath722 , @xmath319 , we have @xmath723 p. schapira : http://people.math.jussieu.fr/~schapira/respapers/tomoln.pdf[_tomography of constructible functions _ ] .
applied algebra , algebraic algorithms and error - correcting codes ( paris , 1995 ) , 427435 , lecture notes in comput .
, 948 , springer , berlin , 1995 . | we present a very short complete proof of the existence of the normal cycle of a subanalytic set .
the approach is morse theoretic in flavor and relies heavily on techniques from @xmath0-minimal topology . |
exclusive nonleptonic b decays have provided a fertile field to investigate the cp violation and search for new physics .
the hadronic matrix elements of the effective operators play a key role in the study of b meson decays , but it is difficult to calculate them precisely due to the long distance qcd dynamics . the factorization approach ( fa ) @xcite based on the color transparency mechanism has been applied to many decay modes , and it works well in many channels .
but it suffers from some problems such as infrared - cutoff and scale dependence . to solve these problems and make more accurate predictions , the perturbative qcd approach ( pqcd ) @xcite , the qcd improved factorization (
qcdf ) @xcite as well as the soft - collinear effective theory ( scet ) @xcite have been developed in the recent years .
pqcd is based on @xmath6 factorization theorem @xcite .
the decay amplitude is factorized into the convolution of the mesons light - cone wave functions ( see @xmath7 ) , the hard scattering kernels and the wilson coefficients , which stand for the soft , hard and harder dynamics respectively .
the transverse momentum is introduced so that the endpoint singularity which will break the collinear factorization is regulated and the large double logarithm term appears after the integration on the transverse momentum , which is then resummed into the sudakov form factor .
the formalism can be written as : @xmath8,\end{aligned}\ ] ] where the @xmath9 is the conjugate space coordinate of the transverse momentum , it denotes the transverse interval of the meson .
@xmath10 is the energy scale in hard function @xmath11 .
the jet function @xmath12 comes from the summation of the double logarithms @xmath13 near the endpoint , called threshold resummation @xcite .
the factorization theorem guarantees the infrared safety and the gauge invariance of the hard kernel and has been proved to all order of @xmath14 @xcite .
many hadronic two body @xmath15 decays have been studied in pqcd approach @xcite .
most predictions are consistent with the current experiments .
the @xmath1 decays are important to extract ckm phase angles and study the cp violation . as @xmath1 meson is not in the energy scale of the high luminosity b factories slac and kek , it is more difficult to be produced and measured now .
we can study the @xmath1 decays more precisely in the very near future with the increase of luminosity at tevatron and the upcoming large hadron collider ( lhc ) .
@xmath1 meson is different from b meson due to the heavier strange quark ( compare to @xmath16 , @xmath17 quark ) which induces the su(3 ) symmetry - breaking effect .
this effect is considered to be small and the distribution amplitude of @xmath1 meson(given in the following formula ) should be similar to that of the b meson , @xmath18.\label{bwave}\end{aligned}\ ] ] the upper limit of the @xmath19 branching ratio is @xmath20 @xcite , which constrain the parameter @xmath21 to a lower limit of about 0.5 @xcite .
moreover , in order to fit the branching ratio measured in the @xmath22 decay @xcite , we constrain @xmath21 to about 0.55 @xcite , then we can see that the su(3 ) symmetry - breaking is not negligible . here
we integrate out the variable @xmath23 and show the distribution amplitude of @xmath15 and @xmath1 meson in fig.1 .
we can see that the peak point of the curve of @xmath1 meson s distribution amplitude prefers a larger @xmath24 ( @xmath24 denotes the momentum fraction carried by the light quark ) region comparing to the b meson .
this is consistent with the fact that the s quark much heavier than the d ( u ) quark , should carry more momentum . later in this paper
, we will see that the branching ratios of @xmath1 decays are very sensitive to this parameter .
if measured by experiments , radiative leptonic decays of @xmath1 meson can provide information of this parameter @xcite .
= 4.0 in in this paper , we study @xmath25 decays in the pqcd approach . hopefully the branching ratio is not too small and can be detected by the tevatron or lhcb experiments , then it may allow us to determine the @xmath1 distribution amplitude and su(3 ) breaking effects with much more precision .
moreover , we can also constrain @xmath2 with fewer pollution from this channel .
we use the effective hamiltonian for the process @xmath25 given by @xcite @xmath26 -v_t\sum_{i=3}^{10}c_i(\mu)o_i^{(q)}(\mu)\right\},\end{aligned}\ ] ] where @xmath27 , @xmath28 , @xmath29 are the wilson coefficients , and the operators are @xmath30 here i and j stand for @xmath31 color indices . the decay width for these channels
is : @xmath32 where @xmath33 is the 3-momentum of the final state mesons , @xmath34 , and @xmath35 .
@xmath36 is the decay amplitude , which will be calculated later in pqcd approach .
the subscript @xmath37 denotes the helicity states of the two vector mesons with the longitudinal ( transverse ) components l(t ) . according to lorentz structure analysis
, the amplitude can be decomposed into : @xmath38 we can define the longitudinal @xmath39 , transverse @xmath40 helicity amplitudes @xmath41 where @xmath42 .
they satisfy the relation @xmath43 the leading order diagrams in pqcd approach are shown in fig.2 .
the amplitudes for @xmath44 and @xmath45are written as @xmath46 respectively , where the subscript @xmath11 denote different helicity amplitudes @xmath47 and @xmath48 , and @xmath49 and @xmath50 are the amplitudes from tree and penguin diagrams respectively .
the detailed formulae of @xmath49 and @xmath50 are similar to those in @xmath51 @xcite and @xmath52 @xcite , so we will not show them here .
= 3.0 in the parameters used in our calculations are : the fermi coupling constant @xmath53 , the meson masses @xmath54 @xcite , the decay constants @xmath55 @xcite , the central value of the ckm matrix elements @xmath56 , latexmath:[$|v_{td}|=0.0075,\ |v_{tb}|=0.9992,\ @xmath1 meson lifetime @xmath58 @xcite . .
branching ratio , polarization fraction and direct cp asymmetry dependence on @xmath21 [ cols="^,^,^,^,^,^,^",options="header " , ] as we mentioned before , the decay @xmath25 can be used to determine the @xmath1 meson wave function parameter @xmath21 , or @xmath21 can influence our predictions of @xmath25 decay .
so we show the results in table 1 according to 3 different values of @xmath21 . from the table
we can easily find out the averaged branching ratio for @xmath59 is much larger than the other two , for @xmath59 involve large wilson coefficient @xmath60 for the factorizable part while the other two ( @xmath61 and @xmath62 ) involve a much smaller wilson coefficient @xmath63 ( color - suppressed ) for the factorizable part of the emission diagram . as a result
the first one is tree dominated and has a large branching ratio and small direct cp asymmetry .
while referring to the other two , the contributions from penguin and tree diagrams are at the same order ( @xmath64 ) , hence we can expect a large direct cp asymmetry from eqs.([adircp ] ) .
the polarization fraction difference of these channels are also due to that the main contribution of each channel comes from different topology .
@xmath59 is tree dominated .
the main contribution comes from the factorizable part of the emission diagram , where transverse polarization amplitude is suppressed by a factor @xmath65 ( see formulas in @xcite ) , so the longitudinal polarization dominates and contributes more than @xmath66 of the total branching ratio .
but in @xmath61 and @xmath62 decays , tree emission ( factorizable ) diagram contribution is suppressed due to the cancellation of wilson coefficients @xmath67 .
the left dominant contribution is the non - factorizable diagrams of tree operators and penguin diagrams .
both of these contributions equally contribute to longitudinal and transverse polarizations .
the transverse polarization is not suppressed in those cases , therefore numerically we get a small longitudinal fraction of about 0.4 .
this similar situation is also found in @xmath68 decays @xcite , which are related by su(3 ) symmetry to our @xmath69 decays .
= 3.0 in to extract the cp violation parameters and dependence on ckm phase angle @xmath2 of these decays , we rewrite the helicity amplitudes in ( [ m],[barm ] ) as the functions of @xmath2 : @xmath70where @xmath71 , and @xmath72 is the relative strong phase between tree ( @xmath48 ) and penguin ( @xmath73 ) diagrams . here in pqcd approach , the strong phase comes from the nonfactorizable diagrams and annihilation diagrams .
this is different from beneke - buchalla - neubert - sachrajda @xcite approach . in that approach ,
annihilation diagrams are not taken into account , strong phases mainly come from the so - called bander - silverman - soni mechanism @xcite .
as shown in @xcite , these effects are in fact next - to - leading - order ( @xmath14 suppressed ) elements and can be neglected in pqcd approach .
we give the averaged branching ratios of @xmath74 and @xmath75 as a function of @xmath2 in fig.3 , and the averaged branching ratios of @xmath76 in fig.4 .
= 3.0 in using eqs.([m++],[m ] ) ,
the direct cp violating parameter is @xmath77 notice the cp asymmetry for these channels are sensitive to ckm angle @xmath2 , we show the direct cp asymmetry as a function of @xmath2 in fig.5 .
it is easy to see that the @xmath78 and @xmath79 have large direct cp asymmetries up to 50% , with a relative minus sign . on the other hand
, the @xmath80 decay has small direct cp asymmetry due to only one large tree contribution in this decay .
the uncertainty shown at this table is only from the @xmath1 meson wave function parameter dependence .
in fact , since cp asymmetry is sensitive to many parameters , the line should be more broadened by uncertainties .
the mixing induced cp asymmetry is complicated and requires angular distribution study , similar study may be found in @xcite .
= 3.0 in at last , if we compare our predictions with those of naive factorization @xcite @xmath81 and ones of qcdf @xcite @xmath82 we can see that they are consistent .
it should be noticed that the branching ratios in fa and qcdf strongly depend on form factors . while in pqcd , the branching ratios and form factors depend on wave functions , especially the @xmath1 meson wave function .
nowadays , very few @xmath1 meson decays have been measured , so we can only give rough constraints on the parameters from other channel and permit large errors .
more experimental data can help to constrain the form factors and wave functions , then we can give more precise predictions and the different methods can be tested by the experiments .
although similar results are got by different methods for branching ratios , the polarization fractions are quite different .
the qcdf and naive factorization give only several percent transverse polarization for all three decay modes @xcite , while table 1 shows large transverse contribution for @xmath83 decays in our pqcd approach .
the direct cp asymmetry are not given in ref.@xcite , but they probably also differ from pqcd approach as it happened in @xmath84 and @xmath85 case @xcite .
the numerical results shown here are only leading order ones . for the tree dominated channel @xmath86 , the leading order diagrams should give the main contribution .
but for the other two decays , with a branching ratio as small as @xmath87 , the next - to - leading order and power suppressed contributions should not be negligible , the results may suffer from large corrections when the next to leading order corrections are included @xcite .
current experiments @xcite only give the upper limit for the decay @xmath88 more data are needed to test our calculations .
in this paper we calculate the branching ratios , polarization fraction and cp asymmetries of @xmath89 modes using pqcd theorem in sm .
we perform all leading order diagrams to next to leading twist wave functions .
we also study the dependence of their averaged branching ratios and the cp asymmetry on the ckm angle @xmath2 .
at last we compare our predictions with values from other approaches .
this work is partly supported by the national science foundation of china under grant ( no.90103013 , 10475085 and 10135060 ) .
we thank g .-
l . song for reading our manuscript and giving us many helpful suggestions .
we also thank j - f cheng , h - n li , y. li , and x - q yu for helpful discussions .
we thank the institute for nuclear theory at the university of washington for its hospitality and the department of energy for partial support during the completion of this work .
for longitudinal polarized @xmath90 meson , the wave function is written as @xmath91,\end{aligned}\ ] ] and the wave function for transverse polarized @xmath90 meson reads @xmath92.\end{aligned}\ ] ] the @xmath90 meson distribution amplitudes up to twist-3 are given by ref.@xcite with qcd sum rules .
@xmath93,\\ & & \phi^t_{k^*}(x)=\frac{f_{k^*}^t}{2\sqrt{2n_c}}\left\{0.3(1 - 2x)[3(1 - 2x)^2 + 10(1 - 2x)-1 ] + 1.68c_4^{1/2}(1 - 2x)\right.\nonumber\\ & & \left.\ \ \ \ \ \ \ \ \ \ \ \ + 0.06(1 - 2x)^2[5(1 - 2x)^2 - 3]+0.36\left\{1 - 2(1 - 2x)[1+\ln(1-x)]\right\}\right\},\\ & & \phi^s_{k^*}(x)=\frac{f_{k^*}^t}{2\sqrt{2n_c}}\left\{3(1 - 2x)\left[1 + 0 .. 2(1 - 2x)+0.6(10x^2 - 10x+1)\right ] \right.\nonumber\\ & & \left.\ \ \ \ \ \ \ \ \ \ \ \ -0.12x(1-x)+0.36[1 - 6x-2\ln{(1-x)}]\right\},\\ & & \phi^t_{k^*}(x)=\frac{3f_{k^*}^t}{\sqrt{2n_c}}x(1-x)[1 + 0.6(1 - 2x)+0.04c_2^{3/2}(1 - 2x ) ] , \\ & & \phi^v_{k^*}(x)=\frac{f_{k^*}^t}{2\sqrt{2n_c}}\left\{\frac{3}{4}[1+(1 - 2x)^2 + 0.44(1 - 2x)^3 ] \right.\nonumber\\ & & \left.\ \ \ \ \ \ \ \ \ \ \ \ + 0.4c_2^{1/2}(1 - 2x ) + 0.88c_4^{1/2}(1 - 2x)+0.48[2x+\ln(1-x ) ] \right\},\\ & & \phi^a_{k^*}(x)=\frac{f_{k^*}^t}{4\sqrt{2n_c}}\left\{3(1 - 2x)[1 + 0.19(1 - 2x)+0.81(10x^2 - 10x+1 ) ] \right.\nonumber\\ & & \left.\ \ \ \ \ \ \ \ \ \ \ \ -1.14x(1-x)+0.48[1 - 6x-2\ln(1-x)]\right\},\end{aligned}\ ] ] where the gegenbauer polynomials are @xmath94 for @xmath95 and @xmath96 meson , we employ @xmath97 and @xmath98 .
their lorentz structures are similar to @xmath90 meson , the distribution amplitudes are the same for @xmath95 and @xmath96 and given as @xcite : @xmath99\ ; , \label{pwr}\\ \phi_{\rho}^t(x)&=&\frac{f^t_{\rho}}{2\sqrt{2n_c } } \left\{3(1 - 2x)^2 + 0.3(1 - 2x)^2[5(1 - 2x)^2 - 3]\right .
\nonumber \\ & & \left.+0.21[3 - 30(1 - 2x)^2 + 35(1 - 2x)^4]\right\}\ ; , \label{pwt}\\ \phi_{\rho}^s(x ) & = & \frac{3f_\rho^t}{2\sqrt{2n_{c } } } ( 1 - 2x)\left[1 + 0.76(10x^2 - 10x+1)\right]\ ; , \label{pws}\\ \phi_\rho^t(x)&=&\frac{3f_\rho^t}{\sqrt{2n_c } } x(1-x)\left[1 + 0.2c_2^{3/2}(1 - 2x)\right]\ ; , \label{pwft}\\ \phi_{\rho}^v(x)&=&\frac{f_{\rho}}{2\sqrt{2n_c } } \bigg\{\frac{3}{4}[1+(1 - 2x)^2]+0.24[3(1 - 2x)^2 - 1 ] \nonumber \\ & & + 0.12[3 - 30(1 - 2x)^2 + 35(1 - 2x)^4]\bigg\}\ ; , \label{pwv}\\ \phi_{\rho}^a(x ) & = & \frac{3f_\rho}{4\sqrt{2n_{c } } } ( 1 - 2x)\left[1 + 0.93(10x^2 - 10x+1)\right]\;. \label{pwa}\end{aligned}\ ] ] 99 m. wirbel , b. stech , m. bauer , z. phys .
c29 , 637 ( 1985 ) ; m. bauer , b. stech , m. wirbel , z. phys .
c34 , 103 ( 1987 ) ; l .- l .
chau , h .- y .
cheng , w.k .
sze , h. yao , b. tseng , phys .
d43 , 2176 ( 1991 ) .
a. ali , g. kramer and c.d .
l , phys . rev .
d 58 , 094009 ; c.d .
l , nucl .
proc . suppl .
cheng , et al , phys .
d 60 , 094014 ( 1999 ) . g.p .
leapage and s.j .
brodsky , phys .
d 22 , 2157 ( 1980 ) ; s.j .
brodsky , g.p .
lepage , p.b .
mackenzie , phys .
d 28 , 228 ( 1983 ) .
li and h.l .
yu , phys .
74 , 4388 ( 1995 ) ; phys .
b 353 , 301 ( 1995 ) ; phys . rev .
d 53 , 2480 ( 1996 ) .
keum , h - n .
li , and a.i .
sanda , phys .
b 504 , 6 ( 2001 ) ; phys . rev .
d 63 , 054008 ( 2001 ) ; y.y . keum and h - n li , phys . rev .
d 63 , 074006 ( 2001 ) . c. d. l , k. ukai , and m. z. yang , phys
d 63 , 074009 ( 2001 ) ; c. d. l and m.z .
yang , eur .
j. c 23 , 275 ( 2002 ) .
m. beneke , g. buchalla , m. neubert and c.t .
sachrajda , phys .
83 , 1914 ( 1999 ) ; nucl . phys .
b 591 , 313 ( 2000 ) .
m. beneke , g. buchalla , m. neubert and c.t .
sachrajda , nucl .
b 606 , 245 ( 2001 ) , nucl . phys .
b 675 , 333 - 415 ( 2003 ) . c.w .
bauer , s. fleming , and m. luke , phys .
d 63 , 014006 ( 2001 ) , c.w .
bauer , s. fleming , d. pirjol , and i.w .
stewart , phys .
d 63 , 114020 ( 2001 ) ; c.w . bauer and i.w .
stewart , phys .
b 516 , 134 ( 2001 ) , phys . rev .
d 65 , 054022 ( 2002 ) .
s. catani , m. ciafaloni and f. hautmann , phys .
b 242 , 97 ( 1990 ) ; nucl . phys .
b 366 , 135 ( 1991 ) . j.c .
collins and r.k .
ellis , nucl .
b 360 , 3 ( 1991 ) . e.m .
levin , m.g .
ryskin , yu.m .
shabelskii , and a.g .
shuvaev , sov.j . nucl.phys .
53 , 657 ( 1991 ) .
h - n . li , phys .
d66 , 094010 ( 2002 ) ; h - n .
li , k. ukai , phys .
b 555 , 197 ( 2003 ) .
li and m. nagashima , phys .
d67 ( 2003 ) 034001 c.h .
chen , h - n .
li , phys.rev .
d 66 ( 2002 ) 054013 .
keum , t. kurimoto , h - n .
sanda , phys .
d69 ( 2004 ) 094018 ; c.d .
yang , eur .
j. c23 ( 2002 ) 275 - 287 ; c.d .
shen and j. zhu , eur .
j. c41 , 311 - 317 ( 2005 ) .
a. warburton et al , cdf collaboration , int . j. mod
a20 , 3554 - 3558 ( 2005 ) .
yu , y. li , c.d .
l , phys .
d71 , 074026 ( 2005 ) .
m. rescigno , hep - ex/0410063 , cdf collaboration , d. acosta et al , phys .
95 , 031801 ( 2005 ) .
cheng , c.d .
shen and j. zhu , in preparation .
j. zhu , y.l .
shen and c.d .
l , phys .
d72 , 054015 ( 2005 ) .
particle data group , s. eidelman et al .
b 592 , 1 ( 2004 ) . p. ball and r. zwicky , phys .
d71 , 014029 ( 2005 ) .
heavy flavor averaging group , hep - ex/0412073 .
y. li , c.d .
l , e - print archive : hep - ph/0508032 .
m. bander , d. silverman , and a. soni , phys .
43 , 242 ( 1979 ) .
g. kramer , w.f .
plamer , phys .
d 45 , 193 ( 1992 ) ; c.s .
kim , y.g .
kim , c.d .
lu , t. morozumi , phys .
d 62 , 034013 ( 2000 ) . h.y .
chen , h.y .
cheng , b. tseng , phys . rev .
d 59 , 074003 ( 1999 ) .
lu , and y.d .
yang , phys .
d 68 ( 2003 ) 114015 ; erratum - ibid .
d71 , 019902 ( 2005 ) . | @xmath0 are useful to determine the @xmath1 distribution amplitude , as well as constrain the ckm phase angle @xmath2 .
we study these decays within the perturbative qcd approach(pqcd ) . in this approach , we calculate factorizable , non - factorizable , as well as annihilation diagrams .
we find the branching ratio for @xmath3 is at the order of @xmath4 , and there s large direct cp violation in @xmath5 .
our predictions are consistent with those from other methods and current experiments . |
from the rosat all - sky survey ( @xmath4 150000 sources ) and the tycho catalogs ( @xmath4 1000000 stars ) guillout et al .
( 1999 ) selected about 14000 optical counterparts of stellar x - ray sources , the so - called rastyc sample .
the high x - ray emission level of these sources is compatible with young stars in the solar neighborhoods , or with close binaries of the rs cvn and by dra classes , for which the high magnetic activity level is due to the synchronization of rotational and orbital periods .
the knowledge of the incidence of binaries in this sample is of paramount importance for studying the recent local star formation history .
we thus have started a spectroscopic observation campaign aimed at a deep characterization of a representative sub - sample of the rastyc population . as a matter of fact , the high - resolution spectroscopic observations allow to derive spectral types and luminosity classes , to infer ages ( by means of lithium abundance ) and to single out spectroscopic binaries .
in addition , chromospheric activity levels ( from h@xmath2 emission ) and rotational velocities ( from doppler broadening ) , can be also determined . in this work
we report on the discovery of three single - lined ( sb1 ) and three double - lined ( sb2 ) late - type binaries , which have been observed with the 193-cm telescope of the _ observatoire de haute provence _ ( ohp ) and the 91-cm telescope of the catania astrophysical observatory ( oac ) . for these binaries
we have obtained good radial velocity curves and orbital solutions . an accurate spectral classification has been also performed and the projected rotational velocity @xmath5 has been measured for all stars . the h@xmath2 emission and the lii@xmath6 line
have been used to infer the chromospheric activity level and the lithium content , respectively .
spectroscopic observations have been obtained at the _ observatoire de haute provence _ ( ohp ) and at the _ m.g .
fracastoro _ station ( serra la nave , mt .
etna , 1750 m a.s.l . ) of the catania astrophysical observatory ( oac ) .
at ohp we observed in 2000 and 2001 with the elodie echelle spectrograph connected to the 193-cm telescope .
the 67 orders recorded by the ccd detector cover the 3906 - 6818 wavelength range with a resolving power of about 42000 ( baranne et al . 1996 ) .
the elodie spectra were automatically reduced on - line during the observations and the cross - correlation with a reference mask was produced as well .
the observations carried out at catania observatory have been performed in 2001 and 2002 with the reosc echelle spectrograph at the 91-cm telescope .
the spectrograph is fed by the telescope through an optical fiber ( uv - nir , @xmath7 core diameter ) and is placed in a stable position in the room below the dome level .
spectra were recorded on a ccd camera equipped with a thinned back - illuminated site ccd of 1024@xmath81024 pixels ( size 24@xmath824 @xmath9 m ) .
the echelle crossed configuration yields a resolution of about 14000 , as deduced from the fwhm of the lines of the th - ar calibration lamp .
the observations have been made in the red region .
the detector allows us to record five orders in each frame , spanning from about 5860 to 6700 .
the oac data reduction was performed by using the echelle task of iraf package following the standard steps : background subtraction , division by a flat field spectrum given by a halogen lamp , wavelength calibration using the emission lines of a th - ar lamp , and normalization to the continuum through a polynomial fit .
the photometric observations have been carried out in 2001 and 2002 in the standard @xmath10 system with the 91-cm telescope of oac and a photon - counting refrigerated photometer equipped with an emi 9789qa photomultiplier , cooled to @xmath11c .
the dark noise of the detector , operated at this temperature , is about @xmath12 photon / sec . for each field of the rastyc sources , we have chosen two or three stars with known @xmath13 magnitudes to be used as local standards for the determination of the photometric instrumental `` zero points '' .
additionally , several standard stars , selected from the list of landolt ( 1992 ) , were also observed during the run in order to determine the transformation coefficients to the johnson standard system . a typical observation consisted of several integration cycles ( from 1 to 3 , depending on the star brightness ) of 10 , 5 , 5 seconds , in the @xmath14 , @xmath15 and @xmath16 filter , respectively .
a 21-arcsecond diaphragm was used .
the data were reduced by means of the photometric data reduction package phot designed for photoelectric photometry of catania observatory ( lo presti & marilli 1993 ) .
seasonal mean extinction coefficient for serra la nave observatory were adopted for the atmospheric extinction correction .
6170 ( bottom ) spectral regions together with the `` synthetic '' spectrum over - plotted with a red line .
the latter was built up as the weighted sum of two standard star spectra ( k0v for the primary and k1v for the secondary one ) broadened at the @xmath5 of the respective component and doppler - shifted according to the orbital solution and the phase of the observation ( @xmath17 ) .
the fractional flux contribution of the primary component is 0.62 and 0.58 at 5170 and 6170 , respectively .
* right panels ) * observed elodie spectrum of rastyc 221428 ( thick black lines ) in the mgib ( top ) and in the @xmath36230 ( bottom ) regions .
the `` synthetic '' spectrum ( f8v + k0iii - iv ) is displayed with red lines in the same boxes .
the excellent agreement of the observed and synthetic spectra is apparent.,title="fig:",width=222,height=162 ] 6170 ( bottom ) spectral regions together with the `` synthetic '' spectrum over - plotted with a red line .
the latter was built up as the weighted sum of two standard star spectra ( k0v for the primary and k1v for the secondary one ) broadened at the @xmath5 of the respective component and doppler - shifted according to the orbital solution and the phase of the observation ( @xmath17 ) .
the fractional flux contribution of the primary component is 0.62 and 0.58 at 5170 and 6170 , respectively .
* right panels ) * observed elodie spectrum of rastyc 221428 ( thick black lines ) in the mgib ( top ) and in the @xmath36230 ( bottom ) regions . the `` synthetic '' spectrum ( f8v + k0iii - iv )
is displayed with red lines in the same boxes .
the excellent agreement of the observed and synthetic spectra is apparent.,title="fig:",width=222,height=162 ] 6170 ( bottom ) spectral regions together with the `` synthetic '' spectrum over - plotted with a red line .
the latter was built up as the weighted sum of two standard star spectra ( k0v for the primary and k1v for the secondary one ) broadened at the @xmath5 of the respective component and doppler - shifted according to the orbital solution and the phase of the observation ( @xmath17 ) .
the fractional flux contribution of the primary component is 0.62 and 0.58 at 5170 and 6170 , respectively .
* right panels ) * observed elodie spectrum of rastyc 221428 ( thick black lines ) in the mgib ( top ) and in the @xmath36230 ( bottom ) regions .
the `` synthetic '' spectrum ( f8v + k0iii - iv ) is displayed with red lines in the same boxes .
the excellent agreement of the observed and synthetic spectra is apparent.,title="fig:",width=222,height=162 ] 6170 ( bottom ) spectral regions together with the `` synthetic '' spectrum over - plotted with a red line .
the latter was built up as the weighted sum of two standard star spectra ( k0v for the primary and k1v for the secondary one ) broadened at the @xmath5 of the respective component and doppler - shifted according to the orbital solution and the phase of the observation ( @xmath17 ) .
the fractional flux contribution of the primary component is 0.62 and 0.58 at 5170 and 6170 , respectively .
* right panels ) * observed elodie spectrum of rastyc 221428 ( thick black lines ) in the mgib ( top ) and in the @xmath36230 ( bottom ) regions .
the `` synthetic '' spectrum ( f8v + k0iii - iv ) is displayed with red lines in the same boxes .
the excellent agreement of the observed and synthetic spectra is apparent.,title="fig:",width=222,height=162 ] the projected rotational velocity , @xmath5 , was measured using the elodie ccfs and the calibration relation between ccf width and @xmath5 proposed by @xcite and is reported in table 2 .
the lowest rotation rate ( @xmath184 kms@xmath19 ) has been detected for both components of hd 183957 , which display also the lowest h@xmath2 activity among the six sources .
for sb1 systems observed with elodie we have determined effective temperatures and gravity ( i.e. spectral classification ) by means of the tgmet code , available at ohp ( @xcite ) .
we have also used rotfit , a code written by one of us ( @xcite ) in idl ( interactive data language , rsi ) , which simultaneously find the spectral type and the @xmath5 of the target by searching , into a library of standard star spectra , for the standard spectrum which gives the best match ( minimum of the residuals in the difference ) of the target one , after the rotational broadening . as a standard star library
, we used a sub - sample of the entire tgmet list , composed of 87 stars well distributed in effective temperature and gravity and in a suitable range of metallicities .
the standard star spectra were retrieved from the elodie archive ( @xcite ) .
the rotfit code was also applied to the oac spectra , using standard star spectra acquired at oac with the same instrumental setup .
this was especially advantageous for df cam , for which we have no elodie spectrum .
we have performed an accurate spectral classification also for the components of sb2 systems , thanks to another idl code made by us , compo2 , which searches for the best combination of standard - star spectra able to reproduce the observed spectrum of the sb2 system .
we give the radial velocity and @xmath5 of the system components as input parameters and the code finds , for the selected spectral region , the spectral types and fractional flux contributions which better reproduce the observed spectrum , i.e. which minimize the residuals in the collection of difference , @xmath20 spectra .
results for different elodie orders from blue to red wavelengths are in very good agreement ( fig .
[ fig : synth ] ) .
region ( black lines ) .
the inactive templates built up with rotationally broadened elodie archive spectra are displayed with red lines .
the two components of hd 183957 show only a small filling of their h@xmath2 profiles , while rastyc 072133 displays a double - peaked h@xmath2 emission profile .
rastyc 215940 and 102623 display instead a pure emission h@xmath2 profile.,title="fig:",width=215,height=151 ] region ( black lines ) .
the inactive templates built up with rotationally broadened elodie archive spectra are displayed with red lines .
the two components of hd 183957 show only a small filling of their h@xmath2 profiles , while rastyc 072133 displays a double - peaked h@xmath2 emission profile .
rastyc 215940 and 102623 display instead a pure emission h@xmath2 profile.,title="fig:",width=215,height=151 ] region ( black lines ) .
the inactive templates built up with rotationally broadened elodie archive spectra are displayed with red lines .
the two components of hd 183957 show only a small filling of their h@xmath2 profiles , while rastyc 072133 displays a double - peaked h@xmath2 emission profile .
rastyc 215940 and 102623 display instead a pure emission h@xmath2 profile.,title="fig:",width=215,height=151 ] region ( black lines ) .
the inactive templates built up with rotationally broadened elodie archive spectra are displayed with red lines .
the two components of hd 183957 show only a small filling of their h@xmath2 profiles , while rastyc 072133 displays a double - peaked h@xmath2 emission profile .
rastyc 215940 and 102623 display instead a pure emission h@xmath2 profile.,title="fig:",width=215,height=151 ] region at different orbital phases .
the cooler k0 iii - iv component shows a h@xmath2 line always in emission with a stronger intensity around phase 0.7 , while the hotter f8 v component displays a normal h@xmath2 absorption profile .
* right panel ) * oac spectra of df cam in the h@xmath2 region at different orbital phases .
the h@xmath2 intensity and shape varies with the orbital / rotational phase.,title="fig:",width=215,height=340 ] region at different orbital phases .
the cooler k0 iii - iv component shows a h@xmath2 line always in emission with a stronger intensity around phase 0.7 , while the hotter f8 v component displays a normal h@xmath2 absorption profile .
* right panel ) * oac spectra of df cam in the h@xmath2 region at different orbital phases .
the h@xmath2 intensity and shape varies with the orbital / rotational phase.,title="fig:",width=215,height=340 ] the h@xmath2 line is one of the most important and easily accessible indicators of chromospheric activity .
only the very active stars show the h@xmath2 line as an emission feature always above the local continuum , while in the spectra of less active objects only a filled - in absorption line is observed .
the detection of the chromospheric emission contribution filling in the line core is hampered by the complexity of the observed spectra of double - lined binaries in which the spectral lines of both components , shifted at different wavelengths according to the orbital phase , are simultaneously seen .
therefore a comparison with an inactive " template , composed of two stellar spectra mimicking the two system components in absence of activity , is needed to emphasize the h@xmath2 chromospheric emission .
the inactive templates have been built up with rotationally broadened elodie archive spectra ( hd 165341 , k0v and hd 10476 , k1v for the primary and secondary component of hd 183957 , respectively ; @xmath21 cep , k1iv for 102623 ; @xmath22 boo , g8iii for 072133 ; hd 17382 , k1v for 215940 ) or with oac spectra of @xmath2 ari ( k2iii ) , for df cam , acquired during the observing campaigns .
the two components of hd 183957 show only a small filling of their h@xmath2 profiles ( fig .
[ fig : halpha1 ] ) , i.e. a moderate activity level .
this is in agreement with the low @xmath5 values indicating a system composed of two early - k stars which are probably synchronously rotating with a period of about 8 days .
the filling in of the h@xmath2 line of both components of hd 183957 is lower than that found for hd 166 ( @xcite ) , a k0v star belonging to the local association ( 100 - 300 myrs ) with a rotation period of 6.2 days .
the other rastyc stars display h@xmath2 emission profiles with a variety of shapes , going from a simple symmetric emission profile ( 102623 ) to a strong emission line with central reversal ( 215940 ) .
it has been also observed a very broad , complex feature with a filled - in core and an emission blue wing ( 072133 ) . a h@xmath2 profile similar to that displayed by the latter star has been sometimes observed in some long - period rs cvn s , like hk lac ( e.g. catalano & frasca @xcite ) .
rastyc 072133 was classified as a semi - regular variable after hipparcos , but it displays all the characteristics of a rs cvn sb1 binary .
the elodie spectra of rastyc 221428 in the h@xmath2 region show that the cooler k0 iii - iv component displays a h@xmath2 line always in emission with a stronger intensity around phase 0.7 ( fig .
[ fig : halpha2 ] ) .
the oac spectra of df cam always display a pure h@xmath2 emission line , whose intensity varies with the orbital / rotational phase ( fig .
[ fig : halpha2 ] ) .
similarly to rastyc 072133 , df cam , considered as a semi - regular variable after hipparcos photometry , is very likely an active binary of the rs cvn class .
the equivalent width of the lithium @xmath36708 line , @xmath23 , was measured on the elodie spectra . for the three sources for which we were able to detect and measure @xmath23 ( table 2 ) , we have deduced a lithium abundance , @xmath24 , in the range 1.31.8 , according to @xcite nlte calculations .
such values of lithium abundance indicate ages of at least a few hundreds myrs , so that the x - ray emission and the fairly strong h@xmath2 chromospheric activity detected in these systems should be essentially the effect of the rather fast stellar rotation due to the spin / orbit synchronization .
we are grateful to the members of the staff of ohp and oac observatories for their support and help with the observations .
this research has made use of simbad and vizier databases , operated at cds , strasbourg , france .
1 baranne a. , queloz d. , mayor m. , et al . , 1996 ,
a&as 119 , 373 biazzo , k. , frasca , a. , henry , g.w .
, catalano , s. , and marilli , e. 2004 , 13th cambridge workshop on cool stars , stellar systems , and the sun ( 2004 july 5 - 9 ) , ed . f. favata , ( esa sp ) , in press catalano s. and frasca a. 1994 , a&a 287 , 575 frasca a. , alcala j.m . , covino e. , catalano s. , marilli e. and paladino r. 2003 , a&a 405 , 149 guillout p. , schmitt j. h. m. m. , egret d. , voges w. , motch c. and sterzik m. f. 1999 , a&a 351 , 1003 katz d. , soubiran c. , cairel r. , adda m. and cautain r. 1998 , a&a 338 , 151 landolt , a. u. 1992 , aj , 104 , 340 lo presti , c. , & marilli , e. 1993 , phot .
photometrical data reduction package .
internal report of catania astrophysical observatory n. 2/1993 lucy , l. b. and sweeney , m. a. , 1971 , aj 76 , 544 pavlenko y.v . &
magazzu a. 1996 , a&a 311 , 961 prugniel , p. and soubiran , c. 2001 , a&a 369 , 1048 queloz d. , allain s. , mermilliod j .- c . , bouvier j. and mayor , m. 1998 , a&a 335 , 183 | the cross - correlation between the rosat all - sky survey ( @xmath0 sources ) and the tycho mission ( @xmath1 stars ) catalogs has selected about 14000 stellar x - ray sources ( rastyc sample , guillout et al .
1999 ) .
about 200 - 300 stars have been spectroscopically observed at high resolution both in the h@xmath2 and lii @xmath36708 regions with elodie and aurelie spectrographs of the ohp ( observatoire de lhaute provence , france ) . the aim was to classify the rastyc star sample in terms of age and chromospheric activity level and to detect eventual binary systems .
photometric and spectroscopic follow - up observations of a rastyc sub - sample composed of particularly interesting objects ( binaries and very young stars ) has been performed with the 91-cm telescope of the catania astrophysical observatory . in this work we present some results of this monitoring . in particular , we have obtained good radial velocity curves and solved for the orbits of three sb2 and three sb1 spectroscopic binaries . in addition , for near all sources we have detected a photometric modulation ascribable to photospheric surface inhomogeneities and chromospheric h@xmath2 line variation .
+ stars : binaries : spectroscopic , stars : fundamental parameters , stars : x - ray , stars : activity |
the search of new topological states of matter has never been stopped since the discovery of the quantum hall state in 1980s.@xcite in particular , in recent years the study of `` topological insulators '' ( ti ) becomes one of the most active fields in condensed - matter physics , not only for its importance to fundamental physics but also for its potential application in spintronics or thermoelectrics.@xcite this new insulating phase is distinguished from the conventional one by a non - trivial z@xmath0 topological invariant and robust gapless edge states in two dimensions ( 2d)@xcite or surface states in three dimensions ( 3d)@xcite , against moderate perturbations which preserve time reversal symmetry ( trs ) .
generally , such topological insulating state can occur in a system with strong spin - orbit coupling that explicitly breaks spin rotational symmetry ( srs ) , resulting in the band inversion phenomenon .
typical materials which exhibit ti phase are found in , for instance , the hgte / cdte quantum wells ( 2d ) , bi@xmath1sb@xmath2 , bi@xmath0se@xmath3 , bi@xmath0te@xmath3 ( 3d ) , and so on.@xcite however , an alternative route to ti is possible and it is associated with the concept of topological mott insulator , first studied in ref . [ ] in _ strongly correlated _ systems .
there are two key and generic ingredients in this approach .
first , the fermi surface of the non - interacting system should shrink to discrete points ( 2d ) or lines ( 3d ) , and hence it is semi - metallic ; second , there exists a suitable repulsive interaction , which can dynamically generate spin - orbit coupling through spontaneously broken srs , as first discovered by wu and zhang.@xcite a few pioneering examples along this line of thought have been discussed in various lattice geometries , _
e.g. _ , honeycomb,@xcite checkerboard,@xcite kagome,@xcite and diamond lattices.@xcite here we show that topological states can be generally realized in a system with quadratic band crossing points ( bcps ) , which are symmetry protected at non - interacting level . near such kind of bcp , instability towards phases with broken symmetry symmetries
is inevitable even if there is only weak interaction between electrons . to demonstrate this
, we construct a concrete model with such a bcp on an extended 2d lieb lattice .
there are several reasons for us to choose the lieb lattice .
first , the lieb lattice has three sites per unit cell , as shown in fig .
[ fig : lattice ] . with only nearest - neighbor ( nn )
hoppings , there is a _ dispersionless _ ( flat ) band in the middle of the band structure .
there are three scenarios to alter this non - generic crossing between this flat band and the two other bands , depending on the values of the onsite potentials or the presence of further range hoppings : ( i ) the flat band touches upper and lower linearly dispersing bands at one point when all lattice sites have equal onsite potentials .
( ii ) the flat band can be isolated .
for instance , one can add intrinsic spin - orbit couplings as discussed by weeks and franz.@xcite ( iii ) when the onsite potential on @xmath4 sublattice is not equal to that of @xmath5/@xmath6 sublattices , the flat band only touches one of the other two bands , which becomes quadratically dispersing , instead of linearly dispersing.@xcite thus , these choices could variegate our results .
second , a nearly flat band can have non - zero chern number ( see sec .
finally , the 2d lieb lattice has been the most important building block in many 3d perovskite materials featured with complex phase diagram and strongly electron - electron correlations .
thus study of the model can be viewed as a preliminary investigation of the ti phase especially in layered perovskites composed of weakly coupled 2d planes with lieb lattice structure ( _ e.g. _ , the well - known high-@xmath7 cuprates ) . in this paper , we start with the construction of the explicit model and reveal the topological nature of the bcp at non - interacting level .
we then examine the consequence of such topological bcp , _
i.e. _ , with a symmetry protected berry flux @xmath8 , under the presence of short - range repulsive interactions .
we investigate various symmetry breaking instabilities at bcp within self - consistent mean - field approximation .
note that we mainly focus on type ( iii ) band structure , namely , only two bands touch together , and compare it with the case of type ( i ) when necessary . in principle , for the bcp there are two ways to gain energy : one is to open a gap at bcp , and the other one is to split into two dirac points ( each with berry flux @xmath9 ) , but with the price of broken @xmath10 symmetry . to justify this speculation
, we show phase diagrams for spinless / spinful fermions at 1/3 or 2/3 filling , according to the position of the bcp in the band structure .
in fact , at both fillings the phase diagrams are qualitatively similar with subtle differences due to particle - hole asymmetry introduced by the interactions . in the spinless case ,
the leading order under `` weak '' next nearest - neighbor ( nnn ) repulsion is the quantum anomalous hall ( qah ) insulating state ( trs broken ) . for `` strong '' nnn repulsion
, the ground state evolves into insulating nematic state ( @xmath10 symmetry broken down to @xmath11 ) .
in addition , for intermediate strength , there exists a narrow coexistence region between these two orders . in the spinful case ,
the phase diagrams are more complicated .
besides the phases we find in the spinless case , there are also a quantum spin hall ( qsh ) insulating state and a nematic - spin - nematic semi - metallic phase with dirac nodes.@xcite thus , we clearly demonstrate that , in principle , correlated systems with lieb lattice structure can be a potential host to tis .
this paper is organized as follows . in section
ii , we define the model and demonstrate the topological nature of its bcp from both momentum and real - space point of view . next , the consequences of introducing short - range repulsions are discussed for spinless fermions in section iii a and for spinful fermions in section iii b , respectively .
finally , we discuss some issues and make conclusions in sec iv .
we begin with the tight - binding model for non - interacting fermions , _ 0=-_ijt_ijc^_ic_j + _ i_i c^_ic_i , [ eq : h0 ] where @xmath12 creates a fermion on site @xmath13 of the 2d lieb lattice ,
the unit cell of which is given by @xmath14 sites shown in fig .
[ fig : lattice ] . for simplicity , we take the hopping amplitudes , @xmath15 , between nn sites @xmath16 , and @xmath17 otherwise for the moment . the effect of adding longer - range hopping amplitudes ( but small in magnitude ) will be discussed later when appropriate .
note that the @xmath10 point group symmetry dictates that the onsite potentials on the @xmath5 and @xmath6 are equal , @xmath18 .
although @xmath19 are generically non - zero , only their relative values are essential to determine the symmetry of the lattice , and hence , the band structure . therefore , hereafter we set the units of energy @xmath20 , the lattice constant @xmath21 , and , without loss of generality , @xmath22 . the band structure of eq .
( [ eq : h0 ] ) can be obtained by transforming @xmath23 into momentum space , _ 0=_^_h_0 ( ) _ , [ eq : h0 ] where the fermion spinor , @xmath24 , with sublattice ( basis ) index @xmath14 and @xmath25 . defining the displacement vectors , @xmath26 and @xmath27 ,
@xmath28 is of the form h_0()= ( [ c]ccc_a & -2 t ( _ 1 ) & -2 t ( _ 2 ) + & 0 & 0 + & & 0 ) , where the lower triangular matrix is understood to be filled for keeping the whole matrix hermitian . in this notation ,
the first brillouin zone ( fbz ) is a square with four time reversal invariant momenta ( trim ) : @xmath29 , and @xmath30 [ see the inset of fig .
[ fig : spectrum](a ) ] .
the energy spectrum consists of two dispersive bands , @xmath31 with @xmath32 ^ 2 $ ] , and one _ dispersionless _ flat band , @xmath33 .
an important feature of this model is that the presence or absence of @xmath34 can change electronic properties dramatically .
when @xmath35 , the flat band touches two linearly dispersing bands at @xmath36 point in the fbz [ type ( i ) ] , where the linear bands meet as if there were a `` dirac point '' .
however , the touching point in fact has completely different structure .
it becomes clear once we expand @xmath28 around @xmath36 point with @xmath37 .
to the first order in @xmath38 , @xmath28 can be written as h_0()~v_f , [ eq : spin1h0 ] where the fermi velocity @xmath39 , @xmath40 , and the ( pseudo ) spin-1 matrices are defined as l_x= ( ccc 0 & 1 & 0 + 1 & 0 & 0 + 0 & 0 & 0 ) , l_y= ( ccc 0 & 0 & 1 + 0 & 0 & 0 + 1 & 0 & 0 ) , l_z= ( ccc 0 & 0 & 0 + 0 & 0 & -i + 0 & i & 0 ) , obeying lie algebra of su(2 ) , i.e. , @xmath41=i\epsilon_{ijk}l_k$],@xcite instead of a clifford algebra as in the case of graphene .
this is the fundamental reason why there is no dirac point and hence no fermion doubling problem@xcite on the lieb lattice .
viewing the low - energy effective hamiltonian @xmath42 as a `` spin '' @xmath43 in an external `` magnetic field '' @xmath44 , its eigenvalues can be easily read out as @xmath45 , where @xmath46 are the quantized angular momenta along the axis parallel to @xmath44 in three dimensions .
space : ( a ) the planar vector field @xmath47 ; ( b ) the planar vector field @xmath44 . ] when @xmath48 , however , the spin-1 structure mentioned above is no longer valid .
the flat band now touches only one dispersive ( massive ) energy band either above or below at @xmath36 point , depending on the sign of @xmath34 [ see fig . [
fig : spectrum ] ( b ) ] . to make this structure transparent , we again expand @xmath28 around @xmath36 with small @xmath38 .
assuming @xmath49 and @xmath50 at 2/3 filling , we then integrate out the contribution from basis @xmath4 ( due to almost fully filled @xmath4 sublattice ) and obtain a low - energy effective two - band hamiltonian , h^eff_0()~ ( cc p_x^2 & p_x p_y + p_x p_y & p_y^2 ) = d_i i+d_x_x+d_z_z , [ eq : heff ] where @xmath51 .
in the last equality , we express @xmath52 in terms of the identity and pauli matrices with @xmath53 , @xmath54 , and @xmath55 .
interestingly , if we further allow small third - neighbor hoppings @xmath56 ( but _ forbid _ to hop whenever there is a site in the middle of the path ) , the flat band becomes slightly dispersive and the effective hamiltonian changes to @xmath57 , @xmath54 , and @xmath58 without removing the band crossing point ( bcp ) [ see fig .
[ fig : spectrum ] ( b ) ] .
such point at @xmath59 is the so - called quadratic band crossing point ( qbcp ) , which has been studied recently by several research groups.@xcite one of the key features for qbcp in 2d is that its density of states ( dos ) is non - zero at the crossing point , in sharp contrast to the case of dirac points .
this will lead to essential difference when responding to the weak interactions present in the system . in the following ,
we will mainly focus on the @xmath60 case and show that the bcp in our model is not only topologically non - trivial , but also make the system be a potential host to topological phases under weak repulsive interactions .
the band touching phenomenon on the lieb lattice is quite generic and stable for non - interacting fermions .
such stability deserves a full analysis here .
we shall provide two different approaches to show it : one is based on momentum - space topology , and the other one is based on real - space topology . from the first point of view
, the bcp actually forms a topological defect in the momentum space , similar to a vortex in a 2d superconductor , here with a winding number @xmath61 . to see this ,
let us rewrite eq .
( [ eq : heff ] ) as @xmath62 , where the `` magnetic field '' @xmath63 .
this effective hamiltonian now represents a spin-1/2 particle sitting in a magnetic field @xmath47 , which has a vortex structure at @xmath59 , as shown in fig .
[ fig : vortex](a ) . for comparison ,
recall that for @xmath35 we instead have a spin-1 particle in an external field @xmath44 [ eq . ( [ eq : spin1h0 ] ) ] ,
whose structure is shown in fig .
[ fig : vortex](b ) .
the winding number @xmath64 then can be easily extracted from the figures that in the former case , @xmath65 ; in the latter case , @xmath66 .
however , somewhat counter intuitively , both cases are associated with the same berry phase of the bcp,@xcite which can be calculated precisely by b^n = i_du_n|_|u_n , [ eq : berry ] where @xmath67 is a contour in the @xmath38 space enclosing the touching point , @xmath68 denotes any one of the involved bands , and @xmath69 represents the bloch wave function for @xmath68th band .
a simple argument solves this puzzle .
the line integral along any loop enclosing 0 in @xmath38 space given above is known to be 1/2 ( 1 ) times the solid angle subtended by @xmath70 [ @xmath71 from the origin for a spin 1/2 ( 1 ) particle .
thus , @xmath72 in the former case , which is just equivalent to @xmath73 in the latter one .
in fact , when @xmath74 , any infinitesimal mixing ( perturbation ) between bands would lift the degeneracy . with non - vanishing @xmath75 , we confirm that the bcp on the lieb lattice at non - interacting level is topologically stable ( i.e. , not opening a gap ) as long as the spinless system preserves both trs and @xmath10 point group symmetry .
note that @xmath10 symmetry in our model is quite essential , as a similar qbcp happens in the a - b stacking bilayer graphene ( with @xmath76 symmetry ) while it can easily decay into dirac bcps and thus is topologically unstable.@xcite on the lieb lattice ( @xmath48 ) .
only those sites in the shaded area have non - zero weights , whose magnitudes are denoted by numbers ( up to a normalization factor ) . the weights on @xmath4 sublattice ( blue dots ) are all zero . ]
an alternative point of view for the protection of such bcp on the lieb lattice can be associated with certain topological structure present in the real space , or more specifically , with the existence of the eigenstates which are extended along _ non - contractible _ loops winding around the whole lattice with periodic boundary conditions ( _ i.e. _ , a torus).@xcite to demonstrate this feature , we first take the merit of the flat band , which allows us to construct its corresponding localized , one - particle eigenstates of @xmath23 ( wannier states ) . taking @xmath77 to be the coordinate of the central site of the shaded plaquette shown in fig .
[ fig : wannier ] , we find that the creation operator for the localized eigenstate at @xmath77 can be written as ^_= [ _ j=1 ^ 4(-1)^j(2c^_+_j -c^_+_2j+3 -c^_+_2j+4 ) ] , [ eq : wannier ] where @xmath78 , @xmath79 , @xmath80 . the key reason for these states being localized is rooted on the fact that all a - sites have vanishing amplitudes and remain zero after the action of @xmath23 on them due to destructive interference .
direction ( a ) and @xmath81 direction ( b ) .
only those sites in the shaded area have non - zero weights ( denoted by numbers in the plots ) .
the weights on @xmath4 sublattice ( blue dots ) are all zero . ] in localized - state language , the existence of the bcp in our model with @xmath48 is equivalent to state that the dimension of the space expanded by independent localized eigenstates with zero energy has a dimension which is one larger than the number of unit cells , @xmath82 .
the extra state can not come from the flat band , but from one of the dispersive bands .
the plaquette states we constructed in eq .
( [ eq : wannier ] ) seem to form @xmath82 linearly independent states with zero energy . for our model with periodic boundary conditions
, however , the following relation , ^_=(,)= _ e^i^_=0 , reduces the naive counting by one and hence only @xmath83 states are independent .
the missing two states , in fact , are accounted for by two _ non - contractible _ loops around the whole lattice ( torus ) , as illustrated in figs .
[ fig : ncloop](a ) and ( b ) .
when @xmath23 acts on these states , the destructive interference again guarantees the zero eigenvalue .
now , in total we have @xmath84 independent states . therefore , provided not destroying the flat band , such band touching phenomenon is protected by the topological character of the lattice .
the existence of such symmetry / topology protected bcp on the lieb lattice at the non - interacting level motivates us to further ask if it is stable in the presence of repulsive interactions . to see this
, we will first examine whether generic short - range repulsions are relevant to this bcp from perturbative renormalization group ( rg ) analysis , and next , if the interactions are relevant , we will investigate possible consequences of such instability , i.e. symmetry breaking phases , at mean - field ( mf ) level . to perform rg analysis , we consider a continuum , spin-1/2 hamiltonian , which can be obtained by projecting the original three - band interacting model onto an _ effective two - band theory _ near the bcp in the continuum and @xmath85 limit , ^eff=^eff_0+^eff_int , where the free part @xmath23 reads as ^eff_0=_d^_()h^eff_0 _ ( ) .
the subscript @xmath86 denotes spin polarization and the fermion field @xmath87 with 1,2 representing orbital ( i.e. , two touching bands ) degrees of freedom .
note that in the momentum space the expression for @xmath88 is given by eq .
( [ eq : heff ] ) , and is independent of @xmath86 . for simplicity ,
we take small @xmath89 such that @xmath90 , making the effective theory particle - hole symmetric . in fact , non - vanishing @xmath91 would not change our main conclusion , provided @xmath92 .
the projected interacting part includes only 1 ) intra - orbital and 2 ) inter - orbital contact interactions , ^eff_int&= & ^eff_1+^eff_2 , + ^eff_1 & = & _ a=1 ^ 2ud^_a ( ) ^_a ( ) _ a ( ) _ a ( ) , + ^eff_2 & = & _ , ^g_^d ^_1()^_2^ ( ) _ 2^ ( ) _ 1 ( ) , where @xmath93 and @xmath94 are intra - orbital and inter - orbital coupling parameters , respectively . for the chemical potential @xmath95
, the non - interacting system @xmath96 leads to one fermi point at @xmath59 with non - vanishing dos , instead of a fermi surface . setting the dimension @xmath97=1 $ ] and understanding the dynamical critical exponent @xmath98 due to quadratic dispersion , it is straightforward to see that the dimension @xmath99=1 $ ] ; in the interacting part the coupling constants @xmath100=[g_{\sigma\sigma^\prime}]=0 $ ] , implying that they are superficially _ marginal _ interactions .
as shown in appendix a , we find that they are generically marginally relevant and bring the system to the strong coupling regime .
more explicitly , up to one - loop order , the coupled rg equations for the coupling parameters are & = & ( 2-)u^2 + 2g_u-2g_g _ , + & = & g^2_-2g_u+g^2 _ , + & = & ( -2)g^2_-2g_u+2g_g _ , [ eq : rgeq ] where we have used the fact that @xmath101 and @xmath102 .
@xmath103 denotes the momentum rescaling @xmath104 and the coefficients @xmath105 and @xmath106 , with @xmath107 [ @xmath108 , the elliptic function of first ( second ) kind and @xmath109 .
in fact , no new fixed point ( fp ) is produced in this set of rg equations , except for the non - interacting one at which @xmath93 and @xmath94 vanish .
moreover , we find that given generic bare coupling parameters ( @xmath110 ) , at least one of them diverges first when reaching low enough energy scale .
this indicates that the non - interacting fp is an _ unstable _ fp , which can drive the system to strong coupling regime in the presence of short - range repulsions .
in addition , it is worth mentioning that by setting @xmath111 , we reduce the rg equation back to the spinless case , = g^2 _ with @xmath112 , which is consistent with the work done in ref . and ,
importantly , it implies that short - range repulsions are again _ marginally relevant_. from above we know that a qbcp is generally unstable against weak repulsive interactions .
we now discuss its consequence on the lieb lattice and explore possible symmetry breaking phases driven by interactions at mean - field level .
we warm up with the spinless case to gain some physical insights before including spin degrees of freedom .
the lattice model we study is given by eq .
( [ eq : h0 ] ) , with short - range interacting terms , _
int= v_1_ij n_i n_j+v_2_ij n_i n_j , [ eq : int ] where @xmath113 and @xmath114 are repulsive coupling constants for nn and nnn interactions , respectively .
@xmath115 is the number operator on the site @xmath13 .
the chemical potential is suitably chosen to keep the system at 2/3 ( 1/3 ) filling for @xmath50 ( @xmath116 ) .
we proceed by treating @xmath117 in the mf approximation , including both the on - site and bond mf decoupling particle - hole channels , n_i n_j & & n_in_j+ n_jn_i- n_in_j , + n_i n_j & & -_ijc^_j c_i - ^*_ij c^_i c_j + _ ij^*_ij , where @xmath118 represent certain current / bond order with @xmath119 belonging to nn and nnn bonds . note that in this work only _ translation - invariant _ mf ansatz is considered .
the repulsive interactions can produce the following possible phases : \(i ) _ nematic state_. this is a phase associated with broken @xmath10 symmetry down to @xmath11 .
it behaves like an anisotropic metal ( one qbcp splits into two dirac points ) or an insulator ( two dirac points meet at zone boundary and end up with a gap ) , depending on the strength of repulsive interaction .
there are two types within this phase .
type i is `` site '' nematic with order parameter , @xmath120 , where @xmath121 denoting four nnn bonds of b - site and @xmath14 are sublattice indices . without loss of generality , at 2/3
filling we can set the charge density , @xmath122 , @xmath123 , and @xmath124 . as @xmath116 , at 1/3
filling we simply replace 2/3 by 1/3 in the above expression .
note that the use of the parameter @xmath125 is to take into account the renormalization of the onsite potentials due to interactions .
the nonzero expectation value of it does not break any symmetry of the model .
on the other hand , type ii is `` bond '' nematic with order parameter either in the form of @xmath126 $ ] , or in the form of @xmath127 for @xmath128 and @xmath129 .
the subscript of the order parameters indicates their origin of either @xmath113 or @xmath114 repulsion .
and others are zero ( qah state ) ; ( b ) @xmath130 and others are zero ( varma @xmath131 loop)(c ) @xmath132 ( varma @xmath133 loop ) ; ( d ) @xmath134 . ]
\(ii ) _ current loop state .
_ this type of phase is featured by spontaneously trs breaking . the most probable current patterns which preserve translational invariance with no ( charge )
source and drain present on the lattice sites are shown in figs .
[ fig : current](a)-(d ) .
each state basically comes from non - vanishing imaginary part of certain bond orders in the mf decouplings and may behave hall - insulating [ 3(a ) ] , semi - metallic [ 3(b),3(d ) ] , or insulating [ 3(c ) ] .
in particular , the most significant one is case ( a ) , which exhibits quantum anomalous hall ( qah ) effect with order parameter , @xmath135 $ ] .
topological _ state is known to be characterized by quantized hall conductance without landau levels ( or equivalently , by non - zero chern number ) and has topology - protected , gapless chiral edge modes.@xcite we compute the chern number for each band within this state and find that 1 ) for @xmath136 , the previous two touching bands now carry chern numbers @xmath137 separately .
in particular , one of two bands is ( nearly ) dispersionless .
the third one simply carries zero ; 2 ) for @xmath35 , the middle flat band carries zero chern number , while the upper and lower bands carry @xmath138 , respectively .
the other possible current loop states , however , are not topological insulating . for case ( b ) ( varma @xmath131 loop state@xcite ) , it is semi - metallic with order parameter given by , simultaneously , @xmath139 $ ] and @xmath140 .
for case ( c ) ( varma @xmath133 loop state@xcite ) , it is insulating with broken inversion symmetry ( is ) as well but is invariant under the combined trs and is .
thus , there is no hall or uniform kerr response by noticing that for any given momentum @xmath141 it changes sign under trs or is.@xcite this order can be described as , @xmath142 , where @xmath143 $ ] , @xmath144 $ ] , and @xmath145 $ ] .
finally , case ( d ) is a semi - metallic state with broken is and hence no net hall current in it .
its order is described as @xmath146 , where @xmath147 $ ] . in momentum space , the mean - field hamiltonian at 2/3 filling can be now written in the matrix form , _
mf=_^_(h_-i)_+e_0 , with the fermion spinor , @xmath148 , and e_0 & = & - - + & + & 4nv_2(t^2 + q_2 ^ 2+_2 ^ 2+_2 ^ 2+_2 ^ 2+_2
^ 2 ) + & + & 4nv_1(t^2+q_1 ^ 2+_1 ^ 2+^2_1+^2_1 ) , where @xmath149 ( @xmath150 ) represents a renormalization due to nn(nnn ) repulsions .
the hamiltonian matrix @xmath151 reads h_= ( [ c]ccc |_a1+_a & _ x & _ y + & |_b1+|_b2+_y & _ xy + & & |_c1+|_c2+_x , ) , where @xmath152 , @xmath153 , @xmath154 , @xmath155 , @xmath156 , and @xmath157 ; @xmath67 parameters are given by _ x , y & = & ( -2t2v_1(q_1-i_1 ) ) + & - & 2v_1(^_1^_1 ) , + _ xy & = & 4iv_2_2^ + 4v_2_2^ + & - & 4v_2(q_2-i_2 ) + & - & 4v_2_2^. thus , the mean - field free energy can be expressed as f=-_(1+e^-(e_-))+e_0 , [ eq : free_energy ] where @xmath158 and @xmath159 are eigenvalues of @xmath151 . the ground state with given coupling parameters
can then be determined by minimizing the free energy with respect to each order parameter , yielding a set of coupled gap equations .
notice that at 1/3 filling we follow the same procedure and only the diagonal part of @xmath151 and @xmath160 need to be changed accordingly due to the shift of the average charge density . ,
@xmath161 and on the ( b ) 2/3-filled lattice with @xmath162 , @xmath163 .
the shaded area represents the coexistence region . ]
we numerically solve the coupled gap equations self - consistently and obtain the zero - temperature @xmath113-@xmath114 phase diagrams at both 1/3 ( @xmath164 ) and 2/3 ( @xmath165 ) fillings , as seen in figs .
[ fig : spinlesspd](a ) and ( b ) .
note that in this study , only weak short - range repulsions , i.e. , @xmath166 are considered . at 1/3 filling
, we find that there are three phases in the absence of @xmath113 : qah phase , coexisting qah+nematic phase , and nematic phase . beginning with @xmath167
, the infinitesimal instability of qbcp leads to qah phase by the second order phase transition , with a @xmath168 gap , @xmath169 , where @xmath170 denotes the finite dos at qbcp and @xmath171 is an energy cutoff ; on the other hand , for @xmath172 the ground state breaks @xmath10 symmetry spontaneously down to @xmath11 and exhibits insulating nematic phase with a gap @xmath173 . in this phase , we find that the site - nematic order ( @xmath174 ) is the dominant one and a small component of the bond - nematic order ( @xmath175 ) accompanies with it .
in fact , we notice that the bond - nematic order can not be induced by @xmath113 itself .
finally , with intermediate value of @xmath114 , there exists a narrow window for the coexistence of both qah and nematic orders .
in addition , we also notice that there is no room for the current loop states other than qah state for systems with relatively large @xmath176 and weak repulsions . at 2/3
filling , the phase diagram is qualitatively similar to that at 1/3 filling .
however , there are a few remarks worth mentioning here : 1 ) although the non - interacting energy spectrum for both fillings can be related by translating `` particle '' into `` hole '' language , which causes @xmath177 , @xmath178 , @xmath179 , and @xmath180 . the interactions given in the present form ruin such relation and hence
the two phase diagrams must be different.@xcite 2 ) the response to the weak @xmath113 repulsion is different for the two cases . at 1/3 filling with @xmath181 , since the charge density on @xmath4 sublattice is much less than 1/3 , the stronger @xmath113 is , the more energy cost the system suffers from @xmath114 .
thus , the region of the qah phase reduces and the ground state evolves into the nematic phase faster than the case with @xmath182 .
in contrast , at 2/3 filling with @xmath183 , the presence of small @xmath113 can only reduce the charge density slightly on @xmath4 sublattice , and therefore its effect on the energy cost from @xmath114 is relatively small .
3 ) when @xmath184 , the spin-1 structure near band touching point , as discussed in the previous section , is recovered .
our mean - field study shows that the infinitesimal instability ( near 1/3 or 2/3 filling ) is absent due to the vanishing dos of the dispersive bands and the semi - metallic phase is robust until @xmath114 reaches certain critical value .
moreover , for @xmath185 , we find that the qah phase only survives in a negligible window of @xmath114 , and the nematic phase becomes the dominant one in the phase diagram ( not shown ) .
this result is similar to the work done by q. liu _
et al._@xcite on the 2/3-filled kagome lattice with dirac bcps .
we now take the spin degrees of freedom into account .
the model hamiltonian again consists of the free and interacting parts , i.e , @xmath186 . the free part is again given by eq .
( [ eq : h0 ] ) with extra spin index @xmath86 in the fermion creation / annihilation operators ; the interacting terms now contain _ int & = & _ ij_,^v_1n_i , n_j,^+ _ ij_,^v_2n_i , n_j,^ + & + & _ iu_in_i , n_i , .
in addition to @xmath113 ( @xmath114 ) denoting the coupling constant of nn ( nnn ) repulsion , we also consider the repulsive hubbard ( @xmath187 ) terms . for simplicity
, we will assume uniform onsite repulsions , i.e. , @xmath188 .
different from the spinless case , there are not only spin-_singlet _ order parameters ( as we had before ) , but also spin-_triplet _ order parameters within mean - field approximation .
the possible phases under _ translation - invariant _ ansatz are classified below : \(i ) _ charge nematic state _ ( cn ) .
this phase is associated with spontaneously rotational ( @xmath10 ) symmetry breaking .
one can either have the site - nematic or bond - nematic state , whose order parameter is a spin - singlet and simply the same as that for spinless fermions with additional summation on spin @xmath86 times a normalization factor 1/2 .
note that for the site - nematic case , the driving force is now from both @xmath189 and @xmath114 terms , combined together to give out an effective nnn repulsion , @xmath190 , playing similar role of @xmath114 in the spinless model .
, @xmath163 , and @xmath182 .
the short - hand notations represent : nsn = nematic - spin - nematic ; qah = quantum anomalous hall ; qah = quantum anomalous hall ; cn = charge nematic .
the shaded area denotes the coexistence region . ]
\(ii ) _ nematic - spin - nematic state _ ( nsn ) .
this phase breaks @xmath10 symmetry in the spin sector , _ not _ in the charge sector .
consequently , it turns the ( spin ) doubly degenerate qbcp into four dirac points ( two pairs with opposite spin polarizations ) , and @xmath10 symmetry of the band structure remains intact .
similar to its charge counterpart in ( i ) , there are two types : one is the site - nsn with a spin - triplet order parameter , @xmath191=@xmath192 , where @xmath193=@xmath194 with @xmath195 and @xmath196 , the pauli matrices .
note that this phase can occur simply due to the presence of the hubbard term , which provides a spin - triplet channel , @xmath197 with @xmath198 representing usual spin operator .
the other one is the bond - nsn , which is described by @xmath199=@xmath200 [ @xmath201-@xmath202 $ ] or @xmath203=@xmath204 ( breaking @xmath10 along a diagonal direction of the lattice ) .
\(iii ) _ charge current - loop state with broken trs_. as mentioned in the spinless model , among all the current patterns the case ( a ) in fig .
[ fig : current ] is of most interest .
this state , characterized by spontaneously broken trs and parity symmetry , exhibits qah effect with a spin - singlet order parameter , @xmath205 $ ] .
mainly driven by @xmath114 terms , fermions with opposite spin polarizations flow in the same way and hence provide the same flux pattern , penetrating the whole lattice .
in addition , we investigate the possibility for the other current - loop states [ e.g. , from fig .
[ fig : current](b)-(d ) ] and find that none of them is stabilized by the presence of short - range repulsions in our mf study
. therefore , we will only consider case ( a ) hereafter .
\(iv ) _ spin current - loop state with trs_. the key difference of this phase from its charge counterpart ( qah ) is trs unbroken .
one can view it as a combined double - layer qah system : fermions of opposite spin polarizations , residing in different layers , producing just opposite flux patterns separately .
thus , for the whole system trs is preserved .
in fact , this is known as quantum spin hall ( qsh ) phase with spin - triplet order parameter described by @xmath206 $ ] .
two remarks deserve mentioning here .
first , both qah and qsh are _ topological _ phases , characterized by non - trivial topological index with robust edge states . however , the former one acquires non - vanishing chern number , while the latter one has zero chern number due to trs .
thus , a new topological index , called z@xmath0 index @xmath207 , needs to be introduced.@xcite as detailed in appendix b , the qsh phase on the lieb lattice indeed acquires non - trivial @xmath208 .
second , it is straightforward to see that at mf level , the energy spectra for both qah and qsh are the same . as a result , they have equal energy gain from @xmath114 repulsion and hence one can not distinguish them in the mf phase diagram .
if there were an extra nnn exchange coupling @xmath209 present in the system , the qah would be favored for @xmath210 ; reversely , the qsh would be favored for @xmath211 due to its spin - triplet nature . under our assumption of translational invariance within our mean - field study
, we do not consider any charge or spin density wave order .
however , it is still worth pointing out that if @xmath212 and hence makes the @xmath4 sublattice be effectively decoupled from rest of the lattice sites , at large @xmath189 the ( 0,0 ) antiferromagnetic order could be realized at 1/3 ( 2/3 ) filling with @xmath116 ( @xmath50 ) , as guaranteed by the lieb s
theorem.@xcite following the same procedure as in the spinless case , we decouple the interacting terms within mf approximation and obtain the mf free energy in similar form of eq .
( [ eq : free_energy ] ) , where we have used the mf ansatz , @xmath213 , @xmath214 , and @xmath215 for the 2/3-filled lattice . note that , for simplicity , we have assumed that spin points to @xmath216-direction after su(2 ) symmetry breaking . such treatment will be applied to other spin - triplet order parameters as well .
@xmath160 in the spinful case becomes & = & - -+8v_1(t^2+q_1 ^ 2+q_1z^t2 ) + & + & 8v_2(t^2 + q_2 ^ 2+_2 ^ 2+q_2z^t2+_2z^t2 ) + & + & ( s^2_az+s^2_bz+s^2_cz+8^t2_z ) + & - & u [ + ] , where @xmath217 , and @xmath218 are defined as before .
@xmath159 are eigenvalues of the mf hamiltonian , which is now a @xmath219 matrix . by minimizing free energy with respect to various order parameters , the @xmath168 mean - field @xmath189-@xmath114 phase diagram at 2/3 filling ( @xmath165 ) for spinful fermions
is shown in fig .
[ fig : spinfulpd ] .
we first notice that in the absence of @xmath189 there again exist three phases : qah / qsh phase , coexisting qah / qsh+cn phase , and cn phase from weak to strong @xmath114 repulsion .
in particular , the fact that topological qah / qsh phase can arise from infinitesimal instability of qbcp further justifies the interaction - driven scenario as a promising way for producing ti . in the presence of @xmath189
, however , nsn phase begins to compete with the topological phase and clearly dominates over qah / qsh whenever @xmath220 . on the other hand , as @xmath221 ,
the insulating cn phase takes over the phase diagram and this is consistent with the result shown in the spinless model .
two remarks are worth mentioning here .
the first one is about the effect of nn repulsion @xmath113 . as in the spinless case , at 2/3 filling the phase diagram is not sensitive to the presence of @xmath113 .
especially , @xmath113 itself does not lead to any order .
however , when @xmath113 becomes stronger , it is quite possible that the system might gain certain energy by opening a gap due to translation symmetry breaking , and hence beyond our current consideration .
the second point is about the system at 1/3 filling with @xmath164 .
in fact , the phase diagram is qualitatively similar to that of 2/3 filling and therefore we omit it without further discussion .
the model we have solved on the lieb lattice demonstrates that the ti ( qah / qsh state ) can be induced by appropriate interactions through spontaneously symmetry breaking mechanism , which dynamically generates spin - orbit couplings necessary for a topological insulator .
it is then natural to ask how it can be realized experimentally .
we notice that in the interaction - driven scenario there are two key conditions that a system has to fulfill : 1 ) the band structure should contain suitable band crossing point at which two touching bands have opposite curvatures .
2 ) the system should have weak ( or no ) spin - orbit coupling [ spin su(2 ) symmetry is preserved ] and nnn repulsions need to be more significant than the other short - range repulsions . while the condition 2 ) is tricky and we have to reserve it for future investigation , we would like to comment on some possible routes for condition 1 ) below .
first of all , the most promising candidate , we believe , is from cold atom system . as discussed by goldman et al . in ref . , the lieb lattice may be constructed as an optical lattice created by properly arranged laser beams .
in particular , the spin - orbit interaction , which might be an issue in traditional materials , now becomes irrelevant .
another potential way for realizing lieb lattice may come from layered perovskites .
a well - known example is the cuo@xmath0 plane in high-@xmath7 cuprates such as la@xmath2sr@xmath222cuo@xmath223 or yba@xmath0cu@xmath3o@xmath224 , whose electronic structure might be captured by a three - band model with @xmath225 and @xmath226 orbitals . here ,
for illustration purpose we take a typical three - band ( emery ) model , written in _ hole _ language , having the same form of eq .
( [ eq : h0]).@xcite the corresponding model parameters are given by hybertsen et al.@xcite : @xmath227 , @xmath228 , @xmath229 , and @xmath230 , where @xmath231 denotes nnn hopping .
the band structure in the fbz along high symmetry lines is presented in fig .
[ fig : spectrumreal](a ) .
as one can see , there is indeed a bcp at @xmath36 , but with `` wrong '' curvatures for two touching bands . as a result
, cuprates may be impractical to produce topological phase as we desire . to overcome this issue , we offer two speculative suggestions .
the first way out is to add decorated elements with suitable orbital nature ( e.g. , @xmath232-orbital ) between atoms on @xmath5 and @xmath6 sites that could change the sign of @xmath231 .
this change leads to the band structure shown in fig .
[ fig : spectrumreal](b ) with `` right '' curvatures now . keeping the same orbital characters ,
the second way is to search for a new system among perovskite materials , whose model hamiltonian is similar to that given by sun et al.s recent work in appendix f.@xcite
the key feature of such three - band model is that the dominating transfer integrals are now associated with distance @xmath233 after hopping .
for instance , if , by certain geometric reason , the relevant orbital on @xmath5 ( @xmath6 ) sublattice becomes @xmath234 ( @xmath235 ) , instead of @xmath235 ( @xmath234 ) as in the cuprates , the above consideration could be plausible . before concluding our work
, we would like make a brief remark for the possible new physics brought by the ( nearly ) flat band appearing in the lieb lattice .
consider @xmath116 and only nn hoppings in our spinless model . now , the middle band is completely flat and touches with the lower dispersive band at @xmath36 in fbz . at 1/3 filling
if we turn on @xmath114 , the ground state would enter qah state with an energy gap @xmath236 opened at @xmath36 . as we have mentioned in sec .
iii a , the middle and lower bands acquire chern numbers @xmath138 , while the upper one has zero chern number in this phase .
moreover , the flatness of the middle band is approximately determined by the ratio of band - gap to bandwidth @xmath237 . in other words , with appropriately chosen @xmath238 , one can produce a nearly flat band with non - trivial chern number .
this result is quite significant since it may provide an opportunity to realize _ fractional _ qah state ( or _ fractional _
z@xmath0 ti in the spinful case ) when such band is partially filled.@xcite in conclusion , we have studied interacting spinless / spinful fermions on the ( extended ) lieb lattice and have explored the possibilities of various spontaneously broken symmetries associated with a bcp in the band structure .
due to the topological nature of the bcp , namely , with berry phase @xmath8 , we have seen that in the @xmath168 phase diagram the system can exhibit topological qah / qsh , nematic , and nematic - spin - nematic phases , depending on the strengths of short - range density - density interactions . in particular , the existence of ti phase firmly justifies the interaction - driven scenario .
moreover , for a quadratic bcp ( as @xmath239 ) , only weak interaction is necessary for inducing the ti phase , which is in sharp contrast to the systems with dirac points .
in addition , in our model there exists a flat band , which is interesting on its own right and we argue that in principle , one can obtain a nearly flat _ topological _ band without external magnetic field , a starting point to realize exotic correlated phases of matter .
still , there are many open issues and they deserve further investigation . for instance , one could consider the effect of the chemical potential away from the bcp or the instability to superconductivity from both repulsions and attractions , with special focus on the possibility of any topological nature . finally , we hope our work can further stimulate people to search for a new family of tis in the perovskite - related materials , where , due to the presence of transition elements , the strong correlations may not be ignored .
we thank l. fu , h. lin , c .- k .
lu , y. ran , k. sun , and c. xu for helpful discussions and , especially , for k. sun s early collaboration in this work .
wft would like to thank for the hospitality of iop , chinese academy of sciences during his visit , where part of the present work was done .
this work was supported in part by the nsc in taiwan with grant no .
100 - 2112-m-110 - 001-my2
. _ note added . _ after this work is completed , we were informed by steve kivelson that there is a partly related paper by m. h. fischer and eun - ah kim , arxiv:1106.6060 .
in this section , we derive rg equations given in eq .
( [ eq : rgeq ] ) in path integral formulation and show the rg flows for two typical cases .
we begin with defining the action and the shorthand notations below . at zero temperature ,
the full action is given by s = s_0+s_int , where the free action @xmath240 reads s_0=_|_(,)[^_0 ( , ) ] ^-1 _ ( , ) , in momentum space .
the grassmann variables , @xmath241 with 1,2 ( labeled by ` @xmath233 ' for later use ) representing orbital degrees of freedom .
as mentioned in the main text , for simplicity , we consider the inverse of the non - interacting green s function @xmath242^{-1}=i\omega i -d_x\sigma_x - d_z\sigma_z$ ] , which is suitable for a particle - hole symmetric qbcp with @xmath54 , @xmath243 , and @xmath244 and is independent of spin polarization @xmath86 .
we will fix @xmath245 as our unit in the analysis .
the action of interactions can be written as , s_int&= & s_1+s_2 , + s_1 & = & -_a=1 ^ 2du|_a(4)|_a(3)_a(2 ) _
a(1 ) , + s_2 & = & -_^dg_^ |_1(4)|_2^(3)_2^(2 ) _
1(1 ) , where @xmath246 and the shorthand notation @xmath247 . note that `` 4=1 + 2 - 3 '' is understood by energy and momentum conservation and we ignore the momentum dependence of @xmath93 and @xmath94 , which turns out to be irrelevant in rg sense .
in addition , it is worth mentioning here that the present rg analysis is much simpler than that of usual fermi liquids in the sense that there is _ no _ fermi surface ( @xmath248 ) .
this fact thus waives the complexity brought by the fermi surface , a situation similar to @xmath249-theory .
let us now sketch how to obtain one - loop rg equations , following the standard perturbative rg procedure by shankar.@xcite defining a momentum cutoff @xmath171 and @xmath250 ( @xmath251 ) as slow ( fast ) modes with @xmath252 ( @xmath253 ) , the key formula we use to derive the renormalized action by integrating out the fast modes is given by z & = & [ |_>_>;| _ < _ < ]
e^s[|_<+|_>,_<+ _ > ] + & = & [ |_<_<]e^s^ , where e^s^&= & z_0(,/s ) e^s_0[|_<,_<]e^s_int_0 , > + & = & z_0(,/s ) e^s_0[| _ < , _ < ] + & & e^[s_int_0,>+(s_int^2_0,>-s_int_0,>^2)+ ] .
note that in deriving last equality we have used the cumulant expansion .
@xmath254 denotes the non - interacting partition function and @xmath255 represents average with respect to the fast modes of action @xmath240 .
the collected exponent of the exponential terms now results in the renormalized action @xmath256 . before examining the renormalization of coupling parameters @xmath93 and @xmath94
, we first notice that , to one - loop level , the quartic terms in @xmath257 do induce quadratic terms from the tadpole diagram shown in fig .
[ fig : tadpole ] , which are momentum independent .
this would indicate there is no non - interacting fixed point in our system .
however , explicit calculations show that all contributions from such type of diagram are zero and hence our starting fp survives without flowing ( @xmath258 , @xmath259 are not renormalized ) , in contrast to the case of the luttinger liquid .
next , we turn our attention to the diagrams that renormalize coupling parameters . as shown in figs .
[ fig : diagram](a)-(c ) , they are , respectively , total contributions to the renormalized @xmath93 , @xmath260 , and @xmath261 . obviously , they have standard structures such as ` bcs ' , ` zs ' ( zero sound ) , and ` zs@xmath262 ' used in shankar s seminal work.@xcite note that we have taken the convention that @xmath263 and vice versa .
the essential fermion loop integrals are listed below ( without vertex ) : & : & _ q g_0aa^(q)g_0|a|a^^(-q ) = dl(|t ) , + & : & _ q g_0a|a^(q)g_0|aa^^(-q ) = dl(|t ) , + & : & _ q g_0aa^(q)g_0aa^|(-q ) = dl , + & : & _ q g_0aa^(q)g_0|a|a^^(q ) = -dl , + & : & _ q g_0a|a^(q)g_0|aa^(q ) = dl(|t ) , + & : & _ q g_0aa^(q)g_0aa^^(q ) = -dl(|t ) , where @xmath264 and @xmath258 , @xmath265 , @xmath266 are defined in the main text [ just below eq .
( [ eq : rgeq ] ) ] .
the superscript , @xmath267 , of @xmath268 can be either @xmath86 or @xmath269 .
the presence of @xmath270 is due to the approximation we have used , @xmath271 in the limit @xmath272 . in order to evaluate these integrals
, we have set all external @xmath273 and made @xmath274 .
combining all the feynman diagrams in fig .
[ fig : diagram ] and keeping in mind the order of fermion operators , the straightforward algebra gives rise a set of rg equations , as we desire , eq .
( [ eq : rgeq ] ) . , ( b ) @xmath260 , and ( c ) @xmath261 .
different types of vertices are shown on the left .
the labels for external legs in each one - loop diagram are the same as the vertex on the left and hence are omitted . ] finally , we investigate the rg flow beginning with weak coupling regime for two typical cases : ( a ) @xmath275 ; ( b ) @xmath276 .
as one can see in fig .
[ fig : rgflow ] , in either case , they all flow to strong coupling regime and hence the non - interacting fixed point is actually unstable against short - range repulsions .
( solid , blue ) , @xmath277 ( dotted , black ) , and @xmath278 ( dashed , red ) with interaction profile ( a ) @xmath279 ; ( b ) @xmath280 . in both cases , @xmath244 . ]
the existence of the qsh state can be justified by a non - trivial @xmath281-valued invariant @xmath208 .
this @xmath281 index represents the topological nature of the qsh state and leads to topologically protected gapless edge states when the 2d system is made open with boundaries . here
, we present a method , first invented by fu and kane@xcite , to derive @xmath281 index explicitly .
this method essentially takes the advantage of _ inversion symmetry _ such that the index can be related to the parity eigenvalues @xmath282 of @xmath283 occupied energy bands at four trim : @xmath284 , and @xmath36 . for the purpose of instruction ,
we focus on 1/3 filling with @xmath285 and hence only the lower band is relevant .
as shown in sec iii , the presence of @xmath114 can dynamically generate the `` spin - orbit '' interaction , which is equivalent to add the following term into eq . ( [ eq : h0 ] ) .
_ so=_,_^h_so ( ) _ [ eq : hso ] with @xmath286 _ so ( ) , where @xmath287 with @xmath288 , and @xmath289 .
the @xmath290 sign refers to spin up ( down ) fermions .
the eigenvalues of @xmath291 are @xmath292 and @xmath33 and their corresponding eigenstates can be written in the form , u_n = g_n ( q_a , q_b , q_c)^ , [ eq : eigen ] where the expressions of the components @xmath293 ( @xmath294 ) and the normalization factor @xmath295 for each band @xmath68(@xmath296 ) are given in table [ eigenstate_so ] . at trim , we have eigenstates @xmath306 , @xmath307 , @xmath308 , and @xmath309 according to table [ eigenstate_so ] , respectively .
now , let us make a gauge transformation on eigenstates such that only one coordinate @xmath77 would be assigned to each unit cell ( instead of @xmath310 for each atom within the cell ) , and the bloch hamiltonian @xmath311 now obeys @xmath312 and , with the inversion operator @xmath313 , @xmath314 , resulting in @xmath315=0 $ ] .
thus , the parity eigenvalue now becomes a good quantum number at trim .
under such transformation the eigenstates are rewritten as @xmath316 , @xmath307 , @xmath317 , and @xmath318 . picking up any one of a - sites as our inversion center
, we can determine the parity eigenvalue for each trim by examining the wave functions ( eigenstates ) in real space .
since moving a unit cell by @xmath77 is simply multiplying our eigenstate at @xmath319 by a factor @xmath320 , we now see the parity eigenvalues for @xmath319=@xmath321 , are , respectively , @xmath322=@xmath323 . the @xmath281 index @xmath207 is then determined by ( -1)^=_i=1 ^ 4_i=-1 , where @xmath324 , and @xmath325 indicating either spin - up or spin - down band in our example here .
therefore , @xmath208 suggests that the ground state of our system at 1/3 filling has topological nature .
we summarize the parity eigenvalue of each band in table [ parityev ] .
finally , for @xmath48 , one can follow similar procedure and the conclusion of @xmath208 is unchanged .
99 see , for instance , k. von klitzing , rev .
phys . * 58 * , 519 ( 1986 ) ; h. l. stormer , d. c. tsui , and a. c. gossard , rev .
phys . * 71 * , s298 ( 1999 ) .
j. e. moore , nature * 464 * , 194 ( 2010 ) . for recent reviews ,
qi and s .- c .
zhang , arxiv:1008.2026 ; m. z. hasan and c. l. kane , rev .
phys . * 82 * , 3045 ( 2010 ) . c. l. kane and e. j. mele , phys .
lett . * 95 * , 146802 ( 2005 ) . c. l. kane and e. j. mele , phys .
lett . * 95 * , 226801 ( 2005 ) .
b. a. bernevig and s .- c .
zhang , phys .
. lett . * 96 * , 106802 ( 2006 ) .
b. a. bernevig , t. l. hughes , and s .- c .
zhang , science * 314 * , 1757 ( 2006 ) .
liang fu , c. l. kane , and e. j. mele , phys .
lett . * 98 * , 106803 ( 2007 ) .
liang fu and c. l. kane , phys .
b * 76 * , 045302 ( 2007 ) . c. wu , b. a. bernevig , and s .- c . zhang , phys .
* 96 * , 106401 ( 2006 ) .
s. raghu , x .-
qi , c. honerkamp , and s .- c .
zhang , phys .
lett . * 100 * , 156401 ( 2008 ) .
m. koenig , s. wiedmann , c. bruene , a. roth , h. buhmann , l. w. molenkamp , x .-
qi , and s .- c .
zhang , science * 318 * , 766 ( 2007 ) .
d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava and m. z. hasan , nature ( london ) * 452 * , 970 ( 2008 ) .
d. hsieh , y. xia , d. qian,5 , l. wray , j. h. dil , f. meier , j. osterwalder , l. patthey , j. g. checkelsky , n. p. ong , a. v. fedorov , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nature * 460 * , 1101 ( 2009 ) .
y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nat .
* 5 * , 398 ( 2009 ) .
y. l. chen , j. g. analytis , j .- h .
chu , z. k. liu , s .- k .
qi , h. j. zhang , d. h. lu , x. dai , z. fang , s .- c .
zhang , i. r. fisher , z. hussain , and z .- x .
shen , science * 325 * , 178 ( 2009 ) . c. weeks and m. franz , phys .
b * 81 * , 085105 ( 2010 ) .
cenke xu , arxiv:1010.0455 .
kai sun , hong yao , e. fradkin and s. a. kivelson , phys .
lett . * 103 * , 046811 ( 2009 ) .
qin liu , hong yao , and tianxing ma , phys .
b * 82 * , 045102 ( 2010 ) .
yi zhang , ying ran , and ashvin vishwanath , phys .
b * 79 * , 245331 ( 2009 ) .
o. vafek and kun yang , phys .
b * 81 * , 041401(r ) ( 2010 ) . c. weeks and m. franz , phys . rev .
b * 82 * , 085310 ( 2010 ) .
d. green , l. santos , and c. chamon , phys .
b * 82 * , 075104 ( 2010 ) . v. oganesyan , s. a. kivelson , and e. fradkin , phys . rev .
b * 64 * , 195109 ( 2001 ) .
s. a. kivelson , i. p. bindloss , e. fradkin , v. oganesyan , j. m. tranquada , a. kapitulnik , and c. howald , rev .
phys . * 75 * , 1201 ( 2003 ) . c. wu , k. sun , e. fradkin , and s .- c .
zhang , phys .
b * 75 * , 115103 ( 2007 ) . y. d. chong , x .-
wen , and m. soljacic , phys .
b * 77 * , 235125 ( 2008 ) .
kai sun and e. fradkin , phys .
b * 78 * , 245122 ( 2008 ) .
s. uebelacker and c. honerkamp , arxiv:1105.1989 .
f. d. m. haldane , phys .
93 * , 206602 ( 2004 ) .
d. l. bergman , congjun wu , and l. balents , phys .
b * 78 * , 125104 ( 2008 ) .
n. goldman , d. f. urban , and d. bercioux , arxiv:1101.4500 .
h. nielssen and n. ninomiya , phys . lett .
* 130b * , 389 ( 1983 ) . f. d. m. haldane , phys .
61 * , 2015 ( 1988 ) . c. m. varma , phys .
b * 73 * , 155113 ( 2006 ) . if one replaces @xmath332 by @xmath333 with @xmath334 denoting the average charge density and rewrites the density - density repulsions in terms of it in eq .
( [ eq : int ] ) , the mapping between 1/3 and 2/3 fillings by particle - hole language translation will become exact .
e. h. lieb , phys .
rev . lett . * 62 * , 1201 ( 1989 ) .
v. j. emery , phys .
. lett . * 58 * , 2794 ( 1987 ) .
e. dagotto , rev .
mod . phys . * 66 * , 763 ( 1994 ) and references therein .
r. thomale and m. greiter , phys .
b * 77 * , 094511 ( 2008 )
. m. s. hybertsen , m. schluter , and n. e. christensen , phys .
b * 39 * , 9028 ( 1989 ) .
kai sun , w. vincent liu , and s. das sarma , arxiv:1011.4301 .
e. tang , j .- w .
mei , and x .-
wen , phys .
* 106 * , 236802 ( 2011 ) .
t. neupert , l. santos , c. chamon , and c. mudry phys . rev . lett .
* 106 * , 236804 ( 2011 ) .
kai sun , z .- c .
gu , h. katsura , and s. das sarma , phys .
* 106 * , 236803 ( 2011 ) .
d. n. sheng , z .- c .
gu , kai sun , and l. sheng , nature communications 2 , 389 ( 2011 ) .
n. regnault and b. andrei bernevig , arxiv:1105.4867 ; yang - le wu , b. andrei bernevig , and n. regnault , arxiv:1111.1172 .
lu and y. ran , arxiv:1109.0226 .
r. shankar , rev .
phys . * 66 * , 129 ( 1994 ) . | we show that topological states are often developed in two dimensional semimetals with quadratic band crossing points ( bcps ) by electron - electron interactions . to illustrate this
, we construct a concrete model with the bcp on an extended lieb lattice and investigate the interaction - driven topological instabilities .
we find that the bcp is marginally unstable against infinitesimal repulsions .
depending on the interaction strengths , topological quantum anomalous / spin hall , charge nematic , and nematic - spin - nematic phases , develop separately .
possible physical realizations of quadratic bcps are provided . |
the lorenz dynamical system , originally introduced by lorenz @xcite in order to describe in a very simplified way the rayleigh - bnard problem @xcite , immediately became important in itself as one of the most studied low - dimensional chaotic systems . still today the lorenz model represents a paradigmatic example for both theoretical and numerical investigations in checking some results in chaos theory @xcite , in the study of geometrical properties of dynamical systems @xcite , in nonlinear analysis of time series @xcite , in the stabilization and synchronization of coupled systems @xcite , and so on .
+ nevertheless , despite the great attention attracted over past decades , some fundamental and rigorous results have been obtained quite recently , as for instance the proof of the existence of the lorenz attractor @xcite , usually using somewhat sophisticated mathematical tools . on the contrary ,
our aim here is to provide a description of lorenz system dynamical features , which requires quite simple analytical tools and , at the same time , allows a very intuitive inspection in lorenz - like chaos .
preliminary results have been reported in a short communication @xcite . here
we shall give more details and new results .
our interpretation will base itself upon the fact that in the `` steady state '' , i.e. , far from the initial transient and when the memory of the initial conditions has been lost , the lorenz system is equivalent to a suitably constructed second order integral - differential equation .
this equation can be regarded , for instance , as a customary second order one - dimensional classical mechanics equation with a peculiar forcing term .
the corresponding dynamics can be interpreted as the one - dimensional motion of a particle in a conservative quartic two - well potential , subjected to a viscous damping and to an additional force resulting from the past history of the motion .
the latter force turns out to be essential for chaos to emerge as it acts as an `` endogenous '' forcing able to permanently sustain the motion even in presence of friction .
the previous interpretation of lorenz dynamics actually leads to a generalization of the lorenz model to a wider class of systems showing similar dynamical properties .
we shall introduce a particular system belonging to such class which , because of its simplicity ( piecewise linearity ) , will allow us to study the dynamics of the original model using analytical tools .
indeed , in the steady chaotic regime ( i.e. when the system permanently lies on its attractor set ) the evolution of a point in the lorenz phase - space can be roughly described as a somewhat regular amplified oscillation around one of its fixed points , followed by a sudden entry in the control basin of a twin fixed point symmetrical with respect to the origin , by a new amplified oscillation around the latter point , by a sudden return in the control domain of the former , and so on .
+ the most evident aspect of the chaotic regime is just the unpredictability of the instant at which the center of the aforementioned amplified oscillations changes . the choice of a piecewise linearized version of the original model will allow us to highlight this point , while keeping unchanged the peculiar topological properties of the lorenz dynamics .
the exact equation ruling the instant of change of the oscillation center will be derived and a discontinuous dependence of this instant on the initial conditions will be highlighted .
it will also be possible to write the analytical equations that defines the first - return two - dimensional poincar map for the piecewise linearized system , which in turn synthesizes the main chaotic features of the model dynamics .
moreover , we shall show that , under suitable and reasonable conditions , the evolution of the system completely reduces to a one - dimensional chaotic map .
+ to summarize , starting from our interpretation , we shall be able to propose a piecewise linearized version of the lorenz model , which on one hand has the same dynamical properties of the original system and , on the other hand , will provide analytical tools to explicate the emergence of chaos in lorenz - like systems .
the original lorenz system @xcite consists of the three first order ordinary differential equations @xmath0 where the dots indicate time derivatives and @xmath1 , @xmath2 and @xmath3 are positive parameters originally related to the fluid properties and to the boundary conditions in the rayleigh - bnard problem .
( lorenz @xcite used @xmath4 , @xmath5 and @xmath6 ) .
+ the fixed points of system ( [ lorenz ] ) and their corresponding stabilities depend on @xmath3 . for @xmath7
there is only one ( stable ) fixed point in @xmath8 .
for @xmath9 the origin looses its stability and a pair of new fixed points appear : @xmath10^{1/2}$ } , \mbox{$\pm [ b(r-1)]^{1/2}$ } , \mbox{$r-1)$}$ ] , which are stable in the range @xmath11 with @xmath12 .
( note that the critical value @xmath13 exists if and only if @xmath14 ) . for @xmath15
all fixed points are unstable , and the lorenz system can exhibit either periodic or chaotic behavior ( see , e.g. , ref .
@xcite for a comprehensive exposition on the matter ) .
since we are interested in the case @xmath9 , we can suitably define the scaled coordinates @xmath16^{1/2}$ ] , @xmath17^{1/2}$ ] , @xmath18 , so that the system becomes @xmath19 with fixed points @xmath8 and @xmath20 .
we now reduce the lorenz system to a unique differential equation for @xmath21 , whose solution makes it possible the direct calculation of @xmath22 and @xmath23 . by inserting into the third equation of the expression for @xmath24 in terms of @xmath25 obtained from the first , one easily gets @xmath26,\ ] ] whose general solution is given by @xmath27 we have already stated that we are interested in the evolution of the lorenz system in the chaotic steady state , i.e. far from the initial transient .
thus , we let @xmath28 move back to @xmath29 and obtain the steady - state expression for @xmath23 , which can be written in the form @xmath30_b(t),\ ] ] where we use the square brackets notation @xmath31_k(t)$ ] to indicate the exponential average of any suitable time function @xmath32 on its past history : @xmath33_k(t)\equiv k\int_0^\infty ds\;e^{-ks}f(t - s)\;.\ ] ] from now on we shall refer to @xmath31_k$ ] as the _ @xmath34-exponentially vanishing memory _ of the function @xmath32 or in short its _ memory_. note that , for the memory @xmath31_k$ ] to exist , it suffices that @xmath35 , with @xmath36 , as @xmath37 ( the time functions we shall be dealing with are even bounded in this limit ) .
+ to summarize , we insert in the second equation of @xmath22 obtained from the first equation and @xmath23 given by ( [ steady ] ) and get for the variable @xmath21 alone in chaotic steady state the following second order differential equation with memory ( in fact , an integral - differential equation ) @xmath38_b x\right)\;.\ ] ] one can interpret the previous equation as the dynamical equation of a ( unit mass ) particle , viscously moving in a compound potential energy field consisting of a weighted average ( in the chaotic regime @xmath39 ) of a quartic potential energy field independent of time ( fig . [ quartico ] ) @xmath40 and of a quadratic potential energy field @xmath41_b \dfrac{(x^2 -1)}{2},\ ] ] whose curvature is given by a suitable memory function of the past motion . without the memory term
the particle would stop in one of the minima of the bistable potential @xmath42 due to the damping term @xmath43 .
indeed , the @xmath44 contribution yields , through an exponential average on the past evolution , an `` endogenous '' forcing term which can permanently sustain the motion .
the particle oscillates with growing amplitude around the minimum of one of the potential well until its energy is sufficient to allow the crossing of the barrier in @xmath45 . as already said , the chaotic behavior of the system emerges just in the unpredictability of the instant at which the particle moves from one well to the other .
trajectories relative to very slightly different initial conditions can produce strongly different sequences in the number of oscillations in each well . in the next section
we shall explain this fact analytically . in order to simplify and to make the analysis standard ,
let us rescale the time @xmath46 in eq . as @xmath47^{1/2 }
\,t$ ] , to obtain the equation ( hereafter called _ lorenz equation _ ) @xmath48 _ \beta\ : x , \label{eq.lorenz}\ ] ] where @xmath49 note that , given @xmath2 and @xmath3 , @xmath1 ( the viscosity parameter in the original problem ) affects both the friction term and the forcing term in the lorenz equation .
this fact shows how much these opposite " contributions are in fact strictly related if the equation is to be viewed as representative of the original lorenz system .
even if one now considers the lorenz equation as the main subject of the study , one must note that not all the ( positive ) values of @xmath50 , @xmath51 , @xmath52 are consistent with their definition in terms of the original parameters @xmath2 , @xmath1 , @xmath3 .
in particular , in order to observe chaotic behavior , the following inequality must hold @xmath53 ( further details on the relation between the two sets of parameters are given in appendix [ app : b ] ) .
equation allows us to highlight the role of the memory forcing term in the lorenz system dynamics . for this purpose
it is interesting to compare the lorenz equation with other examples of chaotic nonlinear ( nonautonomous ) systems as , for instance , the inverse duffing equation @xmath54 , which describes a sinusoidally forced quartic oscillator @xcite , or more appropriately the parametrically forced equation @xmath55 note that in the latter cases the motion is sustained by externally assigned forcing terms , while in the case of the lorenz equation the motion is self - sustained by the endogenous term @xmath56_\beta\ , x$ ] . in figs .
[ xv - lorenz-1 ] and [ duffing ] we give a numerically obtained comparison between the phase portraits of the lorenz equation and of equation ( [ unnamed ] ) with parameters , and suitably chosen in order to get similar ranges of motion .
note that the endogenous lorenz forcing term , mimed by @xmath57 , is in fact neither monochromatic nor with vanishing average value .
the lorenz equation can be usefully recast in the form @xmath58_\beta\,\right)\ , \phi'(x)\:=\,0\ ] ] where the prime indicates the derivative with respect to @xmath25 and in our case @xmath59 and @xmath60 .
such an equation can be interpreted as the description of the motion of a unit mass particle subjected to a viscous force @xmath61 and interacting with a potential field @xmath62 through a `` dynamically varying charge '' @xmath63_\beta$ ] .
this charge both depends on the instantaneous particle position @xmath64 ( by means of the term @xmath65 ) and on the past evolution ( by means of the _ memory charge _
@xmath66_\beta$ ] ) .
the coupling of @xmath67_\beta$ ] with the fixed potential field @xmath62 acts as an endogenous forcing term which can sustain the motion even in the presence of friction , and the chaotic behavior can actually arise from the synergy between this term and the viscosity .
+ put in the form , the lorenz equation is arranged to be generalized to a generic charge @xmath65 interacting with a generic potential field @xmath68 .
correspondingly , it is possible to obtain a _ generalized lorenz system _ whose `` @xmath25 - projection '' ( far from the initial transient ) yields eq . .
indeed , by inverting the calculation followed to derive eq . from system
, one easily gets the generalized lorenz dynamical system @xmath69 . \end{array } \right.\ ] ] therefore , the specific lorenz model can be viewed as singled out from a quite general class of dynamical systems which can exhibit chaotic behavior , their common essential property being an exponentially vanishing memory effect together with a viscous damping .
equations and are related to eq . by assuming as potential energy field the quantity @xmath70 with @xmath71_\beta \,\phi\ ; . \end{array}\ ] ] obviously , any choice of @xmath72 and @xmath62 should maintain the main properties of the lorenz model , i.e. , correspond to a two - well piecewise differentiable potential energy @xmath73 , such that @xmath74 as @xmath75 .
we shall now focus our attention on a particular choice for @xmath72 and @xmath62 , that will maintain all the qualitative properties of the lorenz system and , at the same time , will allow us to deal with chaos analytically .
as already noticed , the chaotic behavior of the lorenz system essentially depends on the unpredictability of the instants when @xmath25 change its sign : as long as it keeps constant sign the system evolution is certainly nonlinear , and nevertheless not chaotic " at all .
this fact suggests a slight modification of the original form of the lorenz system , in order to single out analytically the origin of chaos without to be faced with the difficulties arising from nonlinear features .
we thus set in eqs .
and @xmath76 and @xmath77 obtaining ( for @xmath78 ) the piecewise linear lorenz - like equation @xmath79_{\beta } \,\right\ } \mbox{sgn}(x)=0,\\ \intertext{where $ \text{sgn}(x)\equiv |x|/x$. the corresponding piecewise linearized dynamical system , with the original choice of parameters , is then given by } \label{sist.lin . }
\left\ { \begin{array}{l } \dot{x}=\sigma ( y - x ) \\ \\ \dot{y}=-y+x + ( r-1)(1-z ) \mbox { sgn}(x)\\ \\ \dot{z}=-bz+b\,\mbox{sgn}(x)\dfrac{x+y}{2}. \end{array } \right .
\end{gathered}\ ] ] our assumptions on @xmath62 and @xmath72 correspond in eq . to @xmath80 and @xmath81_b \left(\ , |x|-1\,\right)$ ] .
thus , we are faced with a simplified model , obtained by replacing the constant - in - time quartic potential with a piecewise quadratic one resulting from the superposition of two parabolas with vertex in @xmath82 and truncated at @xmath45 ( fig . [ parabole ] ) . the two - well character of @xmath42 is obviously maintained as well as the piecewise differentiability .
the replacement of the original potential actually corresponds to a linearization of the system around both unstable fixed points @xmath83 , with the metching performed in @xmath45 .
it appears that the chaotic behavior of the original model does not depend on the differentiability in @xmath45 .
one can guess that also other classes of dynamical systems can be transformed in a piecewise linearized version by means of the same operations .
it is easy to check that the fixed points of are @xmath84 and that the equation ruling their local stability is the same as for the original lorenz system with parameters @xmath2 , @xmath1 , @xmath85 .
in particular , if @xmath14 , the critical value of @xmath3 for our piecewise linear system is given by @xmath86 . in figs .
[ xv - lorenz-2 ] and [ xv - lorenz - lin ] two chaotic phase portrait for systems ( [ lorenz scalato ] ) and ( [ sist.lin . ] ) are shown , corresponding to the same choice of @xmath2 and @xmath1 , and different choices of @xmath3 in order to preserve the relationships between @xmath3 and the proper corresponding critical value .
+ when dealing with system , the main simplification is that it is separately linear for @xmath87 and @xmath88 and can thus be analytically solved in each region . indeed , by applying the operator @xmath89 to each side of eq .
one obtains the equation @xmath90 which can be explicitly solved separately on each side of @xmath45 .
the nonlinearity of the model is simply reduced to a change of sign of the forcing term @xmath91 when @xmath25 crosses the plane @xmath45 , henceforth denoted with @xmath92 @xcite .
as we shall see , the crossing times are somewhat unpredictable , as they result from the discontinuous solutions of an ( incidentally transcendental ) equation .
our piecewise linearized system will thus turn out to be an important tool to analytically investigate the emergence of chaos in lorenz - like systems .
the main advantage of the piecewise linearization is that one has to deal only with the simplest nonlinearity , i.e. an isolated ( not eliminable ) discontinuity ( see , e.g. , refs .
@xcite for other examples of piecewise linear chaotic dynamical systems ) .
let us now consider in detail the second order integral - differential equation , which describes the evolution in time of @xmath25 for the piecewise linearized system in the chaotic steady state .
it is equivalent to a third order nonlinear differential equation whose phase space is described by the coordinates @xmath25 , @xmath93 , @xmath94 ( with a small abuse of notation from this time on we shall indicate with the dot the differentiation with respect to @xmath95 ) .
solutions of eq . can be easily calculated with tools of the customary analysis on each side of @xmath45 . to obtain a global solution ,
such partial solutions should be matched at @xmath45 under the reasonable assumptions that the position @xmath25 , the velocity @xmath93 , and the memory @xmath96_\beta$ ] are continuous @xcite .
notice that , in contrast , when crossing the plane @xmath92 , the acceleration @xmath97 turns out to be undefined .
however , if @xmath98 , referring to the left and right time limits @xmath99 and @xmath100 ) , it appears from eq .
that they are related by the equation @xmath101 in the sequel we shall refer to @xmath99 and @xmath102 as the acceleration `` immediately '' before and `` immediately '' after the crossing , respectively . since eq .
is invariant under the transformation @xmath103 , @xmath104 , we can focus our attention only on one of the two regions , e.g. @xmath88 , and describe the motion in this half - space ( the evolution in its twin half - space being recovered through the change @xmath103 ) . clearly , this is equivalent to put a `` rigid wall '' in @xmath45 and look at the crossing of @xmath92 as an elastic collision .
+ in summary , according to the previous scheme , the system evolution in time is completely described by the following steps : 1 .
motion for @xmath88 ; 2 .
collision against @xmath92 and discontinuity of @xmath97 ; 3 .
inversion @xmath105 and matching with a new solution defined in the region @xmath88 again .
if we define @xmath106 , eq . assume a simple form @xmath107 that is a linear third order differential equation ( homogeneous and with constant coefficients ) .
as a consequences of the routh - hurwitz theorem @xcite , a critical value @xmath108 for the parameter @xmath50 exists , i.e. @xmath109 for @xmath110 the fixed points are unstable saddle focus with a real negative eigenvalue @xmath111 and a pair of complex conjugates eigenvalues @xmath112 , @xmath113 .
for the sake of simplicity we shall indicate with @xmath114 the time interval between two consecutive collisions in the steady - state of the system .
without loss of generality we assume @xmath45 and @xmath115 .
then , it can be easily shown that in the time interval @xmath114 the motion in the phase space @xmath116 , @xmath117 , @xmath118 ( @xmath119 ) is completely described by the equation @xmath120 where we have defined @xmath121 \ddot{\xi}(0^+ ) \end{pmatrix } , \\
\intertext{and } { \cal m}(\tau)\:=\ : \begin{pmatrix } \mbox{re } \left(e^{\lambda_1 \tau}\right ) & \mbox{im } \left(e^{\lambda_1 \tau}\right ) & e^{-\lambda_0 \tau}\\ \\ \mbox{re}\left(\lambda_1 \ , e^{\lambda_1 \tau}\right ) & \mbox{im } \left(\lambda_1\,e^{\lambda_1 \tau}\right ) & -\lambda_0 \,e^{-\lambda_0 \tau}\\ \\ \mbox{re } \left(\lambda_1 ^ 2 \ , e^{\lambda_1 \tau}\right ) & \mbox{im } \left(\lambda_1 ^ 2\,e^{\lambda_1 \tau}\right ) & \lambda_0 ^ 2 \,e^{-\lambda_0 \tau } \end{pmatrix}\ , .
\end{gathered}\ ] ] ( note that , for @xmath110 , @xmath122 is always invertible since @xmath123 $ ] and @xmath124 ) .
the eigenvalues of the matrix @xmath125 , which connects the vector @xmath126 to its initial value @xmath127 , are @xmath128 , @xmath129 , @xmath130 , with corresponding ( constant - in - time ) eigenvectors @xmath131 \lambda_1 ^2 \end{pmatrix } , & \boldsymbol{v}_2\equiv \begin{pmatrix } 1\\ \bar{\lambda}_1 \\[0.05 cm ] \bar{\lambda}_1 ^2 \end{pmatrix } , & \boldsymbol{v}_3\equiv \begin{pmatrix } \
1\\ -\lambda_0 \\[0.05cm]\quad
\lambda_0 ^ 2 \end{pmatrix } \end{array } .\ ] ] with respect to the base @xmath132 one has @xmath133 for the saddle point @xmath134 there exist a stable one - dimensional manifold @xmath135 corresponding to @xmath136 @xmath137 the evolution of the linearized system in the interval @xmath114 is the combination of an exponential decay along @xmath135 and of an amplified rotation on @xmath138 . to determine the relative quickness of each component of the motion with respect to the other , consider that from eq .
it is easily checked that @xmath139 , and so @xmath140 @xmath141 .
therefore , the exponential decay along the stable manifold @xmath135 is always more rapid than the exponential oscillating growth on the unstable manifold @xmath138 , the parameter controlling this difference being @xmath142 .
+ although the fixed points are saddle focus , it appears that the piecewise linearized system does not exhibit shilnikov chaos @xcite owing to the absence of an homoclinic orbit . to complete the description of the piecewise linearized system dynamics
, we consider the instant @xmath143 at which the first collision occurs . at this time , as already remarked , the trajectory coming from the half - space @xmath88 must be matched with the solution defined again in the same region , but with corresponding new `` initial conditions '' @xcite @xmath144 with @xmath145 the matrix @xmath146 accounts for the acceleration discontinuity in @xmath45 , while @xmath147 yields the sign inversion after the impact . from eqs . and
it follows that the velocity and the acceleration immediately after the collision are related to the initial conditions by the operator @xmath148 , according to the formula @xmath149 notice the highlighted dependence of @xmath143 on @xmath150 , which reveals the nonlinear character of this important relationship .
starting from the above results , the very origin of chaos in the piecewise linearized lorenz system will be identified and discussed in the next section .
we shall also show that the basic mechanisms for chaos to emerge apply also to the original lorenz system .
as already observed , the instant at which the plane @xmath92 is crossed , strongly depends on the initial conditions .
this fact is strictly related to the chaotic behavior of lorenz - like systems .
let us now study in some detail this topic for the piecewise linearized model .
the instant @xmath143 at which the first collision occurs is defined by the condition @xmath151 . from the first line of eq .
@xmath143 is thus the smallest positive solution of the transcendental equation @xmath152 where @xmath153 , @xmath154 , @xmath155 are linearly related to initial conditions .
the _ residence time _
@xmath143 is therefore the first intersection of the graphs of @xmath156 and @xmath157 . since @xmath158 is a decreasing exponential function and @xmath159 an oscillating function with growing amplitude
, one can easily understand why even a little modification of initial conditions can produce a discontinuous variation of @xmath143 ( fig .
[ intersez ] ) . in conclusion , the unpredictability of the residence time is closely connected to the discontinuous character of the solutions of eq . .
the chaotic behavior for the the piecewise linearized system clearly stems from such unpredictability .
our claim is that an entirely analogous situation exists also for the original lorenz model .
we shall return however in more detail on this important analogy in the sequel .
we now explicitly investigate the dependence of @xmath143 on the initial velocity and acceleration .
as previously remarked , the function @xmath162 is implicitly defined as the smallest positive solution of eq . . since @xmath153 , @xmath154 , @xmath155 are linearly related to the initial conditions , eq .
( [ trasc . ] ) can be rewritten in terms of @xmath160 and @xmath163 . in this form
it describes a family @xmath164 of straight lines in the plane @xmath165 parametrized by @xmath143 @xmath166 [ see appendix [ app : fascio ] for the explicit expression of the coefficients @xmath167 , @xmath168 , @xmath169 .
+ let us denote with @xmath170 the ordered sequence of the zeros of @xmath168 .
note that one always has @xmath171 .
the slope of the straight lines of @xmath172 , @xmath173 , and their ordinate for @xmath174 , i.e. @xmath175 , have their singularities in @xmath176 .
both these functions are only asintotically periodic because of the presence of terms proportional to @xmath177 , which become negligible only for large @xmath143 ( figs .
[ coeff ] and [ intercetta ] ) .
+ in each interval @xmath178 , @xmath179 , the function @xmath173 is everywhere growing ( see appendix [ app : fascio ] ) and varies from @xmath29 and @xmath180 .
therefore , as @xmath143 increases , the straight lines of @xmath172 rotate anticlockwise and at the same time translate starting from the @xmath181-axis ( obtained for @xmath182 ) ( fig .
[ ragnatela ] ) .
+ the envelope of the family @xmath172 is a curve @xmath183 which looks similar to an elliptic spiral ( fig .
[ inviluppo ] ) .
( a parametrical representation of @xmath184 is given in appendix [ app : fascio ] ) . from eq .
the contour lines of the function @xmath143 are parts of straight lines in the plane @xmath185 .
indeed , according to the definition , each line of @xmath172 is in fact a contour line for @xmath143 only where its points do not belong to another line corresponding to a smaller value of @xmath143 and , furthermore , only where their abscissas correspond to positive values of @xmath160 . rather than a family of straight lines ,
the function contour lines are raies or segments according to the constant value of @xmath143 . + for @xmath143 ranging from @xmath186 to @xmath187 these lines perform a complete anticlockwise `` rotation , '' starting from the .
thus , they cover the entire half - plane @xmath188 except the region inside their envelope ; from this first rotation one obtains a set of raies ( fig .
[ semirette ] ) .
+ subsequently , for @xmath143 ranging from @xmath187 to @xmath189 the contour lines become segments of variable orientation and contained in the region delimited by the curve already generated from the first rotation and the new envelope of the second set of straight lines .
this behavior repeats itself @xmath190 $ ] , @xmath179 . because of the aperiodic character of the functions involved , the curve @xmath184 outlines in the positive half - plane a structure consisting of `` pseudo - elliptic '' annula . moving anticlockwise along each of these annula
, @xmath143 grows continuously . on the contrary , passing through the border that separates two different bands , one meets discontinuities in the dependence of @xmath143 on initial conditions .
+ the curve @xmath184 `` winds '' round the point @xmath191 ( see fig . [ inviluppo ] ) . for @xmath192
one has @xmath193 , since for these initial conditions one obtains @xmath194 and the system exactly lies on the stable manifold @xmath135 .
its motion is in this case an exponential decay towards the fixed point @xmath134 .
the previous observations allow to easily guess the structure of the graph of @xmath195 shown in fig .
[ piramide ] .
it should be noted that @xmath143 shows instability with respect to the initial conditions only in a limited subset of the half - plane @xmath188 . as we shall see , it is natural to expect that in the chaotic regime the system is quickly attracted inside this region . at this point of our study
we have analyzed in some detail the unpredictability of the time at which the system crosses the plane @xmath92 . to completely motivate the chaotic dynamics of the piecewise linearized system and thus of the original model , we must , however , add some further results on the attracting set of the system .
all results obtained in the previous section can be easily extended to the collision against @xmath92 . specifically , denoting with @xmath196 the array assigning position , velocity , and acceleration immediately after the collision @xmath197 we have [ similarly to eq . ]
@xmath198 where we have denoted with @xmath199 the residence time , i.e. , the time interval between the collision and the one .
+ the first line of eq .
gives @xmath199 in terms of @xmath200 and @xmath201 : the dependence of this time on the initial conditions has already been discussed in section [ sec : graph ] .
knowing @xmath199 , the other two lines allow to relate the velocity @xmath202 and the acceleration @xmath203 to @xmath200 and @xmath204 .
. defines a two - dimensional first - return poincar map between @xmath205 and @xmath206 , obtained from the section @xmath207 of the phase space @xmath208 @xmath117 , @xmath118 .
we recall that , in spite of its appearance , the map is in fact non - linear , since the matrix @xmath209 depends on @xmath199 , which is a transcendental discontinuous function of the @xmath210-th initial conditions .
+ indeed , under suitable conditions , the two - dimensional map for the linearized system practically reduces to a one - dimensional map .
it has been already shown that for @xmath211 the motion in the phase space has the form @xmath212 where @xmath213 and @xmath214 satisfy the equation @xmath139 , which in turn implies @xmath140 . as a consequence
the evolution in the phase space @xmath116 , @xmath117 , @xmath118 consists both of a `` rapid '' decay towards @xmath134 along the stable manifold @xmath135 and of a `` slow '' amplified oscillation on @xmath138 .
we thus expect the phase trajectories to be strongly attracted on the unstable manifold and , once on @xmath138 , to slowly spiral outwards .
provided trajectories start close enough to @xmath138 , then they meet the @xmath92-plane very close to its intersection with @xmath138 itself , i.e. , along the straight line @xmath215 let us assume that the attraction towards the unstable manifold is very strong and thus it takes place almost instantaneously ( the goodness of this assumption is controlled by @xmath216 ) . under this hypothesis
all trajectories approximately hit @xmath92 along the straight line @xmath217 .
thus , from eq . , it follows that , immediately after each collision , the system necessary lies very close to the straight line @xmath218 therefore , the following relation between velocity and acceleration is expected to approximately hold @xmath219 from eqs . and
an analogous linear dependence between the velocity and the memory follows @xmath220 where @xmath221 denotes the of @xmath116 evaluated at the collision . as a consequence , in the limit
we have considered , the attracting set for the map is @xmath222 and , therefore , it reduces to an one - dimensional map . since in the chaotic steady - state
the velocity and the acceleration that define the initial conditions for the trajectory after each impact are not independent , the time @xmath199 can be expressed as a function of @xmath200 alone : @xmath223 ( see fig .
[ t - v ] ) .
the behavior of this map is easily understood if we refer to the graph of @xmath143 as a functions of @xmath160 and @xmath161 ( see fig .
[ piramide ] ) .
indeed , in the chaotic regime @xmath224 is superimposed to the region of the plane @xmath225 contained by @xmath184 , where @xmath143 shows unstable behavior with respect to a change in the initial conditions ( fig .
[ sovrapp ] ) . for those values of @xmath200 ,
for which @xmath224 passes through the same `` pseudo - elliptic '' corona , @xmath199 slowly changes with varying crossing velocity . on the contrary ,
when @xmath224 intersects the boundary between two different coronas , @xmath199 shows a discontinuity in its dependence on @xmath200 ( fig .
[ t - v ] ) .
+ the linear dependence between velocity and acceleration immediately after each collision implies that the poincar map becomes one - dimensional , e.g. a map between @xmath202 and @xmath200 .
figure [ v - v ] shows this fact for @xmath226 , @xmath227 , @xmath228 . by simple inspection of the derivatives corresponding to the fixed points of the map one can easily check that they are all unstable , so that the map produces a chaotic behavior .
note that the discontinuities simply correspond to the analogous ones in the function @xmath223 , evaluated along the straight line given by and drawn in fig.[sovrapp ] . for the sake of completeness , we turn now to the description of the piecewise linearized system in the original phase space @xmath25 , @xmath93 , @xmath97
. there the system has two fixed points @xmath229
. for each one there exists a stable manifold @xmath230 during the evolution the trajectories are rapidly attracted on the unstable manifold relative to the half - plane where they belong ( fig . [ spaziofasi ] ) .
thus , immediately after the crossing of the @xmath92-plane the phase point lies on one of the straight lines @xmath231 as for the crossing of the @xmath92-plane , the system can therefore be described by a one - dimensional map . to get the maps @xmath232 and @xmath233 from the analogous ones in the rigid wall scheme
, one has simply to consider that @xmath202 coincides with @xmath234 .
the reader should note that all the maps we have drawn in the previous figures have been obtained by numerically solving the analytical equation , which exactly defines @xmath199 , @xmath235 , and @xmath236 , and they have not been computed by numerical integration of the differential system ( as usual for non solvable dynamical systems like the lorenz original one ) .
thus , the use of numerical tools has been required only because of the transcendental character of the concerned equation .
we explicitly remark that in general the attraction towards @xmath138 is not infinite , but nevertheless more rapid than the amplified rotation on the unstable manifold ( remember that @xmath237 @xmath238 ) .
therefore , we can expect that , if @xmath216 is finite , the attractor set is not exactly the straight line @xmath222 , but a narrow `` strip '' that contains @xmath222 .
this strip intersects the region of the graph of @xmath199 where the residence time is strongly dependent on initial conditions and the mechanism for chaos to arise is absolutely the same .
our numerical simulations for finite @xmath216 are in excellent agreement with these predictions and then we do not show their corresponding graphs here , since they are practically indistinguishable from the graphs we have already discussed .
we conclude our analysis by showing some simulated numerical results .
bfore doing that , some remarks are worth discussing .
we previously considered @xmath239 as independent coordinates in the phase space because of their physical meaning : it is more intuitive to speak about acceleration rather than of memory of the system . unfortunately , if we want to `` assign initial conditions '' to the original lorenz system after the crossing of the plane @xmath92 , we can not consider the acceleration and the velocity since from eq .
they are not independent variables in @xmath45 . indeed the relation @xmath240 always holds , while the memory is not known _ a priori_. as a consequence , we have to consider as independent initial conditions position , velocity , and memory .
however , it is worth remarking that the comparison between the lorenz system and the piecewise linearized one is completely meaningful . this because for the latter model @xmath241 is easily known in terms of @xmath200 and @xmath201 from eq . : for the linearized system the choice of @xmath201 or of @xmath241 is absolutely equivalent . since we are interested in the behavior of the lorenz system in the chaotic steady state , all the following results are thus obtained via the numerical integration of the steady - state differential system @xmath242 where now @xmath243_\beta$ ] . this dynamical system , as already proved , is equivalent to the system far from the initial transient
we can now turn to the comparison with the piecewise linearized system . for the original lorenz model we have computed the maps that relate @xmath199 and @xmath234 to @xmath244 , respectively , by numerical integration of the system ( the notations are the same as in the previous section ) .
figures [ t - vl ] and [ v - vl ] show the results we obtained : the qualitative behavior of these maps is very similar to that of the analogous maps for the piecewise linearized system .
as we will see , the mechanism for chaos to arise is indeed the same .
+ observe that the maps in figs .
[ t - vl ] and [ v - vl ] are not perfectly onto : obviously this is due to the fact that , in general , the lorenz system does not reduce exactly to an one - dimensional map when the plane @xmath45 is crossed . from direct computation
it appears that the residence time around a fixed point depends on the initial conditions in a way similar to that of the linearized system ( fig .
[ torrel ] ) .
there is a region in which @xmath143 varies continuously with varying initial conditions and a region ( figs .
[ torrel ] and [ torrel1 ] ) of strong instability . in the latter region
@xmath143 changes continuously along each `` annulus '' of the `` spiral , '' whereas it shows abrupt discontinuities crossing the boundary of each annulus . as expected , the system is attracted on that region ( fig .
[ meml ] ) and the maps in figs .
[ t - vl ] and [ v - vl ] derive from the overplotting between figs .
[ meml ] and [ torrel1 ] .
+ the chaos actually emerges due to the combination of the step - like first - exit - time ( fig .
[ t - vl ] ) and of the piecewise return map ( fig .
[ v - vl ] ) for initial conditions .
this mechanism is the same as for the linearized model , for which we gave an analytical description .
the results we obtained for the piecewise linearized system therefore provide in our opinion a useful enlightening tool to understand both the original lorenz system and its eventual generalizations .
finally , in order to point out the connection between the original lorenz system and its piecewise linearized version , it is illuminating to compare the tent - like map shown by lorenz in his original work @xcite with the analogous map for our simplified system . describing the relation between two consecutive maxima of the coordinate @xmath23 , lorenz found a univocal tent - like map ( fig .
[ tenda ] ) . for our linearized system
we obtain similar results ( see fig .
[ tenda - lin ] ) .
the growing branch of the map is approximately linear , since for small values of @xmath245 the system evolution is confined in a half - space and , there , it essentially consists of an exponentially amplified oscillation . for large values of @xmath245
the map has the same shape of the lorenz tent - like map .
the linear growing of the left branch of the map is due to the simplification we made on the evolution on each side of @xmath45 .
the shape of the right branch shows that the piecewise linear system keeps all the complexity of the lorenz model , which can be attributed to the unpredictability of the @xmath45 crossing , which is the very origin of chaos in lorenz - like systems .
the main conclusions from the present paper can be summarized as follows .
we have reformulated in a classical mechanics form the dynamics of lorenz - like systems and showed that its three - dimensional phase - space dynamics can be mapped into an one - dimensional motion of a particle oscillating in a conservative quartic two - well potential , subjected to a viscous dissipation and to a memory forcing . starting from this interpretation , we have introduced a piecewise linearized version of the lorenz system ( belonging to a larger family of lorenz - like systems ) which substantially has the same properties of the original model with the advantage that it allows an analytical treatment .
+ the most evident aspect of chaotic regime is the unpredictability of the instant at which the center of the particle amplified oscillation changes .
chaos arises due to the combination of the step - like behavior of this time with the piecewise return map defining the crossing conditions .
this aspect has been singled out analytically , focusing on the piecewise linearized version . there , the exact equation for the time , at which the oscillation center changes , has been derived and the discontinuous dependence of this time on the crossing conditions has been shown analytically . + by means of numerical simulations we have verified that the highlighted mechanisms for the chaos to emerge survive also for the fully - nonlinear lorenz system , where analytical techniques are not applicable .
we would like to thank m. la camera and a. vulpiani for useful discussions and suggestions .
this work was partially supported by the infm project gepaigg01 .
d. v. was partially supported by the doctoral grants of the university of nice and by grants of the university of genova .
in the lorenz system , the following conditions for @xmath246 have to be satisfied in order to have three instable fixed points : @xmath247 the parameters @xmath248 are defined as follows : @xmath249 from ( [ prima ] ) and ( [ seconda ] ) it immediately follows that @xmath248 are positive too .
therefore , @xmath250 can be rewritten in terms of @xmath248 in the form @xmath251
r=1-\frac{4}{\beta(\beta(\alpha+1)- 2\eta)},\\[0.2 cm ] \nonumber b=-\frac{2\beta}{\beta(\alpha+1)- 2\eta}.\end{gathered}\ ] ] under the hypothesis @xmath252 , ( [ prima ] ) and ( [ seconda ] ) are equivalent to @xmath253 from which necessarily follows .
the family of straight lines @xmath172 is defined by the equation @xmath254 where @xmath255 + \ , ( \,\lambda_i\,(\lambda_i^2 + ( \lambda_0 + \lambda_r)^2\ , ) .
\end{array}\ ] ] we indicate with @xmath256 the ordered sequence of the zeros of @xmath257 : @xmath258 with @xmath259 the first derivative of @xmath173 is positive over the whole domain of definition . indeed one has @xmath260
\left(\lambda_i^2 + ( \lambda_0+\lambda_r)^2\right ) } { \lambda_i e^{-\lambda_0 \tau_1}+ e^{\lambda_r \tau_1}\big [ -\lambda_i \cos ( \lambda_i \tau_1 ) - ( \lambda_r+\lambda_0 ) \sin ( \lambda_i \tau_1 ) \big]^2 } \end{gathered}\ ] ] with @xmath261 the latter inequality can be easily proved . indeed : @xmath262 a parametrical ( not regular ) representation of the envelope of the family @xmath172 is obtained deriving @xmath160 and @xmath161 with respect to @xmath143 .
this can be done from the system @xmath263 \partial_{\tau_1 } a(\tau_1)\:\dot{\xi } _ 0\ ; + \ ; \partial_{\tau_1}b(\tau_1)\:\ddot{\xi}_0\ ; + \ ; \partial_{\tau_1}c(\tau_1)\,=\,0 \ ; .
\end{array}\right.\ ] ] the result @xmath183 is given by @xmath264 \big\ { -\lambda_0\big ( \lambda_i(1+e^{-\lambda_0 \tau_1 } ) \cos(\lambda_i \tau_1)-\lambda_i(e^{-(\lambda_0+\lambda_r)\tau_1}+ e^{\lambda_r \tau_1 } ) \big)\\[0.15 cm ] -(\lambda_r^2 + \lambda_i^2-\lambda_0\lambda_r ) ( e^{-\lambda_0 \tau_1}-1)\sin(\lambda_i \tau_1 ) \big\ } \bigg/ \\[0.15 cm ] \big ( \lambda_i e^{\lambda_r \tau_1 } + e^{-\lambda_0 \tau_1 } ( -(\lambda_0 + \lambda_r)\sin(\lambda_i \tau_1)-\lambda_i \cos(\lambda_i \tau_1)\big)\end{gathered}\ ] ] and @xmath265 \big\ { \lambda_0 ^ 2 \big ( \lambda_i(\cos(\lambda_i \tau_1)-e^{\lambda_r \tau_1 } ) + \lambda_r\sin(\lambda_i \tau_1)\big)\\[0.15 cm ] + ( e^{-\lambda_0 \tau_1}-1)(\lambda_i\cos(\lambda_i \tau_1 ) -\lambda_r \sin(\lambda_i \tau_1))(\lambda_r^2 + \lambda_i^2)+ \\[0.15 cm ] -\lambda_0\big ( e^{-\lambda_0 \tau_1}((\lambda_i^2-\lambda_r^2 ) \sin(\lambda_i \tau)+ 2\lambda_i\lambda_r\cos(\lambda_i \tau_1 ) ) -2 \lambda_i\lambda_r e^{-(\lambda_0+\lambda_r ) \tau_1 } \big)\big\}\bigg/ \\[0.15 cm ] \big ( \lambda_i e^{\lambda_r \tau_1 } + e^{-\lambda_0 \tau_1 } ( -(\lambda_0 + \lambda_r)\sin(\lambda_i \tau_1)-\lambda_i \cos(\lambda_i \tau_1)\big).\end{gathered}\ ] ] eq .
describes a single loop feedback system ( in the absence of input ) with linear transfer function @xmath266^{-1 } $ ] and nonlinear feedback @xmath267 .
andronov , a.a .
vitt , and s. e. khaiken , _ theory of oscillators _ ( pergamon , oxford , 1966 ) .
c.t . sparrow , j. math .
an . appl . * 83 * , 275 ( 1981 ) .
shaw and p.j .
holmes , j. sound vib . *
90 * , 129 ( 1983 ) .
t. matsumoto , l.o .
chua , and m. komuro , physica d * 24 * , 97 ( 1987 ) .
t. yang and l.o .
chua , int .
j. bifurcations chaos * 10 * , 2015 ( 2000 ) . for the sake of simplicity we frequently speak of `` initial conditions '' for a trajectory after a generic collision .
obviously , we refer to the conditions that a solution , which matches to a preceding trajectory , must satisfy in @xmath45 .
we always consider the evolution of the system in the steady - state and so far from the initial transient . | a novel view for the emergence of chaos in lorenz - like systems is presented . for such purpose ,
the lorenz problem is reformulated in a classical mechanical form and it turns out to be equivalent to the problem of a damped and forced one dimensional motion of a particle in a two - well potential , with a forcing term depending on the `` memory '' of the particle past motion .
the dynamics of the original lorenz system in the new particle phase space can then be rewritten in terms of an one - dimensional first - exit - time problem .
the emergence of chaos turns out to be due to the discontinuous solutions of the transcendental equation ruling the time for the particle to cross the intermediate potential wall .
the whole problem is tackled analytically deriving a piecewise linearized lorenz - like system which preserves all the essential properties of the original model . |
quantum effects strive to disappear for macroscopic objects .
typically , quantum effects become defamed out into their classical averages , and therefore manipulation of quantum states , relevant for quantum computation , becomes unsustainable at the macroscopic scale .
however , bose - einstein condensates ( becs ) constitute one important exception where quantum effects are perceptible on a macroscopic level . with the advent and rapid developing of laser cooling and trapping of neutral atoms over the past decade , micron - sized atomic gases at ultralow temperatures are routinely formed in the laboratory @xcite .
moreover , quantum optic techiques allow for an unprecedent versatile and precise control on internal degrees of freedom , putting cold atoms as one of the most prominent candidates to test complex aspects of strongly correlated matter and to applications in quantum information processing @xcite . quantum information has been introduced in cold atom systems at various levels @xcite .
one way consists in defining a qubit ( a two - state system ) via two internal states of an atom .
this approach , however , requires each atom to be addressed separately .
a similar problem appears when the qubit is introduced via a set of spatially localized states ( e.g. in adjacent wells of an optical lattice potential ) of an atom or a bec .
the complication is due to the fact that the number of atoms in a bec experiment significantly fluctuates from run to run . as a result ,
any qubit system dependent on the number of atoms becomes problematic .
a second way of producing qubits in these systems relies on collective properties of ultracold atoms . here
, a two - state system can be formed by isolating a pair of macroscopic states that are set sufficiently far away from the multiparticle spectrum . at the same time , however , the energy gap between these lowest states must remain small enough to allow measurable dynamics @xcite .
experiments performed in the double - well potential configuration are a pioneer example of such approach @xcite .
nevertheless , despite the appealing similarity with single - particle two - level states , double - well potentials lack to achieve a macroscopic superposition allowing for a measurable dynamics @xcite . to overcome the superposition issue ,
a more recent proposal based on bec superfluid current states in the ring geometry , analogous to the superconducting flux qubit @xcite , has been discussed @xcite .
more recently , the concept of _ phononic _ reservoir via the manipulation of the phononic degrees of freedom has pushed quantum information realizations to another level , comprising the dynamics of impurities immersed in becs @xcite and reservoir engineering to produce multipartite dark states @xcite .
another important difference in respect to quantum optical system is the possibility to use phononic reservoirs to test non - markovian effects in many - body systems @xcite .
the implementation of quantum gates has been recently proposed in ref .
@xcite .
another important family of macroscopic structures in becs with potential applications in quantum information are the so - called dark - solitons ( ds ) .
they consist of nonlinear localized depressions in a quasi-1d bec that emerge due to a precise balance between the dispersive and nonlinear effects in the system @xcite , being also ubiquitous in nonlinear optics @xcite , shallow liquids @xcite , magnetic films @xcite .
quasi one - dimensional becs with repulsive interatomic interaction are prone to inception of dark solitons by various methods , like imprinting spatial phase distribution @xcite , inducing density defects in bec @xcite , and by collision of two condensates @xcite .
the stability and dynamics of ds in becs have been a subject of intense research over the last decade @xcite .
recent activity in the field involve studies on the collective aspects of the so - called _ soliton gases _
@xcite , putting dark solitons as a good candidate to investigate many - body physics @xcite . in this paper , we combine the intrinsic nonlinearity in quasi one - dimensional becs to construct two - level states ( qubits ) with dark solitons .
as we will show , thanks to the unique properties of the ds spectrum , perfectly isolated two - level states are possible to construct . as a result
, a matter - wave qubit of a few khz energy gap is achieved .
the effect of decoherece due to the presence of phonons ( quantum fluctuations around the bacground density ) play the role of a proper quantum reservoir .
remarkably , due to their instrisic slow - time dynamics , bec phonons provide small decoherence rates of few hz , meaning that under typical experimental conditions , ds - qubits have a lifetime comparable to the lifetime of the bec , being only limited by the soliton quantum diffusion ( evaporation " ) . as we show below
, this effect is not critical and the qubit is still robust within the 100 [email protected] s time scale .
the paper is organized as follows : in sec .
ii , the properties of a single ds in a quasi-1d bec immersed in a second condensate are derived .
we start with the set of coupled gross pitaevskii equations and find under which conditions dss can uniquevocally define a two level atom ( qubit ) . in sec .
iii , we compute the coupling between phonons and dss . sec .
iv discusses the weisskopf - wigner theory to determine the emission rate of the qubit .
some discussion and conclusions about the implications of our proposal in practical quantum information protocols are stated in sec . v.
we consider two - coupled quasi-1d becs .
a quasi 1d gas is produced when the transverse dimension of the trap is larger than or of the order of the @xmath1-wave scattering length and , at the same time , much smaller than the longitudinal extension @xcite . at the mean field level , the dynamics of the system
is thus governed by the time - dependent coupled gross pitaevskii equations @xmath2@xmath3where @xmath4 ( @xmath5 ) is the one - dimensional coupling strength between particles in bec@xmath6 ( bec@xmath7 ) and @xmath8 is the inter - particle coupling constant , @xmath9 is the planck constant , @xmath10 is the mass of the atomic species .
we restrict the discussion to repulsive interatomic interacions , @xmath11 . in what follow ,
we assume that a dark soliton is present in bec@xmath6 and @xmath12 such that particles in bec@xmath7 do not interact and can therefore be regarded as a set of free interacting particles ( see fig .
[ fig_scheme ] ) .
thus , eq . ( [ gp2 ] ) can be written as @xmath13 where the soliton profile , resulting a singular nonlinear solution to eq .
( [ gp1 . ] ) , is given by @xcite @xmath14 here , @xmath15 is the background density which is typically of the order of @xmath16 m@xmath17 in elongated becs , and the healing length @xmath18 is of the order @xmath19 @xmath20 m . we also consider the experimentally accesible trap frequencies @xmath21 khz @xmath22 hz and the corresponding length amount to be the value @xmath23 @xmath20 m @xcite .
notice that the previous results can be easily generalized for the case of a gray solitons ( i.e solitons traveling with speed @xmath24 ) by replacing eq .
( [ eq_sol ] ) by @xmath25 , \label{eq_gray sol}\ ] ] where @xmath26 , @xmath27 , and @xmath28 is the bec sound speed @xcite .
therefore , the time - independent version of eq . ( [ sch . eq . without soliton ] ) reads @xmath29where @xmath30 . here , the dark soliton act as a potential for the particles of the reservoir . analytical solutions to eq .
( [ eq_reflectionless1 ] ) can be obtained by casting the potential term in the form of @xcite @xmath31 where @xmath32 .
the particular case of @xmath33 being a positive integer corresponds to the important case of the _ reflectionless _ potential @xcite , for which an incident wave is totally transmitted . for the more general case considered here , the energy spectrum associated to the potential in eq .
( [ potent . ] ) reads @xmath34 where @xmath35 is an integer . the number of bound states is given by @xmath36 , where @xmath37 denotes the integer part .
a two - level system ( qubit ) can be perfectly isolated when the value of @xmath38 ranges as @xmath39at the critical point @xmath40 the two - energy levels merge and the qubit is ill - defined .
finally , for @xmath41 , three - level systems ( qutrits ) can also be formed , but this case is out of the scope of the present work and will be discussed in a separate publication .
the features of the spectrum ( [ energy eigen states ] ) are illustrated in fig .
[ fig_spectrum ] . , while panel b ) illustrates the case of a two - level system ( qubit ) case obtaind for @xmath42 .
panel c ) illustrates the patological case @xmath43 for which a degeneracy in the two - level system is obtained .
a qutrit case is depicted in panel d ) for @xmath44 . ]
the mean - field soliton soluton in eq .
( [ eq_sol ] ) is accompained by quantum fluctuations ( phonons ) . in that case , the total wavefunction is given by @xmath45 , where @xmath46 with @xmath47 denoting bosonic operators satisfying the commutation relation @xmath48=\delta_{kq}$ ] .
@xmath49 and @xmath50 are amplitudes veryfing the normalization condition @xmath51 and are explicitly given by @xcite @xmath52 \right .
, \end{aligned}\]]and @xmath53 \right .
.\end{aligned}\]]similarly , the particles in bec@xmath7 are eigenstates of eq .
( [ eq_reflectionless1 ] ) , being therefore spannable in terms of the bosonic operators @xmath54 as @xmath55 where @xmath56 and @xmath57 .
the total hamiltonian may then be written as @xmath58 the first term @xmath59 represents the dark - soliton ( qubit ) hamiltonian @xmath60 where @xmath61 is the qubit gap frequency and @xmath62 is the corresponding spin operator .
the second term describes the phonon ( reservoir ) hamiltonian @xmath63 where the bogoliubov spectrum is given by @xmath64 and @xmath65 denotes the chemical potential .
the interaction hamiltonian @xmath66 between qubit and the reservoir is defined as @xmath67 which , with the prescriptions in eqs .
( [ eq_dec1 ] ) and ( [ eq_dec2 ] ) , can be decomposed as @xmath68 respectivelty containing zero , first and second order terms in the operators @xmath47 and @xmath69 .
owing to the small depletion of the condensate , and consistent with the bogoliubov approximation performed in eq .
( [ eq_dec1 ] ) , we ignore the higher - order term @xmath70 . the first part of eq .
( [ inter . hamilt . ] ) corresponds to a stark shift term of the type @xmath71 where @xmath72 .
the latter can be omitted by renormalizing the qubit frequency as @xmath73 in its turn , the first - order term @xmath74 is given by @xmath75 where @xmath76 as we can observe , eq .
( [ one hamiltonian ] ) contains intraband ( @xmath77 ) and interband ( @xmath78 ) terms .
however , for small values of the coupling between the system and the environment , the qubit transition can only be driven by near - resonant phonons , for which te interband coupling amplitude @xmath79 is much larger that the interband terms @xmath80 and @xmath81 ( see fig .
[ fig_rwa ] ) . as such , within the rotating - wave approximation ( rwa )
, we can safely drop the intraband terms to obain @xmath82where @xmath83 , @xmath84 and the coupling constant @xmath85 is explicitly given by @xmath86 we notice that the implementation of the rwa approximation also implied the dropping of the counter - rotating terms proportional to @xmath87 and @xmath88 that do not conserve the total number of excitations .
the accuracy of such an approximation can be verified _ a posteriori _ , provided that the emission rate @xmath89 is much smaller than the qubit transition frequency @xmath90 .
( dashed line ) and @xmath91 ( dotted - dashed line ) and inter - band @xmath92 ( solid line ) coupling functions . around the resonant values @xmath93 ( shadowed region ) ,
the interband term is dominant , allowing us to neglect the intra - band terms within the rotating - wave approximation . ]
neglecting the effect of temperature and other external perturbations , the only source of decoherence of a dark - soliton qubit are te sourrounding phonons . because cold atom experiments are typically very clean , and considering that the zero - temperature approximation is an excellent approximation for quasi-1d becs @xcite , we employ the wigner - weisskopf theory in order to compute the lifetime of the qubit .
we assume the qubit to be initially in its excited state and the field to be in the vacuum state . under such conditons
, the total system+reservoir wavefunction can be parametrized as @xmath94where @xmath95 and @xmath96 are the probability amplitudes .
the wigner - weisskopf ansatz ( [ time depend . eigen fun . ] ) is then let to evolve under the total hamiltonian in eq .
( [ eq_ham_final ] ) , for which the schrdinger equation yields the following evolution of the coefficients @xmath97 due to separation of time scales between the phonons and the decay process , we may assume that the coefficient @xmath95 evolves much slower than @xmath96 , which allows us to evoque the born approximation to write @xmath98where @xmath99 moreover , since we expect @xmath95 to varie at a rate @xmath100 , the relevant decay dynamics is expected to take place at times @xmath101 , which allows us to take upper limit of above integral to @xmath102 ( markov approximation ) .
therefore , we have @xmath103 where @xmath104 represents the cauchy principal part describing an additional energy ( lamb ) shift .
because it represents a small correction to the qubit energy @xmath90 , we do not compute its contribution explicitly .
therefore , the excited state amplitude decays exponentially as @xmath105 where @xmath106 is the population decay rate given as @xmath107 where @xmath108 . as depicted in fig .
[ fig_gamma ] , the decay rate @xmath89 is orders of magnitude smaller than the qubit gap @xmath90 , confirming that the evoquing both the rwa and the born - markov approximation can also be used for phononic systems .
remarkably , for a quasi-1d of chemical potential of few khz , we can obtain a qubit lifetime @xmath109 of the order of a second , a time comparable to lifetime of the bec itself .
notice that the value of @xmath110 ( and consequently the qubit natural frequency @xmath90 and lifetime @xmath111 ) can be experimentally tunned with the help of feshbach resonances .
the only immediate limitation to the performance of our proposal may be related to the dark - soliton quantum diffusion @xcite .
since they interact with the background phonons , they are expected to evaporate within the time scale @xmath112 . for typical 1d becs with @xmath113 @xmath20 m and @xmath114 mm / s , we estimate @xmath115 s , which reduces @xmath111 in about 20% . finally , by putting eqs .
( [ coefficient b ] ) and ( [ coefficient a + gamma ] ) together , we can evalute the evolution of the amplitude coefficient @xmath96 as @xmath116t}dt,\ ] ] which yields the following lorentzian spectrum @xmath117 ( panel a ) ) and decay rate @xmath89 ( panel b ) ) on the coupling constant @xmath110 .
the shadowed region corresponds to the range @xmath118 for which the quibit can be exactly defined .
the case @xmath43 produces a degenerate two - level system . ] as illustrated in fig .
[ fig_lorentzian ] .
it is observed that the lorentzian spectrum is narrower for a weak coupling constant @xmath110 . and @xmath119 .
in conclusion , we have shown that a dark soliton in a quasi one - dimensional bose - einstein condensate can produce a well isolated two - level system , which can act as a matter - wave qubit of energy gap of a few khz .
this feature is intrinsic to the nonlinear nature of bose - einstein condensates and does not require manipulation of the internal degrees of freedom of the atoms .
we observe that the decoherence induced by the quantum fluctuations ( phonons ) produce a finite qubit lifetime .
quite remarkably , leading calculations provide a qubit lifetime of the order of a few seconds , a time scale comparable to the duration of state - of - the art cold atomic traps . the only major limitation to the qubit robustness is the quantum diffusion of the soliton , which is estimated to reduce the qubit lifetime to around 20% its value .
this puts qubits made of dark solitons as good candidates to store information for large times ( @xmath120 s ) , offering an appealing alternative to quantum optical of solid - state platforms .
while dark solitons may not compete in terms of quantum scalability ( the number of solitons in a typical elongated bec is not expected to surpasse a few tens ) , their unprecedent coherence and lifetime will certainly make them attractive to the design of new quantum memories and quantum gates . moreover , due to the possibility of interfacing cold atomic clouds with solid - state and optical systems , our findings may inspire further applications in hybdrid quantum computers .
one of the authors ( h. t. ) acknowledges the security of quantum information group for the hospitality and for providing the working conditions during the early stages of this work . stimulating discussions with j. d. rodrigues
are acknowledged .
the authors also thank the support from the dp - pmi programme and fundao para a cincia e a tecnologia ( portugal ) , namely through the scholarship number sfrh / pd / bd/113650/2015 and the grant number sfrh / bpd/110059/2015.e.v.c .
acknowledges partial support from fct - portugal through grant no .
uid / ctm/04540/2013 .
j. denschlag , j. e. simsarian , d. l. feder , c. w. clark , l. a. collins , j. cubizolles , l. deng , e. w. hagley , k. helmerson , w. p. reinhart , s. l. rolston , b. i. schneider , and w. d. phillips , science * 287 * , 97 ( 2000 ) . | we study the possibility of using dark - solitons in quasi one dimensional bose - einstein condensates to produce two - level systems ( qubits ) by exploiting the intrinsic nonlinear and the coherent nature of the matter waves .
we calculate the soliton spectrum and the conditions for a qubit to exist .
we also compute the coupling between the phonons and the solitons and investigate the emission rate of the qubit in that case .
remarkably , the qubit lifetime is estimated to be of the order of a few seconds , being only limited by the dark - soliton death " due to quantum evaporation . |
in the papers @xcite the following monodromy problem is considered for the liouville theory on the sphere : given the singularities in the standard position @xmath1 and given the class i.e. the trace of the monodromy for a path encircling both singularities at @xmath2 and @xmath0 , find the value of the accessory parameter realizing such data .
such a problem intervenes in the process of the classical limit of the quantum four - point function ; the value of trace of the above described monodromy is then fixed by a saddle point procedure @xcite . in @xcite
the problem is solved by going over to the quantum formulation for the four - point function and taking the classical limit i.e. the limit in which the central charge goes to infinity . as the quantum conformal blocks are known as formal power series expansions in @xmath0 also the classical result so obtained is given as a formal power expansion in @xmath0 .
the procedure goes through a process of exponentiation of the quantum conformal blocks after which the classical limit @xmath3 is taken ; in such a limit heavy cancellations take part @xcite . several analytic @xcite and numerical @xcite calculations , also exploiting recursion formulae @xcite for the conformal blocks , support the validity of such a calculational scheme .
one suspects , on the other hand , that the same results should be obtainable just by exploiting the transformation properties of the ordinary differential equations which underlie the liouville theory at the classical level . in this note
we shall in fact consider the problem directly at the classical level .
in addition it appears that working without taking the singular limit in which the central charge goes to infinity one might control better the convergence region of the expansion of the accessory parameter as a function of the modulus @xcite .
the approach followed in this paper consists in computing the monodromy along a contour embracing @xmath2 and @xmath0 through the usual convergent iteration expansion for the solution of the heun equation .
the monodromy is computed along a contour which avoids the neighborhood of the origin where the kernel is singular and then we expand the result in @xmath0 . in so doing one
is faced to first order with the computation of integrals containing the product of two hypergeometric functions ; if one goes to the second order the product of four hypergeometric functions in a double integral appears . in this paper
we show how to compute analytically such integrals , which appear in the expansion of the solution of the volterra equation .
the complete first order result gives as a byproduct the value of the accessory parameter which coincides with the one derived in @xcite and re - derived in @xcite . for computing the integrals appearing in the first order result
, we exploit the transformation property of the solution of the differential equation under @xmath4 . in the calculation of the second order such
a technique is not sufficient and we need a non invertible transformation which at the infinitesimal level is related to the operator @xmath5 .
contrary to the @xmath4 transformations this is not one - to - one in the complex plane . on the other hand
the procedure we shall describe , involves only the solutions along the real @xmath6 axis for @xmath7 , and there for @xmath8 the transformation is well defined . after developing such tools
we give the complete second order computation for the accessory parameter .
in so doing we employ a formalism apt to be extended to higher order computation .
the second order result agrees with the one obtained in @xcite by considering the classical limit of the quantum conformal blocks combined with the exponentiation hypothesis and thus it lends a strong support to the exponentiation hypothesis of the conformal blocks in the @xmath3 limit .
we discuss also the extension of the procedure to higher orders .
obviously , as the determination of the accessory parameter @xmath9 is always obtained through the solution of an implicit equation , the fact that the function @xmath10 which represents the energy momentum tensor , has radius of convergence @xmath11 in @xmath0 , for @xmath12 , does not assure that the expansion of @xmath9 in @xmath0 has the same radius of convergence . for achieving rigorous lower bounds on such a radius of convergence , methods
similar to those developed in @xcite for the convergence in the coupling strength should be applied .
the developed technique can also be applied to the problem of the punctured torus .
the ordinary differential equation associated with the monodromy problem is @xmath13 with @xmath14 where @xmath15 .
@xmath9 is the accessory parameter to be fixed so that the monodromy along a contour encircling both @xmath2 and @xmath0 has trace @xmath16 and as such it will depend both on @xmath0 and @xmath17 .
we have @xmath18 and @xmath9 is related to to one used in @xcite which we call @xmath19 , by @xmath20 and thus @xmath21 and @xmath22'|_{x=0 } -c(0)$ ] . expanding in @xmath0 we have @xmath23
@xmath24 it is our interest to compute the class of the monodromy along a circuit enclosing both the origin and @xmath0 .
working near the origin is difficult due to the singular nature of the kernel .
instead we shall compute the same monodromy along the circuit shown in fig.1 .
the great advantage in performing such a change in the contour is the fact that the expansion in @xmath0 of @xmath10 along the contour is no longer singular and actually is convergent with convergence radius @xmath11 .
@xmath25 will denote the complex @xmath26 being @xmath27 two independent solutions of @xmath28 , canonical at @xmath29 .
they are given by @xmath30 @xmath31 the constant wronskian is easily computed at @xmath29 @xmath32 the unperturbed monodromy is computed as follows .
we start from the @xmath25 for real @xmath6 , @xmath33 .
the continuation to the upper side of the cut @xmath34 is given by @xmath35 @xmath36 whose asymptotic behavior for large @xmath6 is @xmath37 @xmath38 being a well known matrix @xmath39 similarly @xmath40 and @xmath41 we start from @xmath42 eq.([y- ] ) , whose continuation to the upper side of the cut is eq.([y+ ] ) . taking the turn of @xmath43 at infinity
we go back to @xmath42 having encircled the origin and @xmath0 and we obtain the monodromy matrix to lowest order @xmath44 one easily checks that @xmath45 .
the first order corrections to @xmath25 is provided by @xmath46 with @xmath47 @xmath48 . following the procedure illustrated above using ( [ firstiteration ] ) instead of the unperturbed @xmath25 we obtain the new monodromy matrix @xmath49 , @xmath50 with @xmath51 , and we have @xmath52 where @xmath53 @xmath54
we have due to eqs.([y1t1+],[y2t2+],[y1t1y2t2- ] ) @xmath55 and thus @xmath56\ ] ] and to determine @xmath57 we must impose @xmath58 .
the second order result is obtained by iterating once the result with @xmath59 and adding also the contribution obtained by replacing in ( [ firstiteration ] ) @xmath60 with @xmath61 .
the computation of the first order result is reasonably simple .
we shall adopt here a formalism which is apt to be extended to the second and higher order computation .
we are faced to compute the integrals appearing in eqs.([delta11],[delta22 ] ) with @xmath62 given in eq.([qexpansion ] ) .
more generally we shall compute analytically the indefinite integrals @xmath63 for @xmath64 and where @xmath65 and @xmath66 take the value @xmath11 and @xmath67 . for @xmath68
such integrals can be computed by exploiting the transformation properties under @xmath4 of the solutions .
let us consider the equation @xmath69 @xmath70 the solutions of eq.([requation ] ) are @xmath71 and similarly for @xmath72 . then writing @xmath73 @xmath74 using @xmath75 and @xmath76 we have @xmath77 we find @xmath78 on the other hand the change @xmath17 into @xmath79 induces in @xmath80 the change @xmath81 leaving the singularities at @xmath29 and at @xmath82 unchanged .
this time the related change @xmath83 is simply given by @xmath84 and we can again apply eq.([firstintegral ] ) replacing @xmath85 with @xmath86 . combining with ( [ firstintegrals ] )
it provides us with the integrals of type ( [ generalintegrals ] ) with @xmath87 and with @xmath88 appearing in the first order computation , in terms of hypergeometric functions and derivatives thereof .
the integrals appearing in ( [ delta11],[delta22 ] ) have the upper limit infinity , for which the derived formulae ( [ firstintegral ] ) also hold . as the asymptotic behavior of the hypergeometric functions
are given @xcite by simple powers of @xmath6 multiplied by gamma functions , we have that such integrals are expressed in terms of the functions @xmath89 and @xmath90 , where @xmath91 .
we shall denote by @xmath92 the expression @xmath93 they form a basis for the the derivative of @xmath10 with respect to @xmath0 to any order and we shall set @xmath94 explicitly we find @xmath95 and @xmath96 where we defined @xmath97 moreover in eq.([trdeltam ] ) we have @xmath98 @xmath99 with @xmath100 . for future use
we report below also the values of @xmath101 and @xmath102 .
@xmath103 @xmath104 the vectors @xmath105 are given by @xmath106 with @xmath107 @xmath59 , see eq.([qexpansion ] ) , in the basis @xmath108 is represented by the vector @xmath109 and thus , see eq.([trdeltam ] ) , the equation for @xmath57 becomes @xmath110 i.e. @xmath111 giving @xmath112'|_{x=0}-c(0)\ ] ] which is the result of @xcite obtained by taking the @xmath3 limit of the conformal blocks .
the equation which gives @xmath113 is provided by the vanishing of the coefficient of @xmath114 in the expansion of @xmath115 i.e. @xmath116 the second order change in the functions @xmath27 is given by the direct contribution due to @xmath117 and by the second iteration of the contribution of @xmath59 .
with regard to the direct contribution we have to compute @xmath118 with @xmath119 given in eq.([qexpansion ] ) , where the new integrals @xmath120 appear . in the basis @xmath121 @xmath117
is represented by the vector @xmath122 the @xmath123 can not be computed by performing an @xmath4 transformation .
we shall exploit the new transformation @xmath124 and use the schwarzian transformation of @xmath125 and the @xmath126-form nature @xcite of the solutions @xmath127 .
the above transformation gives rise to @xmath128 where @xmath129 is the schwarz derivative of the transformation and the new solutions are @xmath130 reverting to the @xmath6-notation for the variable and denoting with the dot the derivative w.r.t .
@xmath131 we have @xmath132 with @xmath133 from eq.([tkva ] ) we have @xmath134 contrary to the @xmath4 transformations the transformation ( [ lm2transf ] ) is not one - to - one in the complex plane .
nevertheless for @xmath135 the transformation is well defined , i.e. non singular along the line @xmath136 which is our range of application .
we can then exploit again the integration formula ( [ firstintegral ] ) , with @xmath85 replaced by eq.([dottva ] ) and @xmath137 by @xmath138 of eq.([dotr ] ) .
one can also verify the correctness of the result by taking explicitly the derivative w.r.t .
@xmath6 of the obtained formula .
we introduce the three dimensional vectors @xmath139 , with @xmath140 which represent the matrix elements of the variation of @xmath80 under respectively the transformation of eq.([requationr ] ) ( @xmath141 ) , the of eq.([deltanuchange ] ) ( @xmath142 ) and of the transformation of the above eq.([lm2transf ] ) ( @xmath143 ) .
we have @xmath144 with @xmath145 and @xmath146 the inversion of eq.([tjk ] ) provides the values of the fundamental matrix elements @xmath147 , @xmath148 .
we notice that the procedure can be extended to all values of @xmath149 by considering the variation of the equation @xmath150 under the transformation @xmath151 in addition due to the structure of the matrices @xmath152 such a procedure is purely iterative , i.e. known @xmath147 for @xmath153 the computation of @xmath154 , provides directly @xmath155 .
in the present case we have @xmath156 we come now to the second iteration of @xmath59 .
given the green function @xmath157 the expression we are confronted with , for the second order change of @xmath158 is @xmath159 with @xmath160 given by eq.([firstiteration ] ) .
the indefinite integrals appearing in ( [ seconditeration ] ) can be obtained from the equation @xmath161 with @xmath162 where we shall impose @xmath163 with @xmath164 and @xmath57 given by eq.([c1 ] ) . in order to fit the coefficients of @xmath108 in @xmath59 we have to allow in principle for two parameters @xmath165 and @xmath166 .
we find @xmath167 notice that the transformation leading from @xmath168 to @xmath169 is of the same type as the one appearing in ( [ requation],[requationr ] ) of which we know the solutions .
it follows that after replacing @xmath170 and @xmath57 in ( [ newq1 ] ) with their values , the simplest method to compute the matrix elements @xmath171 is to use the transformation ( [ simpleshift ] ) with @xmath131 replaced by @xmath172 .
one can easily prove that @xmath173 has the correct boundary condition @xmath174 with coefficient @xmath11 at @xmath29 , as imposed by the solution of the volterra equation and thus @xmath175 equals @xmath176 i.e. it is the first order correction to @xmath177 .
it is expressed in terms of derivatives of the hypergeometric function .
the same holds for @xmath178 .
then the integrals @xmath179 appearing in ( [ seconditeration ] ) can be computed as follows . using @xmath180 and @xmath181 we have @xmath182
notice that @xmath183 is not equal to the @xmath117 of eq.([qexpansion ] ) but it will be expedient for computing the l.h.s . of ( [ secondintegral ] ) .
@xmath183 in the base @xmath121 , is represented by the vector @xmath184 we know @xmath185 and @xmath186 and as a result we know the l.h.s . of eq.([secondintegral ] ) thus providing the second iteration of the volterra equation in terms of hypergeometric functions and derivatives thereof .
explicitly we find in eq.([secondordereq ] ) @xmath187 the computation of the term @xmath188 in eq.([secondordereq ] ) requires simply the knowledge of @xmath189 which we have already computed and it cancels the term @xmath190 in the curly brackets in the above equation . to summarize the equation for @xmath113
is given by @xmath191 due to the structure of @xmath192 , see eq.([tjk3dim ] ) , the vector @xmath193 has a single entry different from zero and we have for @xmath194''_{x=0}-2 c'(0)-2 c(0)$ ] @xmath195}{\delta_\nu}\\ & -&\frac{(c(0)+ 3\delta-3c^2\delta_\nu ) [ 3\delta_1 ^ 2 + 3\delta_\nu^2 + 3\delta_\infty(1+\delta_\infty)+ \delta_\nu(3 + 2\delta_\infty)-3\delta_1(1 + 2\delta_\nu+2\delta_\infty ) ] } { \delta_\nu(3 + 4\delta_\nu)}\nonumber\end{aligned}\ ] ] where @xmath170 , @xmath57 and @xmath165 are given respectively by eqs.([c0],[c1],[c ] ) .
the value of @xmath113 agrees with the one obtained in @xcite and @xcite by taking the @xmath3 limit of the conformal blocks thus providing strong support to the exponentiation hypothesis .
the procedure can also be pushed to higher order even if we need a systematic organization of the mixed contributions .
integrals similar to those discussed here appear in the accessory parameter problem for the torus , dealt with in @xcite .
there the logarithmic part was derived to all orders and compared with success with the saddle point prediction on the quantum theory , while the presence of integrals of the above mentioned type hampered the analytical evaluation of the @xmath196 term of the expansion , @xmath196 being the nome of the torus .
now we have the possibility of computing analytically not only the @xmath196 term but also the @xmath197 term .
in this paper we developed , for the monodromy problem considered in papers @xcite an analytical technique to compute the expansion of the accessory parameter in term of the invariant cross ratio directly , without taking the limit of the quantum conformal blocks for infinite central charge . in the first order computation
it is shown how the integrals containing products of two hypergeometric functions , which appear in the first iteration of the volterra equation can be computed analytically in terms of @xmath89-functions and derivatives thereof , i.e. @xmath90-functions . the method is to exploit the transformation properties of the kernel under @xmath198 and the @xmath199-form nature of the solutions .
the computation to second order involves double - integrals of products of four hypergeometric functions , and we show how also these can be computed analytically . in such second order computation
we need a transformation which at the infinitesimal level is akin to the @xmath200 generator of the virasoro algebra .
the transformation is not one - to - one on the complex plane but is well defined in the region needed for our computations .
both our first and second order results agree with the ones obtained from the classical limit of the quantum conformal blocks under the exponentiation hypothesis @xcite and thus they lend strong support to such exponentiation hypothesis in the classical @xmath3 limit . with regard to the extension to higher orders
we have shown how the fundamental matrix elements @xmath147 can all be computed by a simple iterative procedure . as it happens already to second order in the @xmath201-th order computation one has both direct contribution from the @xmath201- order term in the expansion of the kernel and mixed contributions due to lower order expansion .
one should need a systematic organization of such mixed contribution to go to arbitrary order . in the full treatment of the torus with one source @xcite , integrals of product of hypergeometric function , similar to those which we discussed above , appeared .
the possibility of computing them analytically will allow an extension of the results reported in @xcite . | we consider the monodromy problem for the four - punctured sphere in which the character of one composite monodromy is fixed , by looking at the expansion of the accessory parameter in the modulus @xmath0 directly , without taking the limit of the quantum conformal blocks for infinite central charge .
the integrals which appear in the expansion of the volterra equation , involve products of two hypergeometric functions to first order and up to four hypergeometric functions to second order .
it is shown that all such integrals can be computed analytically .
we give the complete analytical evaluation of the accessory parameter to first and second order in the modulus .
the results agree with the evaluation obtained by assuming the exponentiation hypothesis of the quantum conformal blocks in the limit of infinite central charge .
extension to higher orders is discussed .
ifup th/2014 - 1 + .8truecm * on the monodromy problem for the four - punctured sphere * 1.2truecm pietro menotti + _ dipartimento di fisica , universit di pisa _
+ _ largo b. pontecorvo 3 , i-56127 , pisa , italy _ + _ e - mail : [email protected]_ + 0.8truecm april 2014 1.2truecm |
the new spectral type l is characterized by the red color possibly due to the dust extinction while type t by the volatile molecules such as methane and water .
initially , we considered a fully dusty model ( model b ) and a dust - segregated model ( model c ) for the l and t type dwarfs , respectively .
however , it was difficult to explain why such different cases are realized in different types of cool dwarfs . also , increasing observations on cool dwarfs could not be explained even by the use of these different kinds of models .
we then showed that all these difficulties can be resolved by models with a warm dust cloud deep in the photosphere ( tsuji 2001 ) , and we extended this idea to a grid of unified cloudy models ( ucms)(tsuji 2002 ) . in this contribution , we reexamine some observed data with our ucms being updated with the use of the new solar carbon and oxygen abundances by allende prieto , lambert , & asplund ( 2001 , 2002 ) .
in the photospheres of l and t dwarfs , dust forms at its condensation temperature ( @xmath1 ) but grows too large to be sustained in the photosphere at the critical temperature ( @xmath2 ) .
for this reason , only small dust grains survive in the temperature range of @xmath3 .
this means a formation of a dust cloud whose temperature is fixed at rather high ( note that @xmath4k ) independently of @xmath5 . since @xmath6 at @xmath7 , the dust cloud appears at the optically thin region ( @xmath8 ) in l dwarfs whose @xmath9 s are relatively high and at the optically thick region ( @xmath10 ) in t dwarfs whose @xmath9 s are lower .
then the dust will give direct observable effect in l dwarfs , which in fact appear to be dusty , but the dust cloud is situated too deep to give significant observable effect in t dwarfs .
we then generated a grid of non - grey model photospheres in radiative - convective equilibrium under the presence of the dust cloud .
we first analyze infrared colors based on our models . as an example , observed values of @xmath11 ( leggett et al .
2002 ) plotted against spectral types ( geballe et al .
2002 ) and the predicted ones based on our cloudy models ( ucms ) plotted against @xmath5 values are compared in fig.1a .
the basic feature of the observed color , which is redder in the later spectral types at first but turns to blueward at about l5 , can be well accounted for by our ucms , but not at all by the fully dusty models b nor by the dust segregated models c. further , the maximum value of @xmath11 at about l5 can be explained by our ucms with @xmath12k , and thus the critical temperature , @xmath13 , can be well constrained empirically . next , effective temperatures can be estimated so that the best fit can be obtained between the observed and predicted colors , and fig.1a already shows the result of such a fit .
the resulting @xmath9 - sp.type relationship is shown in fig.1b by the solid line ( also table 2 in tsuji 2002 ) .
for comparison , empirical @xmath9 values based on the measured bolometric fluxes and parallaxes ( assuming a fixed radius ) by burgasser ( 2002 ) are shown by the filled circles in fig.1b .
the agreement is not so good especially in the early l dwarfs and further works should be needed before we could have a convincing @xmath9 scale .
once the infrared colors can be accounted for by our ucms , the infrared spectra may in principle be accounted for as well . however ,
some details of the molecular spectra depend on the chemical composition besides @xmath5 , and no satisfactory interpretation was possible with the use of the solar abundances by anders & grevesse ( 1989 ) ( see fig.12 in tsuji 2002 ) .
thanks to the recent revisions not only of o but also of c abundances by allende prieto et al.(2001 , 2002 ) , however , the observed spectrum of the l dwarf prototype gd165b ( jones et al .
1994 ) can be reasonably fitted at last with the predicted one based on our ucm ( @xmath14k ) of @xmath15k ( from fig.1b for this l3@xmath161 dwarf ) , but never be fitted with those based on the models b and c , as shown in fig.2 .
a unified spectral classification of l and t dwarfs by geballe et al.(2002 ) is based on the h@xmath17o and ch@xmath18 indices derived from the near ir spectra .
we compare one of them ( h@xmath17o 1.2@xmath19 m index ) with the predicted ones from our model spectra in fig.3a . also , observed ews of ki 1.2432@xmath19 m ( burgasser 2002 ) are compared with the predicted ones in fig.3b .
the rather modest increase of h@xmath17o index in the l dwarf regime is due to the compensation of the increasing h@xmath17o abundance and the increasing dust extinction towards the later l dwarfs , in which the dust cloud is still in the optically thin photosphere .
ki ews show the minimum at about l5 because the dust column density of the cloud in the optically thin region is the largest at about l5 ( or @xmath20k ) . after the dust cloud penetrates into the optically thick region at @xmath21k , the region above the cloud
is dominated by the increasing amount of volatile molecules and mildly volatile atoms ( e.g. k ) towards the later t dwarfs .
this explains the rapid increase of the h@xmath17o indices as well as the upturn of ki ews in the t dwarf regime .
thus , our ucms will provide the physical basis for the spectral classification of l and t dwarfs .
it is to be noted , however , that what matters in the l t classification is the formation of the dust cloud at different optical depths @xmath22 ( though at nearly fixed temperature near @xmath1 ) rather than the amount of dust predicted by the thermochemical theory .
this shows a marked contrast to the classical spectral types o m , which are directly related to the amount of ions , atoms , and molecules predicted by the ionization and dissociation theory .
we showed that the observed spectra and colors of l and t dwarfs can be accounted for consistently with our ucms . on the other hand , our previous models such as the fully dusty models b and the fully dust - depleted models c ( and , by implication , more or less similar models by other authors ) are far too short to explain the observed colors ( fig.1a ) , spectra ( fig.2 ) and spectral indices ( fig.3 ) of cool dwarfs , especially from middle l to early t. thus the presence of the dust cloud deep in the photosphere should be an essential feature of l and t dwarfs , as also noted by marley et al.(2002 ) from a different approach based on the planetary theory .
we followed stellar approach throughout and pursued a simple model as far as possible , but it is clear that this is only an initial step towards understanding the complicated phenomena in ultracool dwarfs . | based on a simple thermodynamical argument , we proposed a cloudy model with a warm dust cloud deep in the photosphere .
we showed , for the first time , that a single grid of model photospheres ( @xmath0k ) offers a natural explanation not only for the division of dwarfs cooler than m into the two distinct types l and t but also for the changes of the spectra and colors along the l t spectral sequence .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
jain and singh [ 1 ] report a study of electron - pair production following 160 a@xmath0gev @xmath1pb bombardment of nuclear emulsions , in which the energy and opening angles of pairs produced at distances greater than 50 @xmath2 from identified interaction vertices were analysed to determine the invariant masses of presumed neutral particles _ x _ decaying by the @xmath3 channel . derived neutral particle masses from 2 to 84 mev / c@xmath4 were reported , with lifetimes ranging from @xmath5 to @xmath6 s ( fig .
1b , 2a , 3 and 4 of [ 1 ] ) .
such large derived masses do not appear to be consistent with the total pair energy and opening angle data that are reported ( fig .
1f of [ 1 ] ) .
a re - analysis of the data presented in fig .
1f indicates presumed neutral particle _ x _ invariant masses of 1.5 to 21 mev / c@xmath4 , with lifetimes between @xmath7 and @xmath8 s. this mass and lifetime range are consistent with previous indications of light neutral particles decaying to @xmath9 from cosmic ray [ 2 - 3 ] , emulsion bombardment [ 4 - 9 ] and nuclear decay [ 10 - 16 ] data .
figure 1f of [ 1 ] shows a scatter plot of total electron - pair energy @xmath10 versus opening angle @xmath11 .
a total of 62 of the reported 1220 @xmath9 pairs fall above borsellino s [ 17 ] most probable opening angle @xmath12 at an invariant mass 1.02 mev / c@xmath4 as plotted together with the data in fig .
1f of [ 1 ] .
these 62 events are , therefore , the only events that can be interpreted as indicating the decays of massive particles into @xmath9 pairs . the total energies and opening angles of these 62 events are reproduced in fig . [
fig : rawdata ] , together with plots of borsellino s most probable opening angle @xmath12 and 1.5 times borsellino s angle .
the invariant mass @xmath13 for a presumed neutral particle _ x _ decaying to @xmath9 pairs was computed as : @xmath14 where @xmath15 is the electron rest mass and @xmath16 and @xmath17 are the pair electron energies , here taken to be equal on the basis of the very small energy divergences reported in fig .
1e of [ 1 ] .
the computed invariant masses are shown in fig . [
fig : massdata ] .
the _ x_-particle lifetimes reported in [ 1 ] could not be confirmed directly , as the measured interaction vertex to pair vertex distances are not given and figs 1b and 1f of [ 1 ] can not be compared on a point - for point basis .
the highest density of events is reported [ 1 ] to be between 50 and 300 @xmath18 m of an interaction vertex , and this distance range was used to calculate _ x_-particle lifetimes of @xmath5 to @xmath6 s [ 1 ] . from the mass range shown in fig .
[ fig : massdata ] , the _ x_-particle lifetime @xmath19 can be estimated to be between @xmath20 for a 1 mev / c@xmath4 _ x_-particle travelling 50 @xmath18 m and @xmath8 for a 10 mev / c@xmath4 _ x_-particle travelling 300 @xmath18 m , i.e. the revised mass calculations yield a median lifetime of approximately @xmath5 s , an order of magnitude lower than reported in [ 1 ] .
evidence for neutral particles with masses less than 10 mev decaying into @xmath9 pairs has previously been reported from cosmic ray [ 2,3 ] , emulsion bombardment [ 4 - 9 ] and nuclear decay [ 10 - 16 ] data .
the structure of these particles is unknown .
however , the angular distributions of @xmath9 pairs emitted in a 10.96 mev 0@xmath21 @xmath22 0@xmath23 magnetic monopole ( m0 ) transition in @xmath24o forbidden to both single-@xmath25 and internal pair - conversion ( ipc ) modes indicates that an _ x_-boson of approximately 10 mev / c@xmath4 is the primary decay product in this transition [ 13 - 14 ] .
similarly , the analysis reported in [ 4,5 ] suggests that 1.1 , 2.1 and 9 mev / c@xmath4 bosons are primary products of @xmath26c and @xmath27ne bombardments of nuclear emulsions .
these candidate light neutral bosons have been tentatively interpreted as potential components of light dark matter [ 18 ] .
the present re - analysis of the 160 a@xmath0gev @xmath1pb emulsion data reported by jain and singh [ 1 ] brings the majority of events observed in these reactions into a mass and lifetime range consistent with earlier reports of light neutral bosons decaying into @xmath9 pairs .
a comprehensive re - analysis of data from multiple sources indicating the existence of such particles is currently under way [ 19 ] .
we wish to acknowledge keith griffioen ( william and mary ) and andries van der schaaf ( university of zurich ) for useful comments .
[ 1 ] jain pl and singh g 2007 _ j. phys .
g : nucl . part . | a re - analysis of data from electron - pair production following 160 a@xmath0gev @xmath1pb bombardment of nuclear emulsions indicates the production and decay of neutral particles of significantly lower invariant mass and shorter lifetimes than previously claimed ( _ j .
phys .
g : nucl . part .
phys . _
* 34 * ( 2007 ) 129 - 138 ) .
# 1nucl .
phys .
* a#1 * # 1nucl .
phys .
* b#1 * # 1phys .
rev .
lett . * # 1 * # 1phys .
rev .
c * # 1 * # 1phys .
rev .
d * # 1 * # 1phys .
lett .
* b#1 * # 1mod .
phys .
lett . *
# 1 * # 1z .
phys .
* c#1 * # 1rev .
mod .
phys . *
# 1 * |
hydrodynamical and microphysical processes of baryonic matter play an important role in structure formation at the smaller sub cluster scales . indeed , microphysics can play a more important role than gravity , especially when the cooling time of the gas is much shorter than the dynamical or hubble times .
protogalactic clouds , which may form , for example , at high redshifts in cdm models can collapse through cooling instabilities to form an early generation of stars .
the feedback from primordial stars can change the physical state of the pre galactic medium and thus have considerable influence over the subsequent formation of stars and galaxies and the general state of the intergalactic medium ( couchman & rees 1986 ; tegmark et al .
microphysical processes are also very important at the center most regions of cosmological sheets or pancakes
. originally studied by zeldovich ( 1970 ) in the context of neutrino - dominated cosmologies , sheets are ubiquitous features in nonlinear structure formation simulations of cdm - like models with gas , and manifest on a spectrum of length scales and formation epochs .
cooling processes occur on very short time scales at the center most densest parts of the pancake structures where stars and galaxies can form from the fragmentation of the gas .
( bond et al .
1984 ; shapiro & struck marcell 1985 ; yuan , centrella & norman 1991 ; anninos & norman 1996 ) .
it is well known that when nonequilibrium atomic reactions are properly taken into account , the cooling time can be shorter than the hydrogen recombination time and that gas which cools to @xmath9 will likely cool faster than it can recombine .
the effect of this nonequilibrium cooling is to leave behind a greater residual of free electrons and ions , as compared to the equilibrium case .
the free electrons can be captured by neutral hydrogen to form @xmath5 that subsequently produce hydrogen molecules .
if large concentrations of molecules can form , the cooling is dominated by the vibrational / rotational modes of molecular hydrogen which acts to efficiently cool the gas to about @xmath10 , thereby reducing the jeans mass of the gas .
hydrogen molecules can therefore play a crucial role in the formation of stars as they provide the means for cloud fragments to collapse and dissipate their energy . however , the typically high computational requirements and technical difficulties needed to solve the chemical rate equations relevant for @xmath7 production in hydrodynamic flows have forced previous authors to impose simplifying assumptions such as the steady state shock condition which reduces the problem to zero dimension . in this case , only the time development of the hydrodynamic and thermodynamic variables are solved ( izotov & kolesnick 1984 ; maclow & shull 1986 ; shapiro & kang 1987 ; kang & shapiro 1992 ) .
all of these studies have consistently found that the mass fraction of @xmath7 can reach @xmath11 behind sufficiently strong shocks , which is adequate to cool the gas to temperatures of order @xmath10 well within a hubble time .
more recently haiman , thoul & loeb ( 1995 ) have investigated the formation of low mass objects in one dimensional numerical calculations .
they confirm the importance of @xmath7 in the collapse of spherically symmetric isolated objects at high redshifts .
although much insight has been gained about the chemical aspects of molecular hydrogen formation and cooling , it remains to incorporate chemical reaction flows in realistic cosmological models .
this paper discusses a method that we have developed for solving the kinetic rate equations with multi species chemistry in nonequilibrium and self consistently with the hydrodynamic and @xmath12body equations in an expanding flrw universe .
the method is based on a backward differencing formula ( bdf ) for the required stability when solving stiff sets of equations , and is designed for both accuracy but especially speed so that it may be used in three dimensional codes with a minimal strain on computational resources . in all , we solve for 28 kinetic reactions including collisional and radiative processes for nine different species : @xmath13 , and @xmath8 , which we track individually with their unique mass transport equations .
we have also implemented a comprehensive model for the radiative cooling of the gas that includes atomic line excitation , recombination , collisional ionization , free - free transitions , molecular line excitations , and compton scattering of the cosmic background radiation by electrons .
the set of hydrodynamic , @xmath12body , kinetic and cosmological equations that we solve are summarized in
[ sec : equations ] .
section [ sec : numerical ] discusses the bdf method and its integration into the hydrodynamic solver .
several tests of our code are presented in
[ sec : codetests ] , including 1d radiative shock waves , 2d cosmological sheets , and fully 3d simulations of cdm cosmological evolutions in which we compare the bdf method to results obtained when the packaged routine lsodar ( hindmarsh 1983 ; petzold 1983 ) is substituted in its place .
we provide concluding remarks in [ sec : summary ] .
finally we note that a companion paper has been written which discusses the chemical model and cooling functions in more detail ( abel et al .
1996 ) . in that paper
, we motivate the model and argue its comprehensive nature in the choice of reactions , insofar as the formation of hydrogen molecules in cosmological environments is concerned .
we also provide more up to date and accurate fits to the different rate coefficients . for completeness ,
we tabulate the reaction list and cooling processes in appendices a and b of this paper , but leave the rate coefficients to abel et al .
the hydrodynamical equations for mass , momentum and energy conservation in an expanding frw universe with comoving coordinates are @xmath14 @xmath15 + 5{\frac{\dot{a}}{a}}{\rho_b}{{v}_{b , i } } = - { \frac{1}{a^2}}{\frac{\partial p}{\partial x_i } } - { \frac{\rho_b}{a^2}}{\frac{\partial \phi}{\partial x_i } } , \label{hydromom}\ ] ] @xmath16 where @xmath17 , @xmath18 and @xmath19 are the baryonic density , pressure and specific internal energy defined in the proper reference frame , @xmath20 is the comoving peculiar baryonic velocity , @xmath21 is the comoving gravitational potential that includes baryonic plus dark matter contributions , @xmath22 is the cosmological scale factor , and @xmath23 and @xmath24 are the microphysical cooling and heating rates . the equations for collisionless dark matter in comoving coordinates are @xmath25 the baryonic and dark matter components are coupled through poisson s equation for the gravitational potential @xmath26 where @xmath27 is the total density and @xmath28 is the proper background density of the universe .
the cosmological scale factor @xmath29 is given by einstein s equation @xmath30^{1/2}\ ] ] where @xmath31 is the density parameter including both baryonic and dark matter contributions , @xmath32 is the density parameter attributed to the cosmological constant @xmath33 , and @xmath34 is the present hubble constant .
in addition to the usual hydrodynamic equations ( [ hydromass ] ) ( [ hydroenergy ] ) , we must also solve equivalent mass conservation equations for the densities @xmath35 of each of the nine separate atomic and molecular species that we track @xmath36 where the signs of each term on the right hand side depend on whether the process creates or destroys the species @xmath35 .
the @xmath37 are rate coefficients for the two body reactions and are functions of the gas temperature @xmath38 .
explicit analytic fits for these coefficients over a broad range of temperatures , and a general discussion of the relevant chemical reactions , can be found in abel et al .
( 1996 ) . in all
, we include 28 rate coefficients , one for each of the chemical reactions shown in appendix a. the @xmath39 in equation ( [ hydrospecies ] ) are integrals due to photoionizations and photodissociations @xmath40 where @xmath41 is the intensity of the radiation field , @xmath42 is the flux , @xmath43 are the cross - sections for the photoionization and photodissociation processes , and @xmath44 are the frequency thresholds for the respective processes .
we note that the nine equations represented by ( [ hydrospecies ] ) are not all independent . the baryonic matter is composed of hydrogen and helium with a fixed primordial hydrogen mass fraction of @xmath45 .
hence we have the following three conservation equations @xmath46 where @xmath47 is the number density of free electrons and @xmath48 the proton mass . to complete the set of equations ( [ hydromass ] ) ( [ hydroenergy ] )
, we must also specify the equation of state appropriate for an ideal gas @xmath49 where @xmath50 is the ratio of specific heats for the baryonic matter , @xmath51 is boltzmann s constant , @xmath38 is the gas temperature , and @xmath52 are the number densities for each of the different species .
we also need to provide the necessary cooling @xmath23 and heating @xmath24 functions to the right - hand - side of equation ( [ hydroenergy ] ) @xmath53 where @xmath54 is the compton cooling ( or heating ) due to interactions of free electrons with the cosmic microwave background radiation , and @xmath55 are the cooling rates from two - body interactions between species @xmath56 and @xmath57 .
the @xmath58 are integrals due to photoionizing and photodissociating heating @xmath59 we include a total of fourteen processes in the cooling function and three processes for heating . the physical mechanisms and mathematical expressions for each process are given in appendix b.
it is well known that the differential equations describing non - equilibrium atomic and molecular rate reactions can exhibit variations on extreme time scales .
characteristic creation and destruction scales can differ by many orders of magnitude among the different species and reactions . as a result , explicit schemes for integration can be unstable unless unreasonably small time steps ( smaller than the shortest dynamical times in the reaction flow ) are taken , which makes any multi - dimensional computation prohibitively expensive . for this reason implicit methods are preferred for stiff sets of equations .
these methods generally involve a newton s iterative procedure to achieve convergence , and for large dimensional jacobian matrices these implicit methods can also be very time consuming .
a number of packaged routines exist which are based on identifying the disparity in time scales among the species and switching between stiff and nonstiff solvers .
an example of such a package is the livermore solver for ordinary differential equations with automatic method switching for stiff and nonstiff problems lsodar ( hindmarsh 1983 ; petzold 1983 ) .
however , an implementation of this solver in multi - dimensions is extremely costly in computer time and an alternative numerical scheme is desirable for fully three - dimensional calculations where computational speed is crucial .
we use an operator and directional splitting of the hydrodynamic equations ( [ hydromass ] ) ( [ hydroenergy ] ) and ( [ hydrospecies ] ) to update the fourteen state variables @xmath17 , @xmath60 , @xmath19 and @xmath35 .
six basic steps are utilized .
first , the source step accelerates the fluid velocity due to pressure gradients and gravity , and modifies the velocity and energy equations to account for artificial viscosity @xmath61 where @xmath62 is a second rank tensor representing the artificial viscous stresses ( stone & norman 1992 ) .
a staggered mesh scheme is utilized whereby the scalar variables @xmath17 , @xmath35 , @xmath19 , @xmath21 and the artificial viscosity are zone centered , while the velocities are located at the zone interfaces . the pressure ,
potential , and viscosity gradients are thus naturally aligned with the momentum terms . in the second cooling / heating step ,
the energy changes are computed from `` pdv '' work and radiative cooling and heating from microphysical processes @xmath63 we discuss solving this equation further in the following subsection .
the third expansion step updates all state variables from the terms arising from the expansion of the universe @xmath64 the homogeneous nature of the expansion allows a simple solution @xmath65 , although a more generalized procedure is required for the energy equation in which an effective adiabatic index must be defined to eliminate the pressure term in the case of ionized gas ( anninos & norman 1994 ) .
the fourth or transport step solves the advection terms @xmath66 several different monotonic schemes have been implemented , including donor cell , van leer , and piecewise parabolic advection .
all results presented here use the second order van leer method which has the best accuracy - to - efficiency performance .
a fifth step evolves the densities of the separate species according to the chemistry of the collisional and radiative kinetic equations @xmath67 the methods for solving these equations is the focus of the next subsection .
the total baryonic density @xmath17 and the density of each individual species @xmath35 are updated independently in the expansion , transport and chemistry steps .
we are thus able to monitor the accuracy of our methods by evaluating the constraint equations for hydrogen , helium and charge conservation , equations ( [ conserveh ] ) ( [ conservene ] ) .
over the course of a typical 3d calculation of approximately one - thousand time steps , we find _ maximum _ errors , which are mostly concentrated at the shock fronts , to be of order 10 to 30% .
however , for increased stability and accuracy , we introduce a sixth step in our scheme to enforce the constraint equations at every time step .
an important point that one must consider in taking this approach is that the errors can be larger than the concentration of those species that are depleted .
it is therefore necessary to modify only the concentrations of the dominant species . for a hot gas in which @xmath0 and @xmath2 are mostly ionized
, the hydrogen and helium constraints should be solved for the ionized components @xmath1 and @xmath4 . in cold ( @xmath68 )
gas the neutral hydrogen and neutral helium concentrations must be adjusted to satisfy the constraints . in this way
, we are guaranteed to make only small ( typically much less than one percent ) fractional changes to any of the species at each timestep .
more specific details of the numerical methods used in the hydrodynamic updates ( [ split1 ] ) ( [ split4 ] ) can be found elsewhere ( stone & norman 1992 , anninos et al . 1994 ) .
here we emphasize the solution to the chemistry step ( [ split5 ] ) .
notice that equations ( [ split5 ] ) can be written schematically as @xmath69 where @xmath70 and @xmath71 is the atomic mass number of the @xmath72th element and @xmath48 is the proton mass .
the @xmath73 are the collective source terms responsible for the creation of the @xmath72th species .
the second terms involving @xmath74 represent the destruction mechanisms for the @xmath72th species and are thus proportional to @xmath52 .
equation ( [ model ] ) suggests a backward difference formula ( bdf ) can be used in which all source terms are evaluated at the advanced time step .
discretization of ( [ model ] ) yields @xmath75 lower order backward differentiation methods when applied to problems of the form @xmath76 are stiffly stable .
this rather restrictive stability property is highly desirable when solving sets of stiff equations ( oran & boris 1987 ) .
we have tried other less stable methods including higher order multi - step predictor - corrector schemes , various runge - kutta and adams - bashforth algorithms , and a newton s procedure to solve the backwards differenced linearized equations .
all of these alternative schemes have either proven to be unstable , less accurate or more expensive computationally compared to the simple bdf method .
the solver can be optimized further by noting that the intermediaries @xmath5 and @xmath6 in the molecular hydrogen production processes have large rate coefficients and low concentrations .
they are thus very sensitive to small changes in the more abundant species .
on the other hand , the low concentrations of @xmath5 and @xmath6 implies that they do not significantly influence the more abundant electron , hydrogen and helium concentrations .
this suggests that the nine species can be grouped into two categories : fast and slow reacting .
the fast reacting group , comprised of @xmath5 and @xmath6 , can be decoupled from the slower network and treated independently since the kinetic time scales for these species are much shorter than the characteristic times of the other seven species and the cosmological or gravitational times . @xmath5 and @xmath6
can thus be considered in equilibrium at all times , independent of the hydrodynamic state variables .
the expressions for the equilibrium abundances of @xmath5 and @xmath6 can be reduced by recognizing that reaction ( 19 ) , according to appendix a , can be neglected as a small order correction to @xmath5 , due to the low concentrations of both species @xmath5 and @xmath6 . neglecting reaction ( 19 )
, the equilibrium abundance of @xmath5 can be written independent of @xmath6 @xmath77 where the variables @xmath78 are the rate coefficients with subscripts referring to the reaction number in appendix a. then given @xmath79 , the equilibrium abundance of @xmath6 can be written with no additional assumptions as @xmath80 the separation into fast and slow reacting systems helps to further increase the accuracy and stability of the bdf method when applied only to the slower network over the longer characteristic time scales required by the hubble , hydrodynamic courant , and gravitational free - fall times in cosmological simulations . due to the intrinsic nonlinearity of equation ( [ model ] ) , not all source terms can be evaluated at the advanced time levels .
significant errors ( of order 20% as measured by the final fractional abundance of hydrogen molecules ) can be introduced if the source terms @xmath73 and @xmath74 are evaluated at the current time level at which the species data is known .
improvements to this crude approximation can be made by sequentially updating each species in order , rather than updating all species simultaneously from the data at the past time step .
for example , the order in which we solve the rate equations was determined finally through experimentation to be @xmath0 , @xmath1 , @xmath2 , @xmath3 , @xmath4 and @xmath8 , followed by the algebraic equilibrium expressions for @xmath5 and @xmath6 , then in the end updating @xmath7 also using the bdf scheme ( [ bdf ] ) .
the updated concentrations of the @xmath72th ( and previous ) species are used as source terms in the equation for @xmath81 .
further improvements in accuracy and stability can be made by mimicking more closely a fully bdf scheme by subcycling the rate solve over a single hydrodynamic courant time step .
the subcycle time steps are determined so that the maximum fractional change in the electron concentration is limited to 10% per timestep , ie .
@xmath82 with @xmath83 .
we note that this same subcylcing procedure can be used to update the energy in the cooling / heating step in equation ( [ split2 ] ) .
the equation of state ( [ eqos ] ) couples the energy and pressure through the gas temperature , and although we have tried a newton - raphson iterative procedure , we found it to converge very slowly or sometimes not at all because the cooling / heating rates are strongly nonlinear and non monotonic functions of temperature ( anninos & norman 1994 ) .
for this reason we solve equation ( [ split2 ] ) with an explicit method that subcycles the cooling source terms .
the timesteps for each subcyle are determined as @xmath84 , where @xmath83 as in the rate equation solve .
this algorithm has been tested to be both fast and accurate .
the robustness of our methods has been verified by switching the order in which the rate equations are solved relative to the other updates .
we have also experimented with the sensitivity to the temperature ( ie .
time centered , retarded and advanced temperatures ) used in updating the rate equations .
the results are stable and unchanged under all these sorts of permutations .
the numerical scheme described in section [ sec : numerical ] has been implemented in two separate codes : zeus-2d and hercules .
zeus-2d is a two dimensional eulerian hydrodynamics code originally developed by stone & norman ( 1992 ) and modified for cosmology by anninos & norman ( 1994 ) .
hercules is a three - dimensional hydrodynamics code derived from a 3d version of zeus , but modified for cosmology and generalized to include a hierarchical system of nested grids by anninos et al .
extensive tests of the two codes can be found in the references provided above . in this paper , we present tests only for the new additions to the codes , namely the multi
species chemistry and the non equilibrium cooling . due to the intrinsic complexity of such systems ,
the diversity of tests is rather limited .
we consider two crucial and relevant ( for cosmology ) tests : radiative shock waves and cosmological sheets .
in addition we have also developed an independent method of solution that is based on the well - tested packaged solver lsodar .
fully three - dimensional calculations of large scale structure formation are presented , comparing results from our method to that of lsodar .
chevalier & imamura ( 1982 ) and imamura , wolff & durisen ( 1984 ) have demonstrated through both analytic and numerical calculations that the fundamental mode of oscillations in one dimensional radiative shocks with cooling rates @xmath85 are unstable if @xmath86 . because the cooling rate is determined from the density of electrons and ions , a comparison to their published work provides an excellent test of both our cooling algorithm and the reaction network .
the initial data for this test is characteristic of pre - shock flows expected from the collapse of zeldovich pancakes @xmath87 corresponding to a uniform flow of gas along the @xmath88 direction .
reflection boundary conditions are imposed at @xmath89 and we use 100 zones to resolve a spatial extent of @xmath90 .
we assume only bremsstrahlung cooling of the form in appendix b , which has an exponent of @xmath91 .
a shock wave forms at the wall ( @xmath89 ) and propagates outward at a velocity @xmath92 . as the heated gas cools
, the shock begins to lose pressure support and slows down . because the cooling rate is proportional to @xmath93 ( ie .
@xmath94 ) , the cooled gas experiences an accelerated energy loss as it gets denser .
eventually the higher density gas nearest the wall loses pressure support and the shock collapses and re - establishes a new pressure equilibrium closer to the wall .
the shock front then begins again to propagate away from the wall , repeating the cycle of oscillations .
this behavior is shown in figure [ fig : radshk ] where the shock position @xmath95 is plotted as a function of time .
our numerical results indicate that the fundamental mode is indeed stable and damped , consistent with the analytic results of chevalier and imamura ( 1982 ) and the numerical simulations of imamura et al .
( 1984 ) .
the shock jump conditions @xmath96 provide a more quantitative check of our numerical results .
for the choice of initial data ( [ initial_data ] ) , the jump conditions give @xmath97 and @xmath98 , in excellent agreement with our numerical results @xmath97 and @xmath99 .
we can estimate the maximum distance the shock front will travel as @xmath100 where @xmath101 is the cooling time and @xmath102 is the shock speed .
substituting the bremsstrahlung cooling formula at the temperature predicted by the jump conditions and assuming a fully ionized gas with @xmath103 and hydrogen mass fraction @xmath104 , gives @xmath105 .
this is again consistent with our numerical result @xmath106 .
chevalier and imamura also characterize their linearized analytic solutions by the frequency of oscillations @xmath107 in units of @xmath108 where @xmath109 is the average shock front position . defining the period @xmath110 of perturbations as the time for the shock to first collapse back to the wall
, we find @xmath111 and @xmath112 , again in good agreement .
we use the zeldovich ( 1970 ) solution to set up a linearized single mode perturbation for the collapse of gas in one dimension @xmath113^{-1},\label{pertd}\end{aligned}\ ] ] where @xmath114 and @xmath115 are the comoving positions and velocities , @xmath116 the proper density , @xmath117 the unperturbed coordinate , @xmath118 , @xmath119 the comoving wavelength of perturbation , @xmath34 the present hubble constant , and @xmath120 the redshift corresponding to the collapse time .
parameters for the calculations presented here are the following @xmath121 at the higher temperatures ( @xmath122 ) characteristic of shock heated gas in high velocity pancake structures , the kinetic time scales are extremely short compared to the hubble time . neglecting the photoionization processes ,
collisional ionization equilibrium is then a good approximation to the following pairs of reactions @xmath123 for a fully ionized gas , we have @xmath124 where @xmath125 is the mass fraction of hydrogen . the corresponding equilibrium fractional abundances from reactions ( [ eq_first ] ) ( [ eq_last ] )
can then be written as @xmath126 which are functions only of the gas temperature . in figure
[ fig : pancake1 ] we show the mass fractions @xmath127 throughout the pancake structure at redshift @xmath128 . at this time
, the shock front is located at a distance of approximately @xmath129 from the central plane at @xmath89 , and is propagating outward at an average comoving velocity of about 110 @xmath130 ( 355 @xmath131 relative to the infalling gas ) .
two distinct cooling layers form , as evidenced by the thick solid line representing the gas temperature .
the gas between @xmath132 first cools mostly by atomic processes to a temperature of about @xmath9 . the second colder layer at @xmath133 results from cooling by hydrogen molecules which form from the residual electrons leftover from the nonequilibrium cooling through the first plateau at @xmath9 .
( we refer the reader to anninos & norman ( 1996 ) for further details concerning the chemistry , dynamics and radiative cooling of cosmological sheets . ) although nonequilibrium effects are important as the gas cools through @xmath9 , collisional ionization equilibrium is a good approximation to the species concentrations in the hot gas between @xmath134 .
the different symbols plotted across this region in figure [ fig : pancake1 ] represent the equilibrium abundances given by equations ( [ eq1 ] ) ( [ eq5 ] ) . differences between the numerical results and the equilibrium estimates are less than 0.5% for all species except @xmath7 which differ by roughly 5% .
however , the larger discrepancy in the @xmath7 mass fraction is not due to numerical errors , but to the inadequacy of equation ( [ eq5 ] ) to fully describe the kinetics .
a more accurate equilibrium ratio is derived by considering all the reactions involving @xmath7 @xmath135 equation ( [ eq6 ] ) agrees with our numerical results to within about 0.1% .
we have also run this same problem using the lsodar routines in place of the bdf method to solve for all nine species in full non - equilibrium .
a comparison of the two results at redshift @xmath128 is shown in figure ( [ fig : pancake2 ] ) .
the symbols represent the bdf calculation and the various line types are the lsodar results for the different species .
the two methods agree to within about 0.5% throughout all the different pancake layers , hot and cold .
notice also the excellent agreement in the mass fractions of @xmath5 and @xmath6 , which is further justification of the hybrid model ( in which the fast reacting species are singled out over cosmological dynamical times to be in equilibrium ) .
finally we point out that the fractional abundance of hydrogen molecules that form in the central cooled gas is consistent with the steady state shock calculations of shapiro & kang ( 1987 ) .
a final test is presented for an actual three dimensional cosmological calculation in which we compare our method of solving the rate equations to that of lsodar .
both accuracy and computational efficiency are stressed in this comparison .
the simulation is performed for a flat ( @xmath136 ) model universe with baryon mass fraction @xmath137 and hubble constant @xmath138 with @xmath139 .
the baryonic matter is composed of hydrogen and helium in cosmic abundance with a hydrogen mass fraction @xmath140 . the initial data is the harrison - zeldovich power spectrum modulated with a transfer function appropriate for cold dark matter ( cdm ) adiabatic fluctuations and normalized to a bias factor of @xmath141 .
we begin the simulations at redshift @xmath142 and evolve to the present time at @xmath143 .
the computational box size is set to @xmath144 ( comoving ) resolved by @xmath145 cells .
our calculation thus has a spatial grid resolution of @xmath146 and baryonic mass resolution of @xmath147 which is just marginally adequate to resolve cooling flows and the formation of hydrogen molecules .
the computational demands of lsodar implemented in three dimensions prohibits comparative calculations of much higher resolution .
we have run two separate simulations with identical initial conditions and model parameters .
the only difference in the two runs is the rate equation solver : in the one case we use our bdf method , in the other lsodar .
the results of comparison are shown in figures [ fig : hist ] to [ fig : cont2 ] .
figure [ fig : hist ] shows the cell distributions ( defined by counting up the number of cells at a particular binned range of values ) for three key variables that would be particularly sensitive to errors in the rate solver : ( a ) gas temperature which is constructed from the concentrations of all the species combined , and ( b ) molecular hydrogen mass fraction @xmath148 and electron number density fraction @xmath149 .
all results are shown at redshift @xmath150 .
the prominent peaks in the @xmath7 and electron fraction distributions correspond closely to values initialized at the start of the runs .
because most of the box volume is comprised of cosmic voids that do not undergo shock heating , the molecular hydrogen and electron fraction do not change significantly in most of the cells .
the relative @xmath151 deviations between the bdf and lsodar results are 0.018 , 0.039 and 0.13 for the gas temperature , molecular hydrogen and electron fractions , respectively .
we note , however , that deviations are due in part to the final output redshifts not being exactly the same in the two runs ; the relative difference is about 0.01 .
the larger differences found here ( as compared to the cosmological sheet calculations ) can also be attributed to the coarse grid resolution and the resulting large courant time step in the hydrodynamical calculation .
the agreements improve with resolution .
figure [ fig : cont ] shows contour graphs of the fractional volume of those cells with a particular combination of temperature and baryonic density , electron number density fraction and molecular hydrogen mass fraction .
two sets of graphs are presented : the three plots in the first column are results from the bdf method , the second column are the lsodar results .
we note that the sharp boundaries in the electron contours at the level @xmath152 correspond to the initialized fraction at @xmath142 : @xmath153 ( peebles 1993 ) .
the volume weighted distribution is concentrated at the initial value for @xmath7 since there is no mechanism to efficiently create nor destroy molecules at the low temperatures of the voids , and slightly lower than the initial value for the electron fraction since the expansion ( cosmological and gravitational ) of the voids continues to cool and recombine the gas .
contours of the spatial distribution of @xmath149 and @xmath148 are shown in figure [ fig : cont2 ] .
the data is projected ( and averaged ) along the @xmath154-axis at redshift @xmath155 .
the first row is the electron fraction , the second molecular hydrogen .
the first column are the bdf results , the second lsodar .
notice that hydrogen molecules form preferentially within the high dense filamentary structures but mostly in the knot like intersections of the filaments .
these are the highest density regions where the gas cools rapidly and the electrons are depleted by recombination .
peaks in the @xmath7 concentration therefore correspond to valleys in the electron distribution . in comparing results from the two solvers , bdf versus lsodar
, we see the distributions in both figures [ fig : cont ] and [ fig : cont2 ] are basically the same .
in addition to the accuracy of the bdf method , another very important point that should be stressed is the amount of computational time required to solve the rate equations .
the bdf run takes about 1.2 cpu hours with the full reaction network on the ncsa convex c3880 , in contrast to the equivalent lsodar calculation which takes about 16.7 hours to complete on the same machine .
the speedup of the bdf method over lsodar is roughly a factor of 14 .
we have developed and tested a new scheme to solve a system of stiff kinetic equations appropriate for chemical reaction flows in cosmological structure formation .
twenty eight chemical reactions for collisional and radiative processes are included in our model , which tracks nine separate atomic and molecular species : @xmath0 , @xmath1 , @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 , @xmath7 , and @xmath8 .
the reaction network is solved in a self - consistent manner with the hydrodynamic , n - body and cosmological expansion equations , and the accuracy of the solver has been verified by performing a series of test - bed calculations that includes radiative shock waves , cosmological sheets and monitering the conservation constraints . we have also implemented a publicly available and well - tested solver called lsodar in place of our scheme and made direct comparisons of the different results in one , two and three dimensions .
we find our methods are both fast and accurate , making fully three - dimensional calculations of non - equilibrium cosmological reaction flows feasible .
we have incorporated the species solver into two separate cosmological hydrodynamic codes : a two dimensional ratioed grid code to model zeldovich pancakes , and a more general three dimensional nested grid cosmological code hercules .
applications to date include investigations of star and galaxy formation in cosmological sheets ( anninos & norman 1996 ) , primordial star formation in cdm models ( zhang et al .
1996 ) , and simulations of the ly@xmath156 forest ( zhang , anninos & norman 1995 ; charlton et al .
1996 ) . in the future
, we plan to extend this work and develop a more sophisticated treatment of radiation to account for self shielding and to more accurately model the microphysics of optically thick gas .
this work is done under the auspices of the grand challenge cosmology consortium ( gc@xmath157 ) and supported in part by nsf grant asc-9318185 .
the calculations were performed on both the c90 at the pittsburgh supercomputing center and the convex c3880 at the national center for supercomputing applications at the university of illinois .
the following is a list of all chemical reactions that we include in our calculations .
further discussions and justification of the completeness of this set of reactions can be found in abel et al .
there we also include explicit formulae for the different rate coefficients .
= ( 16 ) = @xmath4 = + @xmath158 = @xmath159 @xmath4 = + @xmath158 = + ( 1 ) @xmath0 + @xmath160 @xmath159 @xmath1 + 2@xmath160 + ( 2 ) @xmath1 + @xmath160 @xmath159 @xmath0 + @xmath161 + ( 3 ) @xmath2 + @xmath160 @xmath159 @xmath162 + 2@xmath160 + ( 4 ) @xmath3 + @xmath160 @xmath159 @xmath2 + @xmath161 + ( 5 ) @xmath3 + @xmath160 @xmath159 @xmath4 + 2@xmath160 + ( 6 ) @xmath4 + @xmath160 @xmath159 @xmath3 + @xmath161 + + ( 7 ) @xmath0 + @xmath160 @xmath159 @xmath5 + @xmath161 + ( 8) @xmath5 + @xmath0 @xmath159 @xmath7 + @xmath160 + ( 9 ) @xmath0 + @xmath1 @xmath159 @xmath6 + @xmath161 + ( 10 ) @xmath6 + @xmath0 @xmath159 @xmath7 + @xmath1 + ( 11 ) @xmath7 + @xmath1 @xmath159 @xmath6 + @xmath0 + ( 12 ) @xmath7 + @xmath160 @xmath159 @xmath163 + @xmath160 + ( 13 ) @xmath7 + @xmath164 @xmath159 @xmath165 + ( 14 ) @xmath5 + @xmath160 @xmath159 @xmath0 + 2@xmath160 + ( 15 ) @xmath5 + @xmath0 @xmath159 @xmath163 + @xmath160 + ( 16 ) @xmath5 + @xmath1 @xmath159 @xmath163 + ( 17 ) @xmath5 + @xmath1 @xmath159 @xmath6 + @xmath160 + ( 18 ) @xmath166 + @xmath160 @xmath159 @xmath163 + ( 19 ) @xmath166 + @xmath5 @xmath159 @xmath7 + @xmath0 + + + ( 20 ) @xmath0 + @xmath161 @xmath159 @xmath1 + @xmath160 + ( 21 ) @xmath2 + @xmath161 @xmath159 @xmath3 + @xmath160 + ( 22 ) @xmath3 + @xmath161 @xmath159 @xmath4 + @xmath160 + ( 23 ) @xmath5 + @xmath161 @xmath159 @xmath0 + @xmath160 + ( 24 ) @xmath7 + @xmath161 @xmath159 @xmath166 + @xmath160 + ( 25 ) @xmath6 + @xmath161 @xmath159 @xmath0 + @xmath1 + ( 26 ) @xmath6 + @xmath161 @xmath159 @xmath167 + @xmath160 + ( 27 ) @xmath7 + @xmath161 @xmath159 @xmath168 @xmath159 @xmath163 + ( 28 ) @xmath7 + @xmath161 @xmath159 @xmath163 + the cooling rates @xmath23 and photoionization cross sections @xmath43 included in our calculations .
we use units of @xmath169 for the rates , @xmath170 for the cross sections , and degrees kelvin for temperature @xmath38 . also , @xmath171 , @xmath172 with @xmath173 being the threshold frequencies of the @xmath72th species , and @xmath174 , @xmath175 and @xmath176 are the rate coefficients for the ionizing chemical reactions ( 1 ) , ( 3 ) and ( 5 ) listed in appendix a. = @xmath177 ( black 1981 ; cen 1992 ) : + @xmath178 + @xmath179 + @xmath180 + + * collisional ionization cooling * ( shapiro & kang 1987 ; cen 1992 ) : + @xmath181 + @xmath182 + @xmath183 + @xmath184 + + * recombination cooling * ( black 1981 ; spitzer 1978 ) : + @xmath185 + @xmath186 + @xmath187~ \exp{(-470000/t)}~n_e n_{he^+}$ ] + @xmath188 + + * molecular hydrogen cooling * ( lepp & shull 1983 ) .
+ + * bremsstrahlung cooling * ( black 1981 ; spitzer & hart 1979 ) : + @xmath189 ~n_e ( n_{h^+ } + n_{he^+ } + n_{he^{++}})$ ] + + * compton cooling or heating * ( peebles 1971 ) : + @xmath190~n_e$ ] + + * photoionization heating crossections * ( osterbrock 1974 ) : + @xmath191~ [ 1-\exp{(-2\pi/\alpha_h)}]^{-1}$ ] + @xmath192 $ ] + @xmath193~ [ 1-\exp{(-2\pi/\alpha_{he^+})}]^{-1}$ ] + | we have developed a method of solving for multi - species chemical reaction flows in non equilibrium and self consistently with the hydrodynamic equations in an expanding flrw universe .
the method is based on a backward differencing scheme for the required stability when solving stiff sets of equations and is designed to be efficient for three - dimensional calculations without sacrificing accuracy . in all , 28 kinetic reactions are solved including both collisional and radiative processes for the following nine separate species : @xmath0 , @xmath1 , @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 , @xmath7 , and @xmath8 .
the method identifies those reactions ( involving @xmath5 and @xmath6 ) ocurring on the shortest time scales , decoupling them from the rest of the network and imposing equilibrium concentrations to good accuracy over typical cosmological dynamical times .
several tests of our code are presented , including radiative shock waves , cosmological sheets , conservation constraints , and fully three - dimensional simulations of cdm cosmological evolutions in which we compare our method to results obtained when the packaged routine lsodar is substituted for our algorithms . |
the time - domain matsubara correlator of the coupling operator @xmath7 is defined as @xmath12\right\rangle,\label{eq : csxtau}\ ] ] where @xmath15 is the operator at imaginary time @xmath13 with @xmath136 , @xmath14 is an imaginary time , and @xmath16 is the time ordering operator .
we let @xmath11 for simplicity of discussion . with translational symmetry
, this correlator can be simplified as @xmath21 by setting @xmath22 and @xmath23 .
the momentum - domain counterpart of the correlator @xmath21 can be obtained by fourier transformation and can be denoted as @xmath137 .
note that when the operator @xmath7 is bosonic , @xmath138 for @xmath139 ; when the operator @xmath7 is fermionic , @xmath140 for @xmath139 @xcite .
these relations are also valid for the spatial - domain correlator @xmath21 .
the frequency - domain counterpart of the time - domain correlator @xmath137 has the form : @xmath141 where @xmath25 is the quasimomentum and @xmath26 [ @xmath27 for bosonic ( fermionic ) operator @xmath7 with @xmath28 being an integer .
the inverse transformation of ( [ eq : cspomega ] ) can be written as @xmath142 the correlator measured in experiments is the retarded correlator defined as @xmath143\right\rangle,\label{eq : csretxt}\ ] ] where @xmath144 is the heaviside function , @xmath145 , and @xmath146 , @xmath147 are real times .
here we have assumed that the coupling operator @xmath148 is bosonic . for ferminonic operators ,
the retarded correlator is defined as @xmath149\rangle$ ] .
the retarded correlator in the momentum and frequency domains can be denoted as @xmath150 .
the retarded correlator can be obtained from the matsubara correlator via analytic continuation : @xmath151 with @xmath152 being an infinitesimally small positive number @xcite .
because this relation is exact , the matsubara correlators obtained in the main text can be directly converted to corresponding retarded correlators .
our result in the main text can hence be directly verified in experiments .
note that the conversion of a numerical matsubara correlator to a retarded correlator via analytic continuation is nontrivial .
various techniques for such conversion have been developed @xcite .
note that the imaginary - time and real - time correlators of the coupling operator of the detector and their frequency counterparts can be defined similarly and will be discussed in the next section .
the correlator of the detector is defined as @xmath36\right\rangle \label{eq : drxtau}\ ] ] with @xmath37 , where @xmath5 is the total hamiltonian of this system , @xmath2 ( @xmath3 ) is the hamiltonian for the simulator ( detector ) , and @xmath4 is the simulator - detector coupling . for a system with translational symmetry ,
the correlator can be simplified as @xmath38 .
written explicitly , we have @xmath153\right]/z \label{eq : drxtaufull}\ ] ] with @xmath43 $ ] .
define the evolution operator @xmath44 in the interaction picture of the unperturbed hamiltonian @xmath47 .
it can be shown that for @xmath154 @xmath155/\textrm{tr}\left[e^{-\beta h_{0 } } u(\beta , 0)\right ] \nonumber \\ & = & -\textrm{tr}\left[e^{-\beta h_{0 } } u(\beta , \tau ) \hat{\gamma}_{ri}(\tau ) u(\tau , 0 ) \hat{\gamma}_{ri}(0)\right]/\textrm{tr}\left[e^{-\beta h_{0 } } u(\beta , 0)\right ] \nonumber \\ & = & -\left\langle \hat{t}_{\tau}\left[u(\beta , 0)\hat{\gamma}_{ri}(\tau)\hat{\gamma}_{ri}(0)\right]\right\rangle _ { 0}/\left\langle u(\beta , 0)\right\rangle _ { 0},\label{eq : dr}\end{aligned}\ ] ] where @xmath51 is the coupling operator @xmath8 in the interaction picture .
the operator averages in the third row of the above equation are taken over the thermal equilibrium of the hamiltonian @xmath52 with @xmath53/\text{tr}[e^{-\beta h_{0}}]$ ] . in deriving ( [ eq : dr ] ) , we have used the relations @xmath156 , @xmath157 , and @xmath158 .
let @xmath49 be the simulator - detector coupling in the interaction picture .
it can be shown that @xmath159 which leads to @xmath160 .
\label{eq : ubeta}\ ] ] treating @xmath161 as a perturbation , we can expand the operator @xmath54 as @xmath162 + \cdots.\label{eq : ubeta2}\end{aligned}\ ] ] substituting ( [ eq : ubeta2 ] ) into ( [ eq : dr ] ) , we obtain the taylor expansion of the detector correlator @xmath40 in terms of the simulator - detector coupling @xcite .
similar derivation can be applied to the correlator of the simulator .
for a bare cavity mode ( without the bath modes and the simulator ) , the time - domain correlator ( [ eq : drxtaufull ] ) of the coupling operator @xmath63 can be derived as @xmath163 -\theta(-\tau)\left[e^{-\omega_{d}\tau}n_{\textrm{th}}+e^{\omega_{d}\tau}\left(n_{\textrm{th}}+1\right)\right],\label{eq : drc}\ ] ] where @xmath164 is the thermal occupation number of the cavity mode at @xmath18 ( temperature @xmath19 ) . using the transformation @xmath165
, we find the frequency - domain correlator : @xmath166,\label{eq : dr0}\ ] ] with @xmath26 and @xmath28 being an integer .
as discussed in the main paper , the interaction between the cavity and its bath modes modifies the correlator of the cavity .
we denote the correlator of the cavity mode affected by the bath modes as @xmath79 . the diagrammatic equation generated by
( [ eq : dr ] ) shows that @xmath167 all correlators in the above equation have the variable @xmath113 , which is omitted for brevity of discussion . substituting the expressions of @xmath77 and @xmath78 into ( [ eq : drb ] )
, we have @xmath168 with the substitution @xmath169 in the bath correlators , we have @xmath170 the first term on the right hand side of the above equation is a summation of cauchy principle values , and we denote this term as @xmath171 .
this term generates a small shift of the cavity resonance to become @xmath81 with @xmath82 . using the definition of the noise spectrum @xmath67 @xcite and the assumption @xmath68
, we derive that @xmath172 finally , the detector correlator becomes @xmath173 which corresponds to a cavity mode with frequency @xmath174 and bandwidth @xmath84 . | an important step in quantum simulation is to measure the many - body correlations of the simulated model . for a practical quantum simulator composed of finite number of qubits and cavities , in contrast to ideal many - body systems in the thermodynamic limit
, a measurement device can generate strong backaction on the simulator , which could prevent the accurate readout of the correlation functions . here
we calculate the readout of a detector coupled to an analog quantum simulator .
we show that reliable characterization of the many - body correlations in the simulator can be achieved when the coupling operators obey the wick s theorem .
our results are illustrated with two examples : a simulator for an harmonic oscillator and a simulator for the free electron gas . _
introduction._solving quantum many - body problems with classical methods often requires formidable computational resources .
quantum simulators , artificial systems constructed to mimic many - body models or dynamics , can solve such problems efficiently due to their built - in quantum parallelism @xcite . with the rapid experimental progresses , digital @xcite and analog @xcite quantum simulators
can be realized in various physical systems .
arrays of coupled qubits have been demonstrated with trapped ions @xcite and superconducting circuits @xcite . various equilibrium and out - of - equilibrium phenomena have been investigated with atoms and molecules in optical lattices @xcite .
quantum simulation has been applied to problems in condensed - matter physics , quantum chemistry , nonequilibrium dynamics , and _ etc
. _ @xcite .
the measurement of the temporal and spatial correlations in quantum simulators is a crucial component in analog quantum simulation @xcite .
these correlators are often hard to calculate in computationally - demanding problems and can only be extracted from the readout of a detector . for ideal many - body systems in the thermodynamic limit , these correlators can be accurately determined from the detector s readout , where the detector perturbs the system weakly and the detector s backaction on the system can be safely neglected @xcite . for quantum simulators composed of finite number of qubits and cavities
, the interaction between the simulator and the detector can distort the correlators of the simulator .
it thus raises the question of whether the correlators of the simulator can be reliably extracted from the detector s readout . on the one hand ,
the detector relies on the existence of the simulator - detector coupling to retrieve the imprint of the correlators . on the other hand ,
the coupling generates a backaction that could affect the correlators of the simulated model . in @xcite ,
the backaction of a cavity detector on a transverse field ising model with @xmath0 sites was estimated with a perturbative approach and shows a @xmath1-dependence .
previously , quantum non - demolition measurement on many - body hamiltonians was studied in atomic systems @xcite , which can only be applied to a limited group of simulators . here
we study the readout of a detector coupled to a quantum simulator with the matsubara diagrammatic technique for finite temperature systems .
the calculation is conducted on two examples : a simulator for a simple harmonic oscillator and a simulator for the free electron gas .
we show that the unperturbed matsubara correlators of the quantum simulator can be accurately extracted from the detector s readout when the coupling operators obey the wick s theorem @xcite .
whereas when this theorem is violated , the readout returns a correlator that is modified by the detector s backaction . in practice
, wick s theorem can be valid in models with gaussian fluctuations @xcite .
the imaginary - time matsubara correlators can be transformed into real - time retarded correlators that are directly associated with the measurement via analytic continuation @xcite .
our result can thus be verified in experiments and can lead to the design of reliable measurement schemes for quantum simulators .
we want to add that besides the detector s backaction , the reliability of analog quantum simulation can also be affected by noise and disorder @xcite .
several approaches to certifying quantum simulators under realistic conditions have been discussed , such as cross certification between simulators constructed with different systems and comparison with classical solutions in designated parameter regimes @xcite .
note that in our treatment , the system - bath coupling is weak with its only effect being to thermalize the system to equilibrium . the influences of strong simulator - bath coupling and structured spectrum have been studied in @xcite .
_ simulator and detector correlators._consider a quantum simulator with hamiltonian @xmath2 coupled to a detector with hamiltonian @xmath3 via the interaction @xmath4 .
the total hamiltonian of this system is @xmath5 .
the interaction can have the form @xmath6 , where @xmath7 ( @xmath8 ) is the coupling operator of the simulator ( detector ) at position @xmath9 and @xmath10 is the site - dependent coupling constant ( @xmath11 ) .
the coupling of the simulator ( detector ) to bath modes is implicitly included in the hamiltonians @xmath2 ( @xmath3 ) .
the property of a many - body system can be revealed by its correlation functions .
one matsubara correlator for the simulator is defined as @xmath12\right\rangle , \label{eq : csxtau}\ ] ] which describes the correlation between the coupling operators at imaginary times @xmath13 and @xmath14 with @xmath15 and @xmath16 being the time ordering operator .
written explicitly , @xmath17/z$ ] with @xmath18 at temperature @xmath19 and @xmath20 $ ] . for a system with translational symmetry , this correlator can be simplified to be @xmath21 by setting @xmath22 and @xmath23 .
the correlator @xmath21 can be transformed into the momentum and frequency domains as @xmath24 with quasimomentum @xmath25 and @xmath26 [ @xmath27 for integer number @xmath28 when the operator @xmath29 is bosonic [ fermionic ] @xcite .
the correlator @xmath30 is the full correlator including the effect of the simulator - detector coupling . without the coupling ( @xmath31 ) ,
the correlator can be denoted as @xmath32 , which implicitly includes the effect of the simulator s bath modes .
is denoted as @xmath32 [ @xmath33 , the correlator of the detector with the bath modes is @xmath34 , and the full correlator of the simulator [ detector ] is @xmath30 [ @xmath35.,width=302 ] the correlator of the detector is defined as @xmath36\right\rangle \label{eq : drxtau}\ ] ] with @xmath37 . under time
invariance , this correlator can be simplified to be @xmath38 , which transforms to @xmath39 in the frequency domain . here
@xmath40 is the full correlator that includes the effects of the bath modes and the simulator .
the coupling between the detector and the bath modes , e.g. , the dissipation of a cavity detector , can often be engineered and characterized in experiments .
the effect of the simulator - detector coupling on @xmath40 , when properly extracted from the readout , yields a measurement of the many - body correlations of the simulator .
denote the correlator of the bare detector as @xmath41 and the correlator of the detector coupled to the bath modes ( without the simulator ) as @xmath34 .
the relations of these three correlators are illustrated in fig .
[ fig1 ] . _
diagrammatic expansion._the correlator of the detector is affected by the simulator - detector coupling . using ( [ eq : drxtau ] ) , we have @xmath42/z$ ] with @xmath43 $ ] .
the matsubara diagrammatic technique can be applied to calculate this correlator @xcite .
let @xmath44 be the evolution operator from @xmath45 to @xmath46 in the interaction picture of hamiltonian @xmath47 .
it can be shown that @xmath48 $ ] with @xmath49 being the coupling @xmath4 in the interaction picture .
then , @xmath50\right\rangle _ { 0}/\left\langle u(\beta , 0)\right\rangle _ { 0 } \label{eq : dr}\ ] ] with @xmath51 . here
the operator averages are taken on the thermal equilibrium of hamiltonian @xmath52 with @xmath53/\text{tr}[e^{-\beta h_{0}}]$ ] .
all the terms related to the coupling @xmath4 in ( [ eq : dr ] ) are included in the evolution operator @xmath54 .
treating @xmath55 as a perturbation , @xmath54 can be expanded into taylor series of @xmath55 . with the diagrammatic technique
, the correlator @xmath38 can be written as a summation of diagrams in different orders of @xmath55 @xcite . _
cavity detector._consider a cavity mode serving as the detector , which can be realized with optical or microwave cavities @xcite and mechanical resonators @xcite .
the detector together with its bath modes has the hamiltonian @xmath56 , where @xmath57 ( @xmath58 ) is the annihilation ( creation ) operator , @xmath59 is the frequency of the cavity mode , @xmath60 $ ] describes the cavity - bath coupling with @xmath61 being the hamiltonian of the @xmath62th bath mode , @xmath63 is the coupling operator of the cavity , and @xmath64 ( @xmath65 ) is the annihilation ( creation ) operator of the @xmath62th bath mode with coupling coefficient @xmath66 .
the spectrum of the bath modes is @xmath67 @xcite .
we assume the spectrum to be flat in the frequency range of interest with @xmath68 and @xmath69 being the cavity bandwidth .
the detector and the bath modes are in equilibrium at temperature @xmath19 .
we now derive the correlator of the detector coupled to the bath modes with a flat noise spectrum .
this correlator will be utilized to calculate the detector s readout in the presence of a quantum simulator .
the matsubara correlator of the bare cavity ( not coupled to the bath ) can be derived from ( [ eq : drxtau ] ) .
in the frequency domain , the correlator has the form @xmath70 $ ] with @xmath26 and @xmath28 being an integer @xcite , where the temperature dependence enters the expression implicitly through the frequency @xmath71 .
similarly , the correlator of an unperturbed bath mode is @xmath72 $ ] with @xmath73 being the frequency of the @xmath62th bath mode . as illustrated in the diagrammatic equation in fig .
[ fig2 ] , in the presence of the bath modes , the correlator of the cavity satisfies the dyson equation @xmath74 where we omit the dependence on the variable @xmath75 in the correlators for brevity of discussion .
the correlator of the cavity can then be obtained as @xmath76^{-1}{\cal d}_{r0}$ ] . substituting @xmath77 and @xmath78 into @xmath79
, we find that @xmath80\label{eq : drbfinal}\ ] ] with @xmath81 being the cavity frequency shifted by a small cathy principle term @xmath82 @xcite .
the correlator ( [ eq : drbfinal ] ) describes a cavity mode with frequency @xmath83 and bandwidth @xmath84 . and @xmath79 of the cavity and @xmath78 of the @xmath62th bath mode .
( b ) the dyson equation of @xmath79 .
the circles correspond to the coupling coefficients between the cavity and the bath modes.,width=302 ] _ readout of oscillator mode._here the simulator is for a simple harmonic oscillator with hamiltonian @xmath85 , where @xmath86 is the frequency and @xmath87 ( @xmath88 ) is the annihilation ( creation ) operator of the oscillator .
the coupling between the oscillator and the cavity detector has the form @xmath89 with coupling operators @xmath90 and @xmath63 , and coupling constant @xmath91 . without the coupling ( @xmath92 ) ,
the correlator of the operator @xmath29 is @xmath93 $ ] ; the correlator of the operator @xmath8 with the bath modes is given by ( [ eq : drbfinal ] ) .
with the coupling , the oscillator generates a force on the cavity that is proportional to the quadrature @xmath29 , which can be measured in the cavity s readout .
the correlator of the cavity can be calculated with diagrammatic technique . as shown in fig .
[ fig3 ] , the dyson equation is @xmath94 which yields @xmath95 . with @xmath79 in ( [ eq : drbfinal ] ) , the full correlator of the cavity becomes @xmath96^{-1}{\cal d}_{r0}$ ] .
it can be viewed as a cavity coupled to a structured bath that contains the original bath modes and the oscillator . and @xmath40 of the cavity and @xmath97 of the harmonic oscillator .
( b ) the dyson equation of @xmath40 .
the circles correspond to the coupling coefficient between the cavity and the oscillator .
, width=302 ] the correlator of the unperturbed oscillator can then be extracted from the measurement of the cavity correlators @xmath79 and @xmath40 with @xmath98 this relation can be extended to a damped oscillator with finite bandwidth and can be used to characterize the parameters of the oscillator , such as its resonant frequency and bandwidth .
meanwhile , the simulator - detector coupling generates a backaction on the oscillator .
the full correlator of the oscillator can be derived as @xmath99 .
an interesting observation is that the readout of the cavity gives a faithful characterization of the unperturbed correlator @xmath97 , instead of the full correlator @xmath100 , even though the coupling generates a finite backaction on the oscillator .
this result indicates that finite detector backaction does not necessarily prevent the accurate characterization of a simulator .
_ readout of free electron gas._the next simulator is for a free electron gas with hamiltonian @xmath101 , where @xmath102 is the energy and @xmath103 ( @xmath104 ) is the annihilation ( creation ) operator of an electron with quasimomentum @xmath105 and spin @xmath106 .
this model describes electrons in normal metals , where the coulomb interaction between the electrons is largely screened by the interaction between the electrons and lattice ions .
assume that the interaction between the electron gas and the cavity detector has the form @xmath107 , where the coupling operator @xmath108 is the density of the electron gas at position @xmath109 @xcite .
this interaction corresponds to the coupling between the electromagnetic field of the cavity and the electrons at @xmath109 . without the coupling ( @xmath92 ) ,
the correlator of the operator @xmath110 can be written as @xmath111 , where @xmath112 is the propagator of a free electron with quasimomentum @xmath105 , imaginary frequency @xmath113 , and spin @xmath106 @xcite .
the coupling generates a backaction on the simulator , which is manifested by various diagrams .
some of these diagrams are irreducible diagrams that can not be disconnected by cutting off one propagator .
we denote the sum of such diagrams as @xmath114 [ fig .
[ fig4](a , b ) ] . for comparison ,
examples of reducible diagrams are given in fig .
[ fig4](c ) . and @xmath40 of the cavity , the propagator @xmath115 , the unperturbed correlator @xmath97 , and the sum of the irreducible diagrams @xmath114 of the free electron gas .
( b ) @xmath114 with lower - order diagrams .
( c ) examples of reducible diagrams .
( d ) the dyson equation of @xmath40 .
the circles correspond to the coupling coefficient between the cavity and the electrons.,width=302 ] the dyson equation of the full correlator of the cavity can be written as @xmath116 as shown by the diagrammatic equation in fig .
[ fig4](d ) .
we then have @xmath117 . with the measured correlators
@xmath79 and @xmath40 , we can obtain @xmath118 i.e. , the detector s readout returns a correlator modified by the detector s backaction .
_ wick s theorem._for the simple harmonic oscillator , the readout of the cavity gives the unperturbed correlator @xmath97 . whereas the measurement of the free electron gas yields the correlator @xmath114 that is modified by the cavity s backaction .
the distinct difference between these two cases can be explained by the validity of the wick s theorem for the coupling operators in these models @xcite .
for demonstration , assume that the wick s theorem is valid for the coupling operators @xmath29 and @xmath8 with @xmath119 where @xmath120 is the coupling operator in the interaction picture and we assume @xmath121 .
the correlators that contain @xmath122 operators in ( [ eq : wt ] ) will be further factorized using wick s theorem . in the end , each term in the full correlator becomes a product of @xmath28 two - point correlators in the form of @xmath123 .
similar factorization can be performed on the correlators of the operator @xmath8 . using ( [ eq : wt ] ) , the dyson equation of the detector correlator can be derived as @xmath124 .
the detector s readout can hence be used to characterize the unperturbed correlator of the simulator @xmath97 .
the wick s theorem is exactly valid when the coupling operator is a linear combination of the creation and annihilation operators of the quasiparticles in the unperturbed hamiltonian @xmath52 . for the harmonic oscillator ,
the coupling operator @xmath125 obeys the wick s theorem .
the density operator @xmath110 of the free - electron gas , on the contrary , does not obey this theorem , and its correlator includes complicated diagrams resulting from the simulator - detector coupling . in realistic systems , the wick s theorem can be valid for the coupling operators when the fluctuations in the system are approximately gaussian @xcite . under this approximation ,
the dyson equation of the detector correlator depends on the unperturbed correlator of the simulator @xmath97 .
this enables the reliable readout of the correlator @xmath97 by the measurement device , which is the goal of quantum simulation .
furthermore , when @xmath97 is known , the detector s readout can be used to verify the validity of the wick s theorem by comparing the measured correlator with @xmath97 . _
discussions._we have studied the readout of a cavity detector coupled to a quantum simulator at given spatial location for simplicity of discussion .
our result can be applied to other types of detectors and other forms of coupling operators .
for example , the measurement of spatial correlations is of wide interests in condensed - matter systems . by extending the coupling operator to include spatial dependence , i.e. , @xmath126 ,
the impact of the detector s backaction on the readout of the spatial correlators can be studied .
we can also study schemes that involve multiple detectors , each of which is coupled to the simulator via a distinct operator .
in addition , our calculation is based on the imaginary - time matsubara correlators . while the correlators measured in experiments are the retarded correlators defined in real time @xcite .
the retarded correlators can be obtained from the matsubara correlators via analytic continuation .
various techniques have been developed to convert the matsubara correlators to the retarded correlators numerically @xcite . _
conclusions._to summarize , we studied the question of whether the readout of a detector can faithfully reflect the unperturbed many - body correlations in a quantum simulator using the matsubara diagrammatic technique .
our result shows that when the coupling operators obey the wick s theorem , the readout of the detector carries the imprint of the unperturbed correlator of the simulator and can be used to accurately characterize the correlations in the simulator .
in contrast , when the wick s theorem is violated , the measured simulator correlator is modified by the detector s backaction .
our work can shed light on the design of reliable measurement of quantum simulators and can lead to future endeavors in the proper characterization of quantum simulators .
l.t .
is supported by the national science foundation under award no .
nsf - dmr-0956064 .
i.s . acknowledges financial support by friedrich - ebert - stiftung .
99 r. p. feynman , int .
j. theor .
phys .
* 21 * , 467 ( 1982 ) .
s. lloyd , science * 273 * , 1073 ( 1996 ) .
l. garca - lvarez , j. casanova , a. mezzacapo , i. l. egusquiza , l. lamata , g. romero , and e. solano , phys .
rev .
lett . * 114 * , 070502 ( 2015 ) .
see , e.g. , j. i. cirac and p. zoller , nat .
phys .
* 8 * , 264 ( 2012 ) ; and references there in .
p. hauke , f. m. cucchietti , l. tagliacozzo , i. deutsch , and m. lewenstein , rep .
. phys . * 75 * , 082401 ( 2012 ) .
k. kim , m .- s .
chang , s. korenblit , r. islam , e. e. edwards , j. k. freericks , g .- d .
lin , l .- m .
duan , and c. monroe , nature ( london ) * 465 * , 590 ( 2010 ) .
j. w. britton , b. c. sawyer , a. c. keith , c .- c .
joseph wang , j. k. freericks , h. uys , m. j. biercuk , and j. j. bollinger , nature ( london ) * 484 * , 489 ( 2012 ) .
r. blatt , and c. f. roos , nat .
phys .
* 8 * , 277 ( 2012 ) .
m. w. johnson , m. h. s. amin , s. gildert , t. lanting , f. hamze , n. dickson , r.harris , a. j. berkley , j. johansson , p. bunyk , e. m. chapple , c. enderud , j. p. hilton , k. karimi , e. ladizinsky , n. ladizinsky , t. oh , i. perminov , c. rich , m. c. thom , e. tolkacheva , c. j. s. truncik , s. uchaikin , j. wang , b. wilson , and g. rose , nature ( london ) * 473 * , 194 ( 2011 ) .
s. boixo , t. f. rnnow , s. v. isakov , z. wang , d. wecker , d. a. lidar , j. m. martinis , and m. troyer , nat . phys . * 10 * , 218 ( 2014 ) .
p. macha , g. oelsner , j .-
m .
reiner , m. marthaler , s. andr , g. schn , u. hbner , h .-
g .
meyer , e. ilichev and a. v. ustinov , nat .
commun . * 5 * , 5146 ( 2014 ) .
m. h. devoret and r. j. schoelkopf , science * 339 * , 1169 ( 2013 ) .
t. langen , r. geiger , and j. schmiedmayer , annu .
rev . condens .
matter phys .
* 6 * , 201 ( 2015 ) .
i. bloch , j. dalibard , and s. nascimbne , nat .
phys .
* 8 * , 267 ( 2012 ) .
d. s. abrams and s. lloyd , phys .
rev .
lett . * 79 * , 2586 ( 1997 ) .
l .- a .
wu , m. s. byrd , and d. a. lidar , phys .
rev .
lett .
* 89 * , 057904 ( 2002 ) .
a. aspuru - guzik , a. d. dutoi , p. j. love , and m. head- gordon , science * 309 * , 1704 ( 2005 ) .
s. mostame , p. rebentrost , a. eisfeld , a. j. kerman , d. i. tsomokos , and a. aspuru - guzik , new j. phys .
* 14 * , 105013 ( 2012 ) .
v. m. stojanovi , t. shi , c. bruder , j. i. cirac , phys .
rev .
lett . * 109 * , 250501 ( 2012 ) .
f. mei , v. m. stojanovi , i. siddiqi , and l. tian , phys .
rev .
b * 88 * , 224502 ( 2013 ) .
j .- m .
reiner , m. marthaler , j. braumller , m. weides , and g. schn , phys .
rev .
a * 94 * , 032338 ( 2016 ) .
p. j. j. omalley , r. babbush , i. d. kivlichan , j. romero , j. r. mcclean , r. barends , j. kelly , p. roushan , a. tranter , n. ding , b. campbell , y. chen , z. chen , b. chiaro , a. dunsworth , a. g. fowler , e. jeffrey , e. lucero , a. megrant , j. y. mutus , m. neeley , c. neill , c. quintana , d. sank , a. vainsencher , j. wenner , t. c. white , p. v. coveney , p. j. love , h. neven , a. aspuru - guzik , and j. m. martinis , phys .
rev .
x * 6 * , 031007 ( 2016 ) .
r. barends , a. shabani , l. lamata , j. kelly , a. mezzacapo , u. las heras , r. babbush , a. g. fowler , b. campbell , y. chen , z. chen , b. chiaro , a. dunsworth , e. jeffrey , e. lucero , a. megrant , j. y. mutus , m. neeley , c. neill , p. j. j. omalley , c. quintana , p. roushan , d. sank , a. vainsencher , j. wenner , t. c. white , e. solano , h. neven , and j. m. martinis , nature ( london ) * 534 * , 222 ( 2016 ) .
g. d. mahan , _ many - particle physics 3rd ed . _ ( kluwer academic / plenum publishers , new york , 2000 ) .
l. h. du , j. q. you , and l. tian , phys .
rev .
a * 92 * , 012330 ( 2015 ) ; here the perturbative correction to the hamiltonian of the transverse field ising model is given by eq .
( 28 ) and the order of magnitude of the correction is @xmath127 with @xmath128 being the magnitude of the total simulator - detector coupling and @xmath91 being the coupling strength between the detector and a single qubit .
g. m. bruun , b. m andersen , e. demler , and a. s srensen , phys .
rev .
lett .
* 102 * , 030401 ( 2009 ) .
k. eckert , o. romero - isart , m. rodriguez , m. lewenstein , e. s. polzik , and a. sanpera , nat . phys . * 4 * , 50 ( 2008 ) .
g. c. wick , phys . rev .
* 80 * , 268 ( 1950 ) .
y .- l .
wang , s. shtrikman , and h. callen , phys . rev . * 148 * , 419 ( 1966 ) . t. arai and b. goodman , phys
. rev . * 155 * , 514 ( 1967 ) .
p. schad , yu .
makhlin , b.n .
narozhny , g. schn , and a. shnirman , ann .
phys .
( berlin ) * 361 * , 401 ( 2015 ) .
d. s. golubev , m. marthaler , y. utsumi , and g. schn , phys .
rev .
b * 81 * , 184516 ( 2010 ) . d.a .
bagrets , and yu.v .
nazarov , phys .
rev .
b * 67 * , 085316 ( 2003 ) .
j. p. pekola , phys .
rev .
lett . * 93 * , 206601 ( 2004 ) . c. flindt , c. fricke , f. hohls , t. novotny , k. netocny , t. brandes and r. j. haug , proc .
natl .
acad .
sci . * 106 * , 10116 ( 2009 ) .
e. g. klepfish , c. e. creffield , e. r. pike , nucl .
phys .
b ( proc . suppl . )
* 63a - c * , 655 ( 1998 ) .
f. a. wolf , a. go , i. p. mcculloch , a. j. millis , and u. schollwck , phys .
rev .
x * 5 * , 041032 ( 2015 ) .
k. r. brown , r. j. clark , and i. l. chuang , phys
. rev .
lett . * 97 * , 050504 ( 2006 ) .
d. leibfried , nature ( london ) * 463 * , 608 ( 2010 )
. t. h. johnson , s. r. clark , and d. jaksch , epj quantum technology * 1*,10 ( 2014 ) .
q. deng , d. v. averin , m. h. amin and p. smith , sci . rep . * 3 * , 1479 ( 2013 ) .
i. schwenk and m. marthaler , phys .
rev .
b * 93 * , 014305 ( 2016 ) . c. jarzynski , j. stat . mech . * 2004 * , p09005 ( 2004 ) .
w .- m .
zhang , p .- y .
lo , h .-
n . xiong , m. w .- y .
tu , and f. nori , phys .
rev .
lett . * 109 * , 170402 ( 2012 ) .
p .- y .
lo , h .-
n . xiong , and w .- m .
zhang , sci .
rep . * 5 * , 9423 ( 2015 ) .
see the supplementary materials for details .
j. m. raimond , m. brune , and s. haroche , rev .
mod .
phys . * 73 * , 565 ( 2001 ) .
s. m. girvin , in lecture notes on _ strong light - matter coupling : from atoms to solid - state systems _
( world scientific , singapore , 2013 ) .
j. q. you and f. nori , nature ( london ) * 474 * , 589 ( 2011 ) .
m. aspelmeyer , t. j. kippenberg , and f. marquardt , rev .
mod .
phys . * 86 * , 1391 ( 2014 ) .
y. chen , j. phys .
b : at . mol .
. phys . * 46 * , 104001 ( 2013 ) .
a. j. leggett , s. chakravarty , a. t. dorsey , m. p. a. fisher , a. garg , and w. zwerger , rev .
mod .
phys . * 59 * , 1 ( 1987 ) . the density operator of the electron gas can also be written as @xmath129 , where @xmath130 ( @xmath131 ) is the annihilation ( creation ) operator of an electron with position @xmath132 and spin @xmath106 , @xmath133 only when @xmath134 , and @xmath135 otherwise . * supplemental materials for `` detector readout of analog quantum simulators '' * |
the problem of proving or verifying fourier s law , @xmath6 where @xmath7 is the thermal conductivity , in any system evolving through newtonian dynamics has been a challenge for theorists @xcite . so far most studies have been restricted to one - dimensional systems for the simple reason that they are easier to study through simulations and through whatever analytic methods that are available .
also the hope is that such studies on one - dimensional systems provides insights which will be useful when one confronts the more difficult ( and experimentally more relevant ) problem of higher dimensional systems .
for one dimensional systems , some of the most interesting results that have been obtained are : ( 1 ) for momentum conserving nonlinear systems , the heat current @xmath1 decreases with system size @xmath0 as @xmath8 where @xmath9 @xcite .
thus fourier s law ( which predicts @xmath10 ) is not valid .
the exponent @xmath11 is expected to be universal but its exact value is still not known . a renormalization group analysis of the hydrodynamic equations @xcite predicts @xmath12 while mode - coupling theory @xcite gives @xmath13 .
the results from simulations are not able to convincingly decide between either of these .
( 2 ) for disordered harmonic systems we get @xmath8 again but the exponent @xmath11 depends on the properties of the heat baths @xcite
. in dimensions higher than one , there are few detailed studies and it is fair to say that it is totally unclear as to whether or not fourier s law will hold and , if not , then what the value of the exponent @xmath14 is . for nonlinear systems which are expected to show local thermal equilibrium ,
both the hydrodynamic approach and mode coupling theories predict a logarithmic divergence of the conductivity in two dimensions .
there have been simulations by lippi and livi @xcite who find a logarithmic divergence but simulations on larger - size systems by grassberger et al @xcite seem to obtain a power law divergence . a disordered harmonic model in @xmath15 was studied in simulations by yang @xcite who claimed that beyond some critical disorder one gets fourier s law , _
i.e _ @xmath16 .
it is doubtful if this claim is correct .
the data in the paper seems to indicate @xmath17 which is _ not _ fourier s law . besides
, these simulations were done with nose - hoover heat baths and it is known that these can be problematic when applied to harmonic systems @xcite .
simulations by hu et al @xcite on the same model but with stochastic heat baths do not find a fourier behaviour .
finally an older study by poetzsch et al @xcite looked at heat conduction in a @xmath15 system with both disorder and nonlinearity and they give some evidence for fourier behaviour . in this paper
we consider heat conduction in a @xmath15 disordered harmonic system .
let us first try to see what one should expect theoretically .
we expect localization phenomena ( for phonons ) to play an important role .
a renormalization group calculation @xcite predicts : in @xmath18 all modes with @xmath19 are localized ; in @xmath15 all modes with @xmath20^{-1/2}$ ] are localized ; in @xmath21 there is a finite band of frequencies of extended states .
this is similar to results for electron localization with the important difference that here the @xmath22 modes are extended even in @xmath18 and @xmath15 .
also for the case of electrons , only electrons near the fermi - level contribute significantly to transport while in heat transport all phonons contribute . from the localization
results we expect that in @xmath21 the current in a disordered harmonic system should be independent of system size ( @xmath23 ) . in one and two
dimensions it is the small number of low - frequency phonons ( @xmath24 ) which dominate transport properties .
the fact that @xmath25 with increasing @xmath0 immediately implies that @xmath26 . in @xmath18 it has been shown @xcite that the exact value of @xmath14 depends on the low frequency spectral properties of the bath . a similar calculation is not available in the @xmath15 case and we address this specific question .
here we present results from a detailed simulational study to determine the exponent @xmath14 for a mass - disordered harmonic system .
two different kinds of stochastic baths are considered , one with white noise and the other with correlated noise .
we also study a special case where the disorder is correlated . *
definition of model * : we consider heat conduction in a two - dimensional mass - disordered harmonic crystal described by the hamiltonian & h&= _ + _ [ ( x_ij - x_i-1j)^2 + & + & ( x_ij - x_i+1j)^2 + ( x_ij - x_ij-1)^2+(x_ij - x_ij+1)^2 ] where @xmath27 denote the position ( particle displacements about equilibrium positions ) , momentum and mass of a particle at the site @xmath28 .
we set the masses of exactly half the particles to one and the remaining to two and make all configurations equally probable .
heat conduction takes place in the @xmath29-direction and we assume that the ends of the system are fixed by the boundary conditions @xmath30
. we will assume periodic boundary conditions along the @xmath31-direction so that @xmath32 .
the heat baths are modeled through langevin equations and thus we get the following equations of motion : & & m_1j _ 1j=-4 x_1j+x_2j+x_1j-1+x_1j+1 + h^l_j + & & m_l_xj _
l_xj=-4 x_l_xj+x_l_x-1j+x_l_xj-1+x_l_xj+1+h^r_j + & & m_ij _
ij=-4 x_ij+x_i-1j+x_i+1j+x_ij-1+x_ij+1 [ eqnmot ] ( for @xmath33 and @xmath34 ) , where @xmath35 and @xmath36 denote the forces from the heat baths .
we will consider two different models for the heat baths : * ( ii ) * _ gaussian exponentially correlated source_. in this case the bath forces have the forms : h^l_j&=&-_-^t dt ( t - t ) _
1j(t)+^l_j + h^r_j&=&-_-^t dt ( t - t ) _ l_xj(t)+^r_j with @xmath40 , @xmath41 and @xmath42 .
a simple way of implementing correlated baths in the simulations is by introducing new dynamical variables @xmath43 for the bath and setting @xmath44 .
these satisfy the equations of motion @xmath45 , etc . in the long time
limit it can be easily seen that the solutions @xmath46 and @xmath47 have the required properties of correlated baths .
fully disordered lattice , from simulations and from the exact solution .
inset shows the disorder - averaged temperature profiles for different system sizes and seems to approach a linear form.,width=288 ] * simulations with white noise * : equilibration times in simulations of disordered harmonic lattices can be very long and this can sometimes lead to wrong conclusions [ see for e.g @xcite ] . to avoid such problems
we first compare our simulation results with exact numerical results on steady state properties of small systems .
we now briefly describe the numerical technique .
with @xmath48 let us define the new variables @xmath49 . then eq .
[ eqnmot ] can be rewritten in the form : _
l=-_m=1 ^ 2n _ lmq_m+_l [ neqmot ] where the vector @xmath50 has all elements zero except @xmath51 ( for @xmath34 ) and the @xmath52 matrix @xmath53 is given by : = ( ) where the @xmath54 matrices @xmath55 can be labeled by the double indices @xmath28 and are given by & & _
( ij),(ij)= _ ii _
jj m_ij ; _ ( ij)(ij)=4_ii _ jj + & & -_i i(_jj-1+_jj+1)-_j j ( _ i i-1+_i i+1 ) + & & _
( ij)(ij)=_ii_jj ( _ i1/m_1j+_il_x /m_l_xj ) .
[ defn ] in the steady state @xmath56 . from this and using eq .
[ neqmot ] we get the matrix equation @xcite + ^t= = ( ) [ mateq ] where @xmath57 is the correlation matrix with elements @xmath58 and @xmath59 . inverting this equation
one obtains @xmath57 and thus all the moments which includes the local temperatures @xmath60 and currents @xmath61 .
the dimension of the matrix which has to be inverted is @xmath62 and using the fact that it is a sparse matrix we have been able to numerically @xcite obtain @xmath57 for system sizes up to @xmath63 .
the molecular dynamics simulations were performed using a velocity - verlet scheme @xcite .
we chose a step - size of @xmath64 and averaged over @xmath65 time steps ( for @xmath66 we took @xmath67 and @xmath68 time steps ) .
the temperatures at the two ends were set to @xmath69 and @xmath70 . in fig .
[ exact ] , we plot @xmath71 at every site , as obtained from the simulations and from the exact solution , for a particular realization of disorder .
the agreement is clearly very good .
we also find that that current fluctuations decay faster than fluctations of the local temperature .
this is because , in the harmonic model , _ decay of fluctuations takes place only through coupling to the reservoirs _ and this is weak for localized modes that contribute to the temperature .
this means that equilibrated values for the current can be obtained in smaller simulation runs .
simulations were performed for sizes @xmath72 and in fig .
[ jvsl ] we plot the system size - dependence of the current , averaged over @xmath73 samples ( @xmath74 samples for @xmath66 ) .
error - bars shown are those calculated from the disorder averaging , the thermal ones being much smaller . for larger system sizes we find that we need to average over a smaller number of realizations since the _ rms _ spread in the current decreases rapidly . from our data
we estimate @xmath75 .
we briefly note that eq .
[ mateq ] can be solved exactly for the ordered case , using methods similar to those in @xcite .
the current is independent of system size and given by j&= & _ 0 ^ 2 dq _
1(q ) where @xmath78^{1/2 } $ ] and @xmath79 $ ] .
the temperature in the bulk of the system takes the constant value @xmath80 .
* simulations with correlated noise * : in this case the simulations were done by using a slightly modified version of the velocity - verlet algorithm with a step - size @xmath81 and averaging over @xmath65 time steps .
the accuracy of the algorithm was tested in one - dimensions where exact numerical results are available @xcite .
simulations were performed for sizes @xmath82 with disorder averages over @xmath73 samples ( @xmath83 for @xmath84 ) .
the results are plotted in fig .
[ jvsl ] and we estimate the exponent @xmath85 in this case , which is somewhat different from the slope of @xmath86 for the case of uncorrelated noise .
it is possible that the small difference is a finite size effect and for larger system sizes we might see the same exponent .
the error bars given are statistical errors while those from finite - size effects are more difficult to estimate .
the next example throws some light on this aspect .
* correlated disorder * : finally we consider a special case of correlated disorder ( with white noise baths ) which was discussed in @xcite .
this case is analytically tractable and gives us some insights on possible finite size effects . in this model , in a given column , say the @xmath87th , all particles have the same mass @xmath88 .
this case can be reduced to an effective one - dimensional problem @xcite .
using the fact that there is order in the transverse direction , we transform to new variables using an orthogonal basis @xmath89 which satisfies the equation @xmath90 .
we choose the @xmath89 to be real and find that @xmath91 with @xmath92 where @xmath93 .
the new variables @xmath94 satisfy the following equations of motion : m_1 _ 1(q)&=&-(q ) x_1+x_2-_1+^l(q ) + m_l_x _
l_x(q)&=&-(q ) x_l_x+x_l_x-1-_l_x+^r(q ) + m_i _
i(q)&=&-(q ) x_i+x_i-1+x_i+1 [ eqnmot2 ] ( for @xmath33 ) , where @xmath95 . the transformed noise variables @xmath96 satisfy @xmath97 and similarly for @xmath98 .
thus for every @xmath99 we have an equation identical to that of a one - dimensional disordered chain with an additional on - site potential @xmath100 .
the heat current in terms of the transformed variables is : j = _ q ( -_1(q ) + ^l(q ) ) _
1(q ) [ jcor ] fourier transforming eq .
[ eqnmot2 ] using : @xmath101 we get the set of equations : _
1- _ 2 & = & ^l + - _ i-1+[(q)-m_i ^2 ] _ i- _ i+1 & = & 0 + - _ l_x-1+[(q)-m_l_x ^2+i ]_l_x & = & ^r which in matrix notation can be written as @xmath102 where @xmath103 , @xmath104 are the vectors @xmath105 and @xmath106 .
the matrix @xmath107 where @xmath108 and @xmath109 . after some manipulations the current in eq .
[ jcor ] simplifies to give : j= _ q _ -^ d ^2 @xmath110|^{-2}$ ] where @xmath111= d_{1,l_x}+i \g \o ( d_{2,l_x}+d_{1,l_x-1 } ) -\g^2 \o^2 d_{2,l_x-1}$ ] and @xmath112 is defined to be the determinant of the submatrix of @xmath113 beginning with the @xmath87th row and column and ending with the @xmath114th row and column .
these matrix elements can be expressed in terms of products of random matrices @xcite .
using these results one can very efficiently compute the integral in eq .
[ meth2j ] and obtain @xmath1 accurately for quite large system sizes ( @xmath115 ) . in fig .
[ jvsl ] we show the system size dependence of the current as obtained from the exact numerical method and also from simulations .
they agree very well and give @xmath116 .
this value can be understood analytically by noting that the leading contribution to the current in eq .
[ meth2j ] comes from the @xmath117 term ( finite @xmath99 modes decay exponentially with system size ) and this is identical to a pure @xmath18 chain for which @xmath5 .
the fact that the simulation results agree extremely well with the exact numerical results ( for sizes up to @xmath84 ) proves the accuracy of our simulations .
further we see that for the correlated disorder case the asymptotic result for the exponent can already be seen at around @xmath115 .
this gives us confidence that for the case of the fully disordered lattice we might already be close to the asymptotic value .
this is also supported by the fact that the change in slope of the @xmath1-vesus-@xmath0 curves in fig .
[ jvsl ] over the system sizes studied is very small .
* conclusions : * we have performed extensive simulations of heat conduction in a mass - disordered harmonic solid in @xmath15 which give exponents @xmath3 for white noise heat baths and @xmath118 for correlated noise baths .
a system with correlated disorder gives , somewhat surprisingly , a larger exponent @xmath5 .
the combination of simulations and exact numerics gives us confidence on the accuracy of our results and also additional insight .
some interesting open problems are the exact determination of the exponent @xmath14 in @xmath15 , for any heat bath model , and an analytical understanding of dependence of @xmath14 on bath properties .
we acknowledge a. p. young for a critical reading of the manuscript .
lwl acknowledges o. narayan and t. mai for useful discussions .
the work of one of us ( lwl ) is supported by the nsf under grant dmr 0337049 .
we are grateful to the hierarchical systems research foundation ( hsrf ) for the usage of their computing cluster . * * f. bonetto , j.l .
lebowitz , l. rey - bellet , in mathematical physics 2000 , edited by a. fokas et al .
( imperial college press , london , 2000 ) , p. 128 . s. lepri , r. livi , and a. politi , phys .
rep . * 377 * , 1 ( 2003 ) .
t. hatano , phys .
e * 59 * , r1 ( 1999 ) ; a. dhar , phys . rev .
lett . * 86 * , 3554(2001 ) ; p. grassberger , w. nadler and l. yang , phys . rev . lett . * 89 * , 180601 ( 2002 ) ; g. casati and t. prosen , phys . rev .
e 67 , 015203 ( r ) ( 2003 ) ; j. m. deutsch and o. narayan phys .
e 68 , 010201(r ) ( 2003 ) .
o. narayan and s. ramaswamy , phys .
lett . * 89 * , 200601 ( 2002 ) .
s. lepri , r. livi , and a. politi , europhys . lett . * 43 * , 271 ( 1998 ) .
a. dhar , phys .
lett . * 86 * , 5882 ( 2001 ) .
a. j. oconnor and j. l. lebowitz , j. math . phys . *
15 * , 692 ( 1974 ) .
r. rubin and w. greer , j. math .
* 12 * , 1686 ( 1971 ) .
a. casher and j. l. lebowitz , j. math . phys .
* 12 * , 1701 ( 1971 ) .
a. lippi and r. livi , j. stat phys .
* 100 * , 1147 ( 2000 ) .
p. grassberger and l. yang , arxiv : cond - mat/0204247 , ( 2002 ) .
l. yang , phys .
lett . * 88 * , 094301 ( 2002 ) .
d. frenkel and b. smit , understanding molecular simulation ( academic press , 2002 ) ; a. dhar ( unpublished notes ) .
b. hu et al , phys .
lett . * 90 * , 119401 ( 2003 ) .
r. h. h. poetzsch and h. bottger , phys .
b * 50 * , 15757 ( 1994 ) .
s. john , h. sompolinsky and m. j. stephen , phys .
b * 27 * , 5592 ( 1983 ) .
a. dhar , phys .
lett . * 87 * , 069401 ( 2001 ) . z. rieder , j. l. lebowitz , and e. h. lieb , j. math . phys . * 8 * , 1073 ( 1967 ) .
j. w. demmel , j. r. gilbert and x. s. li , superlu users guide ( 1999 ) .
lbnl technical report-44289 .
m. p. allen and d. l. tildesley , computer simulation of liquids ( clarendon press oxford , 1987 ) . | we study the problem of heat conduction in a mass - disordered two - dimensional harmonic crystal . using two different stochastic heat baths , we perform simulations to determine the system size ( @xmath0 ) dependence of the heat current ( @xmath1 ) . for white noise heat baths
we find that @xmath2 with @xmath3 while correlated noise heat baths gives @xmath4 .
a special case with correlated disorder is studied analytically and gives @xmath5 which agrees also with results from exact numerics . |
acceptable theories of structure formation are those whose parameters have been tuned to match the most up - to - date cosmological observations . at high redshift ,
the current leading constraint on models is the data from cobe ( @xcite ) , which fixes the amplitude of the power spectrum on very large scales ( @xmath11mpc ) to an accuracy of about 12% . at low ( essentialy zero ) redshift , we demand that models fit current observations of our local universe , primarily those concerning the distributions of galaxies in @xmath12 space .
these include the abundance of clusters of galaxies , which fixes the amplitude of the power spectrum on scales of @xmath13mpc to about 10% accuracy ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; pen 1996 ) , the power spectrum of galaxies , which constrains the shape of the power spectrum on the intermediate - to - large scale of @xmath14mpc ( peacock & dodds 1994 ; feldman , kaiser & peacock 1994 ) , the ratio of gas to total matter in galaxy clusters , which determines @xmath15 ( @xcite ) .
in addition , the current measurements of the hubble constant ( fukugita , hogan , & peebles 1993 ; freedman 1994 ; @xcite ; hamuy 1995 ) and the age constraint from the oldest globular clusters ( bolte & hogan 1995 ) limit the range for @xmath16 and the combination of @xmath17 and @xmath18 . this observational suite has been examined by ostriker & steinhardt ( 1995 ) to constrain flat cold dark matter models with a non - zero cosmological constant ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
the exercise is repeatable for other models ( not necessarily flat ) including the tilted cold dark matter model ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , the mixed dark matter model ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , the open cold dark matter model ( gott 1982 ; bucher , goldhaber , & turok 1995 ) and the primeval isocurvature baryon model ( peebles 1987a , b ; cen , ostriker , & peebles 1993 ) . the allowed parameter space for the family of gaussian cosmological models with the tunable parameters ( @xmath16 , @xmath17 , @xmath18 , r@xmath19
, @xmath20 , iso , adia ) is thus quite limited . in this context ,
@xmath16 , @xmath17 , and @xmath18 are the hubble constant , the density parameter , and the value of cosmological constant , all in the present epoch .
r@xmath19 is the mass ratio of hot to cold matter , @xmath20 is the asymptotic power spectral index on large scales , and iso and adia denote isocurvature and adiabatic models . although it is possible to further tighten this parameter space simply by making more accurate observations of the forementioned quantities , we should not overlook any other _ independent _ tests which could help to distinguish between contending models .
the @xmath0 forest lines observed in the spectra of high redshift quasars have two unique qualities which distinguish them from other observationally accessible phenomena , at redshifts between the epoch observed by cobe and our local universe .
first , each line of sight indiscriminately samples the distribution of neutral hydrogen gas over a wide redshift range ( @xmath21 ) along random lines of sight ( i.e. , foreground objects are unrelated to the background quasar ) .
second , each individual spectrum contains a large amount of information about low absorption regions ( e.g. voids , fluctuating gunn - peterson absorption ) as well as information from high absorption `` cloud '' lines ( number densities of clouds at different redshifts , @xmath22-parameters of the individual clouds , correlations of the clouds , relationship with other cosmic entities ) . with a database of thousands of observed quasars ,
the total amount of information available is very large , allowing for very detailed statistical studies ( e.g. , carswell 1991 ; rauch 1993 ; petitjean 1993 ; schneider 1993 ; cristiani 1995 ; hu 1995 ; tytler 1995 ) .
these facts , taken together , suggest that the @xmath0 forests constitute perhaps the most rich and unbiased sample available for studying the universe at moderate redshift .
recent cosmological hydrodynamic simulations by several independent groups have consistently shown that @xmath0 clouds are an integral part of the cosmic structure , resulting naturally from the gravitational growth and/or collapse of density fluctuations on small - to - intermediate scales of @xmath23 kpc to a few mpc in comoving units ( cen 1994 , cmor hereafter ; zhang 1995 ; hernquist 1996 ) .
several simple population models have been designed to examine the ( local ) physical as well as ( global ) statistical properties of @xmath0 clouds ( sargent 1980 ; ostriker & ikeuchi 1983 ; ikeuchi & ostriker 1986 ; bahcall & spitzer 1969 ; arons 1972 ; rees 1986 ; ikeuchi 1986 ; bond , szalay , & silk 1988 ) . one of the essential simplifications in almost all these models is to assume that individual @xmath0 clouds are spherical .
while the clouds produced in the new simulations may have some physical properties in common with these simple models , a visual inspection of the new simulation results reveals that the @xmath0 absorbing structures resulting from small - to - intermediate scale structure formation at high redshift are _ far from spherical_. furthermore , they seem to have a wide range of sizes .
we presented these results first in cmor , and then in much more detail in miralda - escud ( 1996 , mcor hereafter ) . here , we present a quantitative study of the topological aspects of the @xmath0 clouds , complementing the topics covered in cmor and mcor . in the redshift range from two to four ,
@xmath0 clouds exhibit a rich spectrum of structure , ranging in shape from semi - spherical to filamentary and even sheet - like .
the spherical structures often reside at density maxima ( with @xmath24@xmath6 ) located in the centers of extended structures , while the filaments and sheets tend to have a low @xmath25 ( @xmath26@xmath6 ) , and form a web - like , interconnecting network covering large portion of the simulation box ( @xmath27mpc comoving ) .
we also show that it is not unusual for pairs of quasar sightlines to contain absorption features at close wavelengths .
in particular , we show that `` common '' absorption features ( coincident lines ) in sightline pairs separated by @xmath28 proper kpc or more are probably due to absorption by _
different clouds _ , while on separations smaller than @xmath29kpc , double sightlines are mostly likely to actually pierce a common cloud . finally , we show that @xmath0 clouds are spatially significantly clustered , with a correlation length of roughly @xmath30mpc comoving ( for the high column density clouds of @xmath5@xmath6 ) at redshift @xmath31 .
this suggests the possibility of an intriguing connection to faint blue objects ( @xcite ; @xcite ; @xcite ) .
this paper is organized as follows .
some brief descriptions of the simulations ( for more details see mcor ) and our cloud identification method are presented in 2 . results and conclusions
are given in 3 and 4 .
we simulate the formation of @xmath0 clouds in a spatially flat cold dark matter universe with a cosmological constant ( @xmath32cdm ) , using the following cosmological parameters : @xmath33km / s / mpc , @xmath34 , @xmath2 , @xmath35 ( cf .
walker 1991 ) , @xmath3 ( the simulations we use in this paper are the same as those used in cmor and mcor ) .
the primary motivation for choosing this model is that it best fits the available observations which were summarized in the introduction .
the simulation box size is @xmath36 comoving mpc per side , and contains @xmath37 cells and @xmath38 dark matter particles .
the cell size is @xmath39 comoving kpc , corresponding to a average baryonic cell mass of @xmath40 , with the true spatial and mass resolutions being about 2 and 8 times worse than those values , respectively . at @xmath41 ,
the jeans length , @xmath42 for @xmath43 , is equal to @xmath44 in comoving units , or 11 cells .
the power spectrum transfer function is computed using the method described in cen , gnedin , & ostriker ( 1993 ) .
we use a new shock - capturing total variation diminishing ( tvd ) cosmological hydrodynamic code described in ryu ( 1993 ) .
all the atomic processes for a plasma of ( h , he ) of primeval composition ( 76%,24% ) in mass are included , using the heating , cooling , and ionization terms described in cen ( 1992 ) .
we calculate self - consistently the average background photoionizing radiation field as a function of frequency , assuming the radiation field is spatially uniform ( i.e. , optically thin ) .
the evolution of the radiation field is calculated given the average attenuation in the simulated box and the emission ( both from the gas itself and from the assumed sources of ionizing photons ) .
the time - dependent equations for the ionization structure of the gas are solved by iteration using an implicit method , to avoid the instabilities that arise in solving stiff equations . in general , the abundances of different species are close to ionization equilibrium between recombination and photoionization after most of the gas has been photoionized .
we model galaxy formation as in cen & ostriker ( 1992 , 1993a , b ) .
the material turning into collisionless particles as galaxies " is assumed to emit ionizing radiation , with two types of spectra : one characteristic of star formation regions and the other characteristic of quasars , with efficiencies ( i.e. , the fraction of rest - mass energy converted into radiation ) of @xmath45 , and @xmath46 , respectively .
we adopt the emission spectrum of massive stars from scalo ( 1986 ) and that of quasars from edelson and malkan ( 1986 ) .
details of how we identify galaxy formation and follow the motions of formed galaxies have been described in cen & ostriker ( 1993a ) .
note that in this simulation supernova energy feedback into the intergalactic medium from aging massive stars is not included .
perhaps the most critical decision which must be made in a study such as this is how to define a @xmath0 `` cloud '' .
this is not as simple as it may seem , because the smooth transition in density between the global intergalactic medium and the local structures within it can blur the distinction between actual clouds and the intercloud medium .
however , in order to examine the structure of @xmath0 clouds in any quantitative way , it is absolutely necessary to adopt some sort of definition of the boundary of a cloud .
the simplest approach which we believe to be meaningful is to identify clouds as regions with densities above a chosen threshold .
there are two major motivations for defining our clouds in this manner .
first of all , this method allows us to associate the clouds found at each density cut with a particular column density , once we obtain information about the characteristic sizes of the clouds .
this is a desirable feature because we can only observe @xmath0 forest clouds through absorption lines in qso spectra .
second , density perturbations with amplitudes larger than a certain threshold become self - gravitating , bound clouds . for a spherical cloud , the average overdensity at which a cloud breaks away from the general hubble expansion and becomes self - gravitating is @xmath47 .
thus , our cloud defintion also has physical motivation in that structures above some high density threshold are gravitationally bound , distinct systems .
having chosen a general strategy for defining our clouds , we now describe the details of the grouping procedure .
cells with a baryonic density below a chosen value ( 3,10 , and 30 will be used ) in units of the global mean of the baryonic density are cut out of the original density array .
then , clouds are defined by grouping the remaining cells using the denmax scheme ( @xcite ) as follows .
first , two sets of baryonic densities are stored .
one is the original density output array from the simulation ( @xmath48 , @xmath49 elements ) , and the second is a smoothed version of @xmath48 with a gaussian smoothing window of radius 1.5 cells ( @xmath50 , @xmath49 elements ) .
the smoothing is performed to eliminate any small , cell - to - cell density fluctuations , which could be physical in the form of small discontinuities or oscillating sound waves , or just numerical noises .
such small - scale fluctuations could result in the identification of small unwanted clouds with sizes on the order of one cell . at the same time , since the true resolution of the simulation is close to 2 - 3 cells , a gaussian smoothing window of radius 1.5 cells should preserve all information regarding `` real '' adjacent structures . a larger smoothing window could merge real and separate entities which are actually well - resolved in the simulation .
we have made rather extensive experiments on the smoothing operation and conclude that the adopted smoothing window size is appropriate , as will be shown below . after the smoothing operation is performed , we cut out all elements of @xmath50 and @xmath48 whose @xmath48 values are below a chosen value , @xmath51 .
we then collect the uncut cells of @xmath48 ( using the denmax scheme ) by propagating each cell along the gradient of @xmath50 until it reaches a local density maximum .
all the cells collected at a particular local maximum are grouped into one cloud " .
once the cells have been grouped , we examine the shapes and sizes of the resultant clouds " by approximating them as _ ellipsoids_. a symmetric @xmath52 moment tensor is created for each cloud of the form @xmath53 ; @xmath54 ; @xmath55 and so on .
this tensor is diagonalized , and the square roots of the eigenvalues yield the lengths of the semi - major and semi - minor axes , @xmath56 , in descending order of length . for cells
where @xmath57 , @xmath58 or @xmath59 , we account for the finite size of a single cell by finding @xmath60 for a 1 cell square and adding that to the tensor component for a particular cloud .
this method underestimates the actual axial lengths of an ellipsoid , because it only computes moments .
we attempt to correct this by scaling up the values of @xmath61 and @xmath62 ( determined above ) by a factor of @xmath63 : @xmath64 where @xmath65 is the actual volume of the cloud computed by summing over all the member cells of the cloud .
we first present a three dimensional visual description of the structure of the @xmath0 clouds produced in the simulation .
figures ( 1a , b , c ) display the isodensity contour surfaces at @xmath41 for @xmath66 , respectively . here , @xmath51 is the baryonic density in units of its global mean .
the box size is @xmath36 comoving mpc .
the most striking visual feature is the network of sheets and filaments spanning the box . in figure
1a we see that most of the structures are connected over the entire box , and it is apparent that most of the covering areas [ i.e. , cross section for quasar lines of sight at low column densities ( @xmath26 @xmath6 ; see below ) ] are sheet - like objects .
although this result is for the particular @xmath32cdm universe simulated here , we expect this qualitative picture to be fairly generic for any cdm - like model ( hdm - like models should show even more prominent sheets , while models like pbi are likely to be less coherent ) . at a higher density of @xmath67 ( figure 1b )
most structures become filamentary and more isolated .
when the isodensity reaches @xmath68 ( figure 1c ) , most of the filaments are replaced by relatively round systems , which are typically isolated with separations of about one comoving megaparsec .
the second noticeable visual feature is the existence of a large , low density region at the center of the box , occupying more than half of the simulation volume .
since the initial condition of the simulation is not constrained ( i.e. , it is only a random realization of the cosmological model in such a volume ) , this void indicates that the simulation box may still not be large enough to contain a fair " volume for the structures under consideration .
it may be necessary to use a larger simulation box , perhaps @xmath69mpc on a side , in order to properly sample the objects in question .
different properties will be affected by the simulation volume to varied degrees .
we expect that the most significant differences will be in quantities like void sizes , or correlations .
let us now examine the individual clouds to assess the general accuracy of our grouping scheme .
figure 2a shows clouds at @xmath70 for two randomly selected slices of size @xmath71mpc@xmath72 with thickness of @xmath73kpc ( all lengths are in comoving units ) at @xmath41 .
the two left panels show the density contours of levels @xmath74 , and the two right panels show the identified @xmath0 clouds in the same slices .
two symbols are used to show two types of clouds : the filled dots represent large " clouds with the dot size roughly proportional to the actual size of the cloud , and the open circles are for the small " clouds .
a cloud is called small " if its smallest axis does not exceed 3 cells in length , and large " if it does .
figure 2b is similar to figure 2a , but with @xmath67 , and for two different slices .
we note that large " clouds are typically embedded in larger , extended density structures while small " ones are more isolated . a close
, one - by - one examination of the density maxima in the density contour plots indicates that each identified cloud corresponds to a well defined density maximum ( some clouds , especially
small " ones , have no corresponding density maxima simply because their densities fall below the contour levels when averaged over the slice thickness to be displayed ) .
it appears that our smoothing operation and cloud identification scheme indeed yield well defined , distinct clouds with one particularly desirable feature : they are neither over - merged in the sense that a single cloud contains multiple structures of comparable size ( over - smoothed ) , nor over - separated in the sense that artificial clouds are created by small - scale small - amplitude density fluctuations ( undersmoothed ) .
when the clouds are classified into large " and small " groups , we find that most of the clouds are
small " : the fraction of small " clouds at @xmath66 at @xmath41 is ( @xmath75 ) , containing ( @xmath76 ) of the total cloud mass .
the fact that a large fraction of the baryonic mass is in small " clouds indicates that a _ higher resolution simulation is perhaps needed _ before we can be absolutely sure that these clouds are resolved properly .
inclusion of more initial small scale power , which is limited by the nyquist frequency of a simulation , might further increase the fraction of small " clouds .
we note in passing that we also experimented with the conventional friends - of - friends grouping algorithm , with a linking length of one cell .
the resulting clouds were obviously not useful for our analyses .
for example , at @xmath70 , most of the volume , as well as mass , wound up in a single supercloud because the distinct clouds as shown in figure 2 were linked together by touching boundaries . in other words , the friends - of - friends grouping scheme picks out large , inter - connected networks ( see figure 1a ) but not the individual structures embedded within them . we have chosen to focus on three quantitative aspects of the identified @xmath0 clouds : their sizes , shapes , and correlations .
in addition , double quasar sightline analysis is performed . to facilitate quantitative discussions ,
we relate the mass density of a cloud to its column density . assuming that the clouds are in photoionization equilibrium , the column density of a cloud can be related to the parallel size and density of the cloud ( assuming that the density is uniform across the cloud along the line of sight ) as @xmath77 where @xmath78 is the baryonic density parameter , @xmath79 the temperature in kelvin , @xmath80 the hydrogen photoionization rate , @xmath81 and @xmath82 the baryonic density of the cloud and the mean baryonic density at the redshift in question , @xmath83 the proper size of the cloud along the line of sight , and @xmath84 the redshift . for a photoionization radiation field with power - law form @xmath85 , @xmath86 sec@xmath87 , where @xmath88 is the intensity of the photoionization field at the lyman limit in the usual units ( @xmath89erg / cm@xmath72/hz / sec / sr ) . taking @xmath91 , @xmath92sec@xmath87 , @xmath93 kelvin , @xmath94 kpc as typical values at @xmath41
, one finds that @xmath95 for @xmath66 . of course , exact column densities would depend on exact values of the assumed quantities as well as the actual density distribution .
we expect that the actual density distribution , which is non - uniform , will increase the column densities .
for example , redistribution of the same mass within @xmath83 into a ( one dimensional ) coreless , non - uniform one would result in a column density @xmath96 times that of the uniform density distribution , where @xmath97 is the slope of the density profile .
we simply increase the above estimates by a factor of four to account for the gradient in the density distribution of a cloud [ note that this factor of four is fairly plausible for a reasonable @xmath97 ( @xmath98 ) in the one dimensional singular case or a relatively lower @xmath97 value for a non - singular profile .
we also note that this factor of four can also be achieved if @xmath78 is twice the value adopted here , in the light of recent new measurements on the deuterium abundance ( @xcite ) ] , thus obtaining @xmath99 we will use the above relation for quantitative discussions . before engaging in a more detailed quantitative discussion of cloud properties , we present a few simple , global quantities . listed in table 1
are the numbers of clouds at each redshift and density threshold , the baryonic mass fraction of such clouds , the mean inter - cloud separation in proper units , and the mean cloud mass .
two points are interesting .
first , while the average cloud increases its mass by a factor of about @xmath100 from @xmath101 to @xmath102 , the number of clouds decreases by a factor two .
this indicates that the merging of old clouds and creation of new clouds are of comparabe importance .
second , the fraction of baryonic mass contained in @xmath0 clouds is large . for example , at redshifts ( 2,3,4 ) , ( 31% , 29% , 26% ) of baryons are found in regions with densities between @xmath70 and @xmath68 ( i.e. , @xmath103@xmath6 ) .
we define the characteristic size " of a cloud as @xmath104 where @xmath105 , @xmath22 , and @xmath106 are the semi - axes of the ellipsoid approximating the cloud , measured in proper physical units ( see 2.2 for definition of @xmath107 ) . for a true ellipsoid cloud , the volume is @xmath108 .
figure 3 shows the cumulative distribution of cloud sizes at @xmath41 for @xmath66 , respectively .
the thin curves are weighted by @xmath109 , where @xmath110 and the thick curves are weighted by mass .
@xmath109 is an approximation of the average covering area of a cloud on the sky , which is proportional to the probability of the cloud intersecting a line of sight .
the @xmath109-weighted and mass - weighted distributions are very similar , implying that @xmath109 and mass are correlated .
clouds are larger at lower @xmath51 , as expected , and they exhibit a wide range in size , from a few to about a hundred proper kiloparsecs .
the median size varies from @xmath111 kpc at @xmath112 to @xmath113 kpc for @xmath114 .
note that @xmath115 is only the cube root of the effective volume of a cloud , so the distribution of cloud sizes would be more varied , if we used @xmath105 ( semi - major axis of an ellipsoid ) instead of @xmath115 as a measure of size .
figure 4 shows the redshift evolution of the median size of the @xmath0 clouds .
note that here the size ( vertical axis ) is in comoving length units , which might be easier to interpret .
our data indicates that @xmath0 clouds grow with time in comoving length units at a moderate rate , and that this rate depends _ weakly _ on the density ( i.e. , three curves for the three density cases are almost parallel to each other ) . the fact that denser regions are not shrinking , but rather expanding at a rate similar to that of less dense regions implies that the merger / accreation rate is _ higher _ in the dense regions than in less dense regions . without merger / accretion effects
, the high density regions would appear to shrink due to gravitational collapse . to summarize , we find that @xmath0 clouds exhibit a wide range of sizes from a few proper kiloparsecs to about a hundred proper kiloparsecs at z=3
. the comoving size of these clouds tends to increase slowly with time , implying that the proper physical size increases with time more rapidly than @xmath116 .
figures ( 5a , b , c ) show the distributions of cloud shapes in the @xmath117 - @xmath118 plane for clouds at @xmath66 at @xmath41 .
each dot represents one cloud and the contours indicate the density of clouds ( weighted by @xmath109 ) in the plane .
the contour levels are incremented up linearly from outside to inside .
we see that the most common cloud at @xmath70 and @xmath119 is an ellipsoid with axial ratios of @xmath120 .
however , at @xmath121 the situation is interestingly different .
two major concentrations of clouds are seen at @xmath122 and @xmath123 , indicating the existence of two distinct populations : filaments and near - spherical ellipsoids . in order to conveniently show the evolution of shapes of the clouds and to relate shapes to other quantities ,
we define a simple shape " parameter of a cloud as @xmath124 for an ideal ( spherical , filamentary , disk ) cloud , @xmath125 . some degeneracy exists for any value of @xmath126 with such a simple definition .
for example , @xmath127 and @xmath128 both give @xmath129 .
figure 6 shows @xmath126 as a function of size @xmath115 at @xmath41 for @xmath66 . there is a trend , albeit with a large scatter , that larger clouds tend to be less spherical .
figure 7 represents the @xmath109-weighted cumulative distribution of @xmath126 for three cases at @xmath41 .
we note two points .
first , we see that there is negligibly small fraction of clouds with the shape parameter @xmath126 greater 1.5 , indicating that spherical clouds ( @xmath130 ) contribute only a tiny percentage to the total covering area of the clouds at all three density cuts .
second , we find a larger fraction of round clouds ( large @xmath126 ) _ and _ filamentary clouds ( @xmath126 close to zero ) at high densities than at a lower densities .
this is due to the fact that the large sheets and filaments present at lower densities are broken up into both filaments and turn into roundish regions at higher densities , respectively .
this quantitative result is consistent with the visual impression that structures go from sheets to filaments and spheres as the density goes from low to high ( figure 1 ) .
the fact that there are more filaments than spheres at all densities is probably a consequence of nonlinear evolution coupled with hydrodynamic ( pressure ) effects .
the situation for dark halos might be different , since dark matter distributions tend to be more unstable in lower dimensional structures . finally , we show in figure 8 the redshift evolution of the median shape paramter @xmath131 .
it indicates that at @xmath70 the median shape of the clouds do not change with time , while at @xmath68 the clouds progress toward the shape of spheres ( larger @xmath126 ) in time with a dramatic upturn from @xmath41 to @xmath102 , possibly related to the collapse / merging of gas along the longest axis .
the situation at the intermediate @xmath67 is between the above two cases . in summary
, we have shown that @xmath0 clouds display a variety of shapes ranging from semi - spherical clouds to filaments and sheets , with the most common shape being a flattened cigar " with an axial ratio of @xmath120 at column density of @xmath132 @xmath6 , and being either a thin cigar with an axial ratios of @xmath133 or near sphere with an axial ratios of @xmath134 at column density of @xmath135@xmath6 .
larger clouds tend to be less spherical .
the shapes of clouds with densities a few times the mean density of the universe evolve very weakly with time , whereas the clouds at higher densities ( @xmath136 ) grow more spherical with time .
none of our analyses so far have been directly comparable with observables .
here we attempt to project our theoretical models onto the observational plane .
cmor and mcor have examined many of the properties of these clouds in detail , and found that the simulation results match observations fairly well . here ,
we focus on the observations of double quasar sightlines ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
let us first describe our procedure for modeling these observations .
once three dimensional clouds in the simulation box are identified ( see 2.2 ) , each cell is labelled with an integer @xmath20 , meaning that it belongs to cloud @xmath20 ( if @xmath20 is zero , it means that the cell does not belong to any cloud , i.e. , its density is below @xmath51 ) .
a random pair of sightlines separated by @xmath137 is selected .
since the simulation box has no preferred orientation , we simply choose a direction perpendicular to one of three faces of the simulation box as the direction for the double sightlines .
\3 . along each of the two sightlines we identify cells whose labels @xmath20 ( see step 1 ) are non - zero , and then separate regions of different @xmath20 into separate clouds , which we call clouds - along - line - of - sight ( calos s ) .
note that each calos inherits its 3-d parent cloud label @xmath20 .
we define the location of each calos in @xmath84-space by finding the @xmath138 weighted center of the calos cells located along the sightline . \5 . for each calos along sightline # 1 , if a calos within a velocity space separation @xmath139 along the sightline # 2 is found , the following operation is performed : an integer counter @xmath140 is incremented if the two calos s share the same @xmath20 , i.e. , they belong to the same 3-d cloud
. otherwise an integer counter @xmath141 is incremented by @xmath142 .
if both a common and a clustering line are found , we consider it to be a common pairing .
the above exercise is repeated for each calos along sightline # 2 .
we note that the bulk peculiar velocity of each calos is ignored in this calculation , i.e. , the line - of - sight velocity difference between a pair of coincident lines simply reflects the line - of - sight real space distance difference .
this should be a good approximation , because the peculiar velocity gradient is typically smaller than the hubble constant on the scale @xmath139 ( see below ) in which we are interested .
we perform this exercise on 10,000 randomly selected pairs of sightlines for each value of @xmath66 at six different perpendicular separations , @xmath143 cells .
figure 9 shows the fraction of lines which are coincident at @xmath41 , @xmath144 , as a function of @xmath137 , where @xmath145 and @xmath146 is total number of lines .
six curves are shown , corresponding to three values of @xmath66 , each at two values of @xmath147 km / s .
we see that _ the line coincident rate decreases with increasing column density_. at a perpendicular proper separation of @xmath28kpc and @xmath70 ( corresponding to clouds with @xmath148@xmath6 ) about @xmath149 to @xmath150 of the sightlines should have coincident absorption features for @xmath151km / s to @xmath152km / s .
the line coincident rates drop to @xmath153 and @xmath154 for @xmath67 and @xmath68 ( corresponding to clouds with @xmath155@xmath6 and @xmath156@xmath6 ) , respectively , at @xmath147km / s .
figure 10 shows the redshift evolution of @xmath144 at @xmath157 kpc in proper units , for four cases with @xmath158 of @xmath159 , @xmath160 , @xmath161 , @xmath162 .
the indicated neutral hydrogen column densities in figure 9 are approximately computed as follows .
combining equation ( 2 ) for @xmath163 at @xmath41 ( with the adopted fiducial values of various quantities ) with the result that the physical sizes of clouds go roughly as @xmath8 ( figure 4 ) gives @xmath164@xmath6 .
we see that , _ at a fixed column density _ , the rate of coincident lines ( at @xmath165 km / s ) _ increases with redshift at a moderate pace _ , and higher column density clouds and/or smaller @xmath139 yield weaker evolution , assuming that @xmath80 is _ constant _ over the redshift range in question .
the assumption of constant @xmath80 is merely for the convenience of illustration .
in fact , the self - consistently produced photoionization field during the simulation is constant from @xmath101 to @xmath102 to within a factor of two .
in other words , the results would not have significantly differed if the actual photoionization field had been used .
given the three dimensional nature of the simulations , we can distinguish between two kinds of coincident lines : one of which ( @xmath140 ) occurs when a pair shares the same 3-d parent cloud , the other of which ( @xmath166 ) occurs when the lines intersect two separate 3-d clouds .
the former set of coincident lines is due to the extended size of a single cloud , while the latter set of coincident lines is due to the clustering of several clouds .
figure 11 shows @xmath167 at @xmath41 .
six curves are shown for @xmath66 with @xmath151 km / s and @xmath152 km / s .
we see that at @xmath168kpc about half of the coincident line pairs share the same clouds and the other half pierce separate clouds . at @xmath157kpc only 10 - 20% of the total coincident line pairs
share the same clouds , i.e. , clustering of separate clouds dominates the coincident events at this and larger separations .
the observational signature of clustering dominated coincident lines is that the difference in line of sight velocity between the two absorption lines should be a weak function of the separation perpendicular to the line of sight .
figure 12 shows rms velocity difference of coincident line pairs along the line of sight , @xmath169 , as a function of the perpendicular separation , @xmath137 , for @xmath151 km / s and 150km / s at @xmath66 .
the two long - dashed horizontal lines show the rms difference for the two @xmath139 values , if clouds are small ( compared to perpendicular separation ) and randomly distributed ( i.e. , no clustering ) .
we see that @xmath169 depends rather weakly on @xmath137 when @xmath170kpc , and it approaches the asymptotic limit of a random distribution of clouds at @xmath171kpc .
we predict that for quasar double sightlines at large separations ( @xmath172kpc ) , the rms velocity difference between absorption features common to both sightlines , @xmath169 , should depend on the range being searched in @xmath84-space , @xmath139 , as @xmath173 figures ( 13a , b ) show the redshift evolution of @xmath174 for @xmath175@xmath6 and @xmath176@xmath6 , respectively .
four cases are shown in each figure with two separations , @xmath177kpc and @xmath157kpc both in proper units , and with @xmath151km / s and @xmath152km / s .
we see that the fraction of coincident lines with a separation of @xmath178 proper kpc due to common clouds _ decreases with increasing redshift at fixed column density _ , presumably because clouds are smaller and more crowded at high redshift than at low redshift .
we remind the reader that the total line coincident rate _ increases with redshift _ ( see figure 10 ) , indicates that a higher cloud number density ( crowding of clouds ) at high redshift dominates the line coincident events .
this is consistent with results shown in figure 11 .
the combined results of our quasar double sightline analysis suggest a few basic trends which should be kept in mind when analyzing qso spectra .
the most important of these is that line coincident events can be caused by two different phenomena . at small perpendicular separations ( @xmath179kpc proper )
a pair of coincident lines is most likely to pierce a common cloud , as is usually assumed when analyzing such observations to infer the actual cloud size .
however , at larger @xmath137 , a pair of coincident lines is more likely to penetrate two different clouds which are spatially clustered . at @xmath157 kpc ,
only 10 - 20% of the total coincident line pairs share the same clouds . and at very large separations of @xmath172kpc , the coincident line events can be explained as the random intersection of two unrelated clouds , whose rms difference @xmath169 should relate to the velocity interval @xmath139 as indicated in equation ( 7 ) .
it is tempting to make a connection between @xmath0 clouds and dwarf galaxies and/or moderate redshift faint blue galaxies .
we approach this problem by examining clustering properties of the @xmath0 clouds .
a few cautionary words about the limitations of the simulation are in order here . the simulation box size ( @xmath27mpc comoving ) places severe limits on our ability to study the clustering properties of clouds on large scales with a high degree of accuracy , because it sets an upper limit on the scale of the input power spectrum .
waves longer than the simulation box size would have made a considerable contribution to clustering on larger scales , since the density fluctuations on scales comparable to the simulation box length have started to approach nonlinearity even at redshift @xmath102 ( @xmath180 at @xmath102 on the box scale ) .
the situation becomes even more serious at lower redshift .
nevertheless , we may gain some insight into the distribution of mass on smaller scales , and say somthing useful about the more local clustering properties of clouds .
the three dimensional correlation of the @xmath0 clouds at @xmath41 is shown in figure 14 for @xmath181 .
higher column density clouds seem to be more strongly clustered than lower column density ones .
since most of the faint blue galaxies are thought to be in the redshift @xmath182 , we show in figure 15 the evolution of correlation length @xmath183 ( solid curve ) .
this is defined as the length where the correlation is unity and is plotted for the redshift range @xmath184 , where we have the most confidence in the accuracy of the simulation .
all lengths are shown in comoving units . in order to estimate the effect of missing power at box - sized scales , we computed the correlation function of the clouds in eight subboxes .
these subboxes were created by dividing the original simulation box into eight equal cubes and pretending that each subbox also has periodic boundary conditions .
the dotted line in figure 15 shows the median value of the correlation length for the subboxes , and the vertical bars indicate the full width of the distribution at each redshift ( i.e. , highest and lowest values of @xmath183 ) .
we see that the median correlation length drops about 30 - 40% for a box of size @xmath185mpc .
our best estimate is that the true correlation length at each redshift would be larger than values indicated by the solid curves by perhaps 50% , were the simulation box sufficiently large .
simple extrapolation of the solid curve in figure 15 to lower redshifts is unlikely to give us accurate values of @xmath183 at @xmath31 .
however , it is reasonable to expect the value of @xmath183 at these redshifts to be significantly larger than @xmath186mpc comoving , when one takes into account the effect of the missing longer waves .
this correlation length is interestingly close to the observed correlation length of the faint blue galaxies ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) , although an accurate conversion from observed angular correlation function @xmath187 to the 3-d correlation function as computed here requires detailed knowledge of the redshift distribution of the observed faint blue objects , which is currently unavailable .
furthermore , the average mass of these objects is close to @xmath188 ( see table 1 ) , which is in accord with what the mass of the faint blue galaxies is thought to be or required [ , e.g. , babul & rees ( 1992 ) starburst dwarf galaxy model ] .
finally , the number density of the clouds ( with a mean separation of @xmath189mpc at z=2 for @xmath68 ) and its increasing trend with redshift indicate that it is probably quite close to what is thought to be for observed faint blue objects .
these three considerations lend tentative support to the conjecture that the high redshift @xmath0 clouds ( @xmath5@xmath6 ) are the progenitors of the faint blue galaxies .
we have presented a quantitative study of the sizes , shapes , and correlations of @xmath0 clouds in a spatially flat cold dark matter universe with a cosmological constant , utilizing state - of - art cosmological hydrodynamic simulations including detailed atomic physics for a plasma of primordial composition .
keeping in mind the unavoidable limitations of such numerical experiments due to limited boxsize and limited resolution as well as some approximate treatment of physical processes , a few findings are probably fairly reliable .
structures formed at high redshift @xmath190 due to gravitational growth / collapse of cosmic density perturbations at small - to - intermediate scales ( 100kpc to a few mpc comoving ) are responsible for the observed @xmath0 forest .
these @xmath0 clouds can not be described by a simple model which characterizes their sizes and shapes .
their sizes vary in a wide range from a few kiloparsecs to about one hundred kiloparsecs in proper units with the median size being @xmath191kpc at @xmath41 , and their shapes vary from nearly - spherical ellipsoids to filaments and pancakes .
a typical @xmath0 cloud resembles a flattened cigar , " with an axial ratio of @xmath120 at column density of @xmath132 @xmath6 , and is either a thin cigar with an axial ratios of @xmath133 or near sphere with an axial ratios of @xmath134 at column density of @xmath135@xmath6 .
larger clouds are , on average , less spherical than smaller ones .
the physical size of a typical cloud grows with time roughly as @xmath8 while its shape hardly evolves ( except for the most dense regions @xmath9 where tend to be more spherical with time ) .
analysis of simulated quasar double sightlines indicates that coincident absorption features observed in the two spectra can have two different causes . at small perpendicular sightline separations ( @xmath179kpc proper ) , a pair of coincident lines most likely represents absorption from a common cloud
this is usually assumed when analyzing such observations to infer the actual cloud size .
however , at larger @xmath137 , a pair of coincident lines most likely samples two separate clouds which belong to the same cloud cluster . in other words ,
clustering of separate clouds dominates the coincident events at larger double sightline separations . at @xmath157kpc ,
80 - 90% of coincident line pairs can be explained in this fashion . at very large separations , @xmath172kpc ,
the coincident line events are entirely due to random intersections of two unrelated , uncorrelated clouds , whose rms velocity difference @xmath169 should relate to the velocity interval being examined , @xmath139 , as @xmath192 .
analyzing observed coincident absorption lines to infer the actual sizes of @xmath0 clouds by assuming a population of clouds with simplified geometry ( e.g. spheres , disks ) is reasonably accurate only when the double sightline separations are small ( @xmath10kpc ) . for larger
@xmath137 , this exercise is not very meaningful .
the inferred sizes are not related to true sizes of the clouds and are grossly inflated .
we attempt to make a connection between the high column density @xmath0 clouds ( @xmath5 ) and the faint blue galaxies .
the supportive evidence for such a conjecture is threefold .
first , when extrapolated to @xmath182 , the correlation of the high column density @xmath0 clouds ( @xmath193mpc comoving ) is close to that of the observed faint blue galaxies ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
second , the typical mass of these objects ( @xmath188 ) is close to that of a dwarf galaxy [ , e.g. , babul & rees ( 1992 ) scenario ] .
finally , the number density of the clouds is reasonably close to what is observed for faint blue objects .
finally , we caution that it may be necessary to extend the simulation s dynamic range to @xmath194 cells in each dimension , in order to have both a larger box ( @xmath195mpc comoving ) and a finer resolution ( @xmath196kpc comoving ) to ensure both a fair " sample and fully resolved structures of the objects in question , while maintaining a sufficient mass resolution .
the work is supported in part by grants nag5 - 2759 , ast91 - 08103 and asc93 - 18185 .
we thank the referee , george efstathiou , for a constructive and pertinent report which improved the quality of the work .
discussions with j.r .
gott , n. katz , g. lake , j. miralda - escud and j.p .
ostriker are gratefully acknowledged .
rc would like to thank g. lake and department of astronomy of university of washington for the warm hospitality , and financial support from the nasa hpcc / ess program during a visit when much of this work was completed .
dum adams , f. , bond , j.r . ,
freese , k. , frieman , j. & olinto , a. , 1993 , preprint babul , a. , & rees , m. 1992 , , 255 , 346 bahcall , n. , & cen , r. 1992 , , 407 , l49 bahcall , n.a .
, & cen , r. 1993 , , 407 , l49 bechtold , j. , crotts , a.p.s . ,
duncan , r.c . , fang , y. 1994 , , 437 , l83 bechtold , j. , & yee , h.k.c .
1996 , apj , in press bertschinger , e. , & gelb , j.m .
1991 , comp .
phys . , 5 , 164 bolte , m. , & hogan , c.j . 1995 , , 376 , 399 bond , j.r . ,
myers , 1996 , , brainerd , t.g . , smail , i , & mould , j. 1995 , preprint bucher , goldhaber , & turok , n. 1995 , preprint carswell , r. f. , lanzetta , k. m. , parnell , h. c. , & webb , j. k. , 1991 , , 371 , 36 cen , r. 1992 , , 78 , 341 cen , r. , & ostriker , j.p .
1993a , , 417 , 404 cen , r. , & ostriker , j.p .
1993b , , 417 , 415 cen , r. , & ostriker , j.p .
1994 , , 431 , 451 cen , r. , gnedin , n.y . , &
ostriker , j.p .
1993 , , 417 , 387 cen , r. , gnedin , n.y . ,
kofman , l.a . , & ostriker , j.p .
1992 , , 399 , l11 cen , r. , ostriker , j.p . , &
peebles , p.j.e .
1993 , , 415 , 423 cen , r. , miralda - escud , j. , ostriker , j. p. , & rauch , m. 1994 , , 437 , l9 couch , w.j .
, jurcevic , j.s . , & boyle , b.j .
1993 , , 260 , 241 cowie , l.l . , lilly , s.j . ,
gardner , j.p . , & mclean , i.s .
1988 , , 332 , l29 cristiani , s. , dodorico , s. , fontana , a. , giallongo , e. , & savaglio , s. 1995 , , 273 , 1016 davis , m. , summers , f.j . , & schlegel , d. 1992 , , 359 , 393 dinshaw , n. , impey , c.d . ,
foltz , c.b . ,
weymann , r.j . , & chaffee , f.h .
1994 , , 437 , l87 dinshaw , n. , foltz , c.d . , impey , c.d . ,
weymann , r.j . , & morris , s.l .
1995 , , 373 , 223 edelson , & malkan 1986 , , 308 , 59 efstathiou , g. , bond , j.r .
, & white , s.d.m .
1992 , , 258 , 1p efstathiou , g. , bernstein , g. , tyson , j.a . , katz , n. , & guhathakurta , p. 1991
, , 380 , l47 eke , v.r . , cole , s. , & frenk , c.s .
1996 , preprint fang , y. , dunca , r.c . ,
crotts , a.p.s . , & bechtold , j. 1996 , preprint feldman , h.a . ,
kaiser , n. , & peacock , j.a . 1994 , , 437 , 56 freedman , m.l . , 1994 , , 371 , 757 gott , j.r . ,
iii 1982 , , 295 , 304 hamuy , m. , 1995 , , 109 , 1 hernquist , l. , katz , n. , & weinberg , d.h .
1996 , , 457 , l51 hu , e. m. , kim , t .- s . , cowie , l. l. , songaila , a. , & rauch , m. 1995 , , 110 , 1526 infante , l. , & pritchet , c.j .
1995 , , 439 , 565 klypin , a. , holtzman , j. , primack , j. , & regos , e. 1993 , , 416 , 1 kofman , l.a , gnedin , n.y . , &
bahcall , n.a .
1993 , , 413 , 1 koo , d. 1986 , , 311 , 651 liddle , a.r .
, lyth , d.h . , & sutherland , w. 1992 , phys .
b 279 , 244 lidsey , j.e . , coles , p. 1992
, , 258 , 57p lucchin , f. , matarrese , s. , & mollerach , s. , 1993 , submitted to apj(letters ) ma , c.p .
, & bertschinger , e. 1994 , , 435 , l9 miralda - escud , j. , cen , r. , ostriker , j. p. , & rauch , m. 1996 , , in press neuschaefer , l.w . , windhorst , r.a . , & dressler , a. 1991 , , 382 , 32 ostriker , j.p . , & steinhardt , p. 1995
, , 377 , 600 oukbir , j. , & blanchard , a. 1992 , a & a , 262 , l21 peacock , j.a . , & dodds , s.j .
1994 , , 267 , 1020 peebles , p.j.e .
1984 , apj , 284 , 439 peeles , p.j.e 1987a , , 327 , 210 peeles , p.j.e 1987b , , 315 , l73 pen , u .- l .
1996 , in preparation petitjean , p. , webb , j. k. , rauch .
m. , carswell , r. f. , & lanzetta , k. , 1993 , , 262 , 499 rauch , m. , carswell , r. f , chaffee , f. h. , foltz , c. b. , webb , j. k. , weymann , r. j. , bechtold , j. , & green , r. f. 1992 , , 390 , 387 riess , a.g . , press , w.h .
, & kirshner , r.p .
1995 , , 438 , l17 roche , n. , shanks , t. , metcalfe , n. , & fong , r. 1993 , , 263 , 360 scalo , j.m .
1986 , fund . cosmic phys . , 11 , 1 schneider , d.p . , et al . 1993 , , 87 , 45 smette , a. , surdej , j. , shaver , p.a . , foltz , c.b . ,
chaffee , f.h . ,
weymann , r.j . , williams , r.e . ,
& magain , p. 1992
, , 389 , 39 smette , a. , robertson , j.g . , shaver , p.a . , reimers , d. , wisotzki , l. , & kohler , t.h .
1995 , in press smoot , g.f .
, 1992 , , 396 , l1 taylor , a.n . , &
rowan - robinson , m. 1992 , , 359 , 396 tyson , j.a .
1988 , aj , 96 , 1 tytler , d. , fan , x .-
burles , s. 1996 , , 381 , 207 tytler , d. , fan , x .- m . , burles , s. , cottrell , l. , davis , c. , kirkman , d. , & zuo , l. 1995 , in _ qso absorption lines _ , proc .
eso workshop , ed .
g. meylan ( heidelberg : springer ) , p. 289 viana , p.t.p , & liddle , a.r .
1995 , preprint walker , t.p . ,
steigman , g. , schramm , d.n . , olive , k.a . , &
kang , h.s .
1991 , , 376 , 51 white , s.d.m . , efstathiou , g. , & frenk , c.s .
1993 , , 262 , 1023 white , s.d.m . ,
navarro , j.f .
, evrard , a.e . , & frenk , c.s .
1993 , , 366 , 429 zhang , y. , anninos , p. , & norman , m.l .
1995 , , 453 , l57 @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 @xmath203 & @xmath198 & @xmath204 & @xmath205 & @xmath206 & @xmath207 @xmath208 & @xmath198 & @xmath209 & @xmath210 & @xmath211 & @xmath212 @xmath197 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath217 @xmath203 & @xmath213 & @xmath218 & @xmath219 & @xmath220 & @xmath207 @xmath208 & @xmath213 & @xmath221 & @xmath222 & @xmath223 & @xmath224 @xmath197 & @xmath225 & @xmath226 & @xmath211 & @xmath227 & @xmath228 @xmath203 & @xmath225 & @xmath229 & @xmath230 & @xmath231 & @xmath232 @xmath208 & @xmath225 & @xmath233 & @xmath234 & @xmath235 & @xmath236 | this study analyzes the sizes , shapes and correlations of @xmath0 clouds produced by a hydrodynamic simulation of a spatially flat cdm universe with a non - zero cosmological constant ( @xmath1 , @xmath2 , @xmath3 ) , over the redshift range @xmath4 .
the @xmath0 clouds range in size from several kiloparsecs to about a hundred kiloparsecs in proper units , and they range in shape from roundish , high column density regions with @xmath5 @xmath6 to low column density sheet - like structures with @xmath7 @xmath6 at z=3 .
the most common shape found in the simulation resembles that of a flattened cigar .
the physical size of a typical cloud grows with time roughly as @xmath8 while its shape hardly evolves ( except for the most dense regions @xmath9 ) .
our result indicates that any simple model with a population of spheres ( or other shapes ) of a uniform size is oversimplified ; if such a model agrees with observational evidence , it is probably only by coincidence .
we also illustrate why the use of double quasar sightlines to set lower limits on cloud sizes is useful only when the perpendicular sightline separation is small ( @xmath10kpc ) .
finally , we conjecture that high column density @xmath0 clouds ( @xmath5@xmath6 ) may be the progenitors of the lower redshift faint blue galaxies .
this seems plausible because their correlation length , number density ( extrapolated to lower redshift ) and their masses are in fair agreement with those observed .
# 1 # 1 , email : ( cen,rasimcoe)@astro.princeton.edu |
low surface brightness galaxies ( lsbgs ) were theorized to exist by @xcite and then verified by the discovery of malin 1 @xcite .
they refer to galaxies which have lower surface brightness than the background @xcite and are traditionally defined as galaxies with central surface brightness @xmath5 fainter than some threshold which varies between 22.0 and 23.0 mag arcsec@xmath6 ( e.g. , * ? ? ?
* ; * ? ? ?
recent studies indicate that lsbgs contribute 20% of the dynamical mass of galaxies in the universe @xcite and 30%60% of the number density of local galaxies @xcite , suggesting that the contribution of lsbgs to the universe can not be negligible . however , selection effects cause lsbgs to be underrepresented in the present - day galaxy catalogs . compared with disk galaxies having high surface brightness ( hsb ) ,
lsbgs generally tend to be dust - poor ( e.g. , * ? ? ?
* ) and h -rich , with diffuse , low - density stellar disks and metallicity @xmath7 solar abundance ( e.g. , * ? ? ?
this suggests that lsbgs have been inefficient star formers over their lifetimes and have low - level star - formation activity at present ( e.g. , * ? ? ?
* ; * ? ? ?
indeed , h surface densities in lsbgs are often observed to be well below the critical threshold for star formation ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
nonetheless , lsbgs typically contain some signatures of ongoing star formation , including modest amounts of h@xmath0 emission from a very small number of h regions ( e.g. , * ? ? ?
* ; * ? ? ?
* ) and blue colors indicative of young stellar populations ( e.g. , * ? ?
* ; * ? ? ?
metallicities of stars and gas within a galaxy provide both a fossil record of its star formation history ( sfh ) and evolution .
for galaxies , metallicities are traditionally represented by the oxygen abundance , 12 + log(o / h ) , which directly reflects the metal content of the gas in galaxies and indirectly reveals the metal content of the stars in galaxies that ionize the gas . for lsbgs
, many previous works have demonstrated that lsbgs are metal - poor .
the lsbg sample of @xcite has been measured to have low oxygen abundances , 12 + log(o / h ) @xmath8 dex .
@xcite found a mean value 12 + log(o / h ) = 8.22 dex for 17 lsbgs .
@xcite studied kkr17 , an lsbg dominated by h , and measured an average 12 + log(o / h ) = 8.0 dex .
the oxygen abundances of @xcite are 12 + log(o / h ) @xmath98.2 dex .
apparently , most lsbgs have 12 + log(o / h ) lower than the solar value of oxygen abundance of 12 + log(o / h ) = 8.87 dex @xcite , and even lower than the transition value ( 12 + log ( o / h ) = 8.4 dex ) between metal - poor and metal - rich branches for the @xmath1 method ( see 4.3 ) . to derive gas - phase metallicities of lsbgs
, spectra should be required .
however , it is not easy to derive spectra of lsbgs due to their low surface brightness .
fortunately , we have carried out spectroscopic observations for a sample of lsbgs . later in this paper
, we will introduce this observation and then study properties which are mainly relevant to metallicities on basis of the observational data . given the faint nature of lsbgs , with peak surface brightness @xmath10 mag arcsec@xmath6 ,
it is extremely difficult to obtain spectra of their stellar continua with relatively high signal - to - noise ratio ( @xmath11 ) because the sky background always dominates the signal .
studies of the stellar populations of lsbgs are currently underrepresented and are based mainly on spectra of their h regions ( e.g. , * ? ? ?
* ) and central portions ( e.g. , * ? ? ?
but the bright h regions contain a large fraction of o- and b - type stars , so stellar populations studied through spectra of the h regions would be biased to younger ages . compared to observing face - on lsbgs ( * ? ? ?
* e.g. , ) or fiber spectroscopy , the spectral qualities can be highly improved by observing edge - on lsbgs , especially those having thin disks , with a single slit of appropriate length and width , positioned along the major axis .
this scheme makes it possible , although still very difficult , to obtain reasonable - quality data .
very fortunately , spectra of several few lsbgs of our sample have relatively reliable stellar continua which can be used for stellar population studies , so we additionally study stellar populations of these few lsbgs and to investigate their star formation histories as byproducts of our observations ( 6.1 ) .
the outline of this paper is as follows .
section 2 describes our sample , and detailed accounts of the observations and data reduction are given in 3 and 4 .
we study the physical properties of our edge - on lsbgs in 5 , including oxygen abundances in 5.1 , the luminosity - metallicity relation in 5.2 , the mass - metallicity relation in 5.3 , and the behavior of gas mass fraction versus metallicity in 5.4 . in section 6 ,
we discuss stellar populations in 6.1 , stellar mass estimation in 6.2 , chemical evolution in 6.3 , and the true nature of our lsbg sample in 6.4 .
finally , we conclude in 7 .
a catalog of edge - on disk galaxies in sdss dr7 ( egis ; * ? ? ?
* ) consists of 5747 genuine edge - on galaxies in the sloan digital sky survey ( sdss ; * ? ? ?
* ) . by analyzing photometric profiles in each of the @xmath12- , @xmath13- , and @xmath14-band images ,
the egis catalog provides estimates for structural parameters of the stellar disks of the galaxies , such as the disk scale height ( @xmath15 ) , radial scale length ( @xmath16 ) , and central surface brightness ( @xmath17 ) . to further study the egis catalog in combination with h data , we correlated it with the @xmath0.40 catalog @xcite , which lists h line sources from 40% of the area of the final blind 21 cm h alfalfa ( arecibo legacy fast arecibo l - band feed array ) survey for local galaxies . using the transformation formula of @xcite
, we derived the central surface brightness in the @xmath18 band , @xmath19 , from those in the @xmath12 and @xmath13 bands .
then , according to a widely used definition of a lsbg , there are 287 galaxies in the [email protected] sample that can be identified as lsbgs with @xmath20 mag arcsec@xmath6 . out of the [email protected] lsbg sample , only 21 galaxies ( blue filled circles in figure [ fig : target_info](a),(b ) )
are in the fall sky region ( @xmath21 ) which matched our available observing time . of the 21 galaxies , 14 galaxies
have @xmath22 mag ( dashed black line in figure [ fig : target_info ] ) , sufficiently blue to be considered as the final targets for our 3 nights in late - september 2015 .
unfortunately , only 50% of the total time was useful for observations because of stormy and cloudy conditions ; we ultimately observed 9 galaxies . comparing with the 14 prepared targets ( magenta lines ) , the observed 9 galaxies ( red ) are generally consistent in distributions of @xmath19 , magnitude , color , and @xmath23 as shown in figure [ fig : target_info](c)(f ) .
furthermore , comparing with the entire [email protected] lsbg sample ( black ) , these 9 galaxies ( red ) sample almost the full range in @xmath19 from 22.5 to 24.5 mag arcsec@xmath6 , but they sample only the fainter part in @xmath13 ( @xmath24 mag ) and only the blue part in @xmath25 ( @xmath26 mag ) . on the other hand , in february 2015 , we had already spectroscopically observed in a similar way several galaxies ( not confined to be lsbgs ) in the [email protected] catalog , but with a different telescope using spare time from another program . from these objects , there are 3 galaxies with @xmath20 mag arcsec@xmath6 .
we overplot them ( blue ) in figure [ fig : target_info ] and find that they are comparable to our observed ( fall 2015 ) 9 lsbgs in distributions of main properties .
so , we add these 3 lsbgs into our 9-galaxy lsbg sample in order to enlarge the overall observed sample in this paper .
we display the sdss images of the 12 edge - on lsbgs in figure [ fig : edgeon - slit ] and list their main properties in table [ tbl-1 ] .
statistically , @xmath19 values of the 12 lsbgs range from 22.6 to 24.4 mag arcsec@xmath6 .
there is one object ( agc 228010 ; see figure [ fig : edgeon - slit ] ) with @xmath27 ; it is a super - thin disk galaxy , previously defined to have an axial ratio between the disk scale length ( @xmath15 ) and the scale height ( @xmath16 ) in the @xmath12 band larger than 10.0 @xcite . additionally , our 12 galaxies have little ( or no ) optical bulges or central light concentrations , with @xmath28 in the sdss @xmath13 band , but they are not the only ones with @xmath28 ( see figure [ fig : target_info](f ) ) .
ccrrccrrrc 191770&145.940&14.680&3823&57.4&23.27&17.21&16.88&6.04&1.60 + & 149.97&0.06&9.07&165&1.50 + 228010&194.909&6.446&6354&95.3&24.31&17.90&17.63&10.39&0.96 + & 41.12&0.04&9.45&165&1.32 + 253926&227.023&6.863&9333&135.9&23.33&17.05&16.46&6.55&1.55 + & 32.85&0.08&10.26&264&4.19 + 12249&343.834&28.346&7549&102.9&23.59&16.08&15.52&11.87&1.93 + & 53.80&0.05&9.73&295&2.15 + 321228&344.597&26.028&7492&102.1&22.84&16.92&16.39&5.92&1.71 + & 42.00&0.06&9.26&104&0.75 + 332804&354.608&15.941&12425&172.3&23.16&16.91&16.43&6.85&1.65 + & 80.07&0.06&9.99&266&1.40 + 332806&354.807&15.697&12055&167.0&22.88&17.33&17.04&4.24&1.05 + & 95.95&0.05&9.83&242&1.03 + 100811&3.981&16.004&8362&114.4&23.67&16.38&15.88&10.44&2.12 + & 57.34&0.06&9.77&239&1.89 + 123165&39.174&28.784&4536&61.9&22.68&16.1500&15.65&7.68&2.83 + & 157.24&0.07&8.95&205&0.99 + 101807&14.957&14.328&11916&165.6&24.40&17.13&16.63&9.37&1.43 + & 175.54&0.04&10.1&308&1.97 + 1181&25.125&14.523&8133&112.0&23.31&16.34&15.89&10.19&2.20 + & 130.95&0.06&9.93&292&2.89 + 748835&31.840&15.429&12793&178.9&23.37&17.69&17.10&5.41&1.20 + & 55.53&0.04&9.96&266&1.22 +
long - slit spectroscopy of three lsbgs in the spring sky were obtained on 2015 feb . 18 ut with the kast double spectrograph @xcite at the cassegrain focus of the 3-m shane telescope at lick observatory .
a long slit 2@xmath29 wide was positioned along the major axis of each target , and the seeing was @xmath30 . for the blue side , a 452 lines mm@xmath31 grating blazed at 3306 was used in first order , yielding a wavelength range of 31506300 with a spectral resolution of 1.41 pixel@xmath31 .
use of a dichroic allows simultaneous observations at longer wavelengths with the red side ; however , lick spectroscopy was obtained only with the blue side .
the other nine lsbgs were observed on 2015 sep .
21 ut ( seeing @xmath32 ) and sep .
23 ut ( seeing @xmath33 ) with the 5-m hale telescope at palomar observatory .
we used the double spectrograph ( dbsp ; * ? ? ? * ) and a @xmath34-long slit of width 2@xmath29 , positioned along the major axis of each target .
a d68 dichroic was used to divide the wavelength coverage into two regions , 36716727 for the blue side ( 600 lines mm@xmath31 grating blazed at 4000 ) and 62239577 for the red side ( 600 lines mm@xmath31 grating blazed at 10,000 ) .
the respective spectral resolutions are 1.078 pixel@xmath31 and 0.819 pixel@xmath31 , so both gratings resolve all possible line features of interest to us .
the total exposure times at lick and palomar for a galaxy were 36007200 s , subdivided into 24 subexposures in order to remove cosmic rays and enhance the signal - to - noise ratio ( s / n ) of the combined spectra .
see table [ tbl-2 ] for a log of observations .
the atmosphere is a dispersive medium . at nonzero zenith angles ,
a stellar image will be spread out into a small spectrum , and this spread can be significant compared to the size of the seeing disk . under these conditions , if the spectrograph slit is not oriented along the direction of atmospheric dispersion ( the parallactic angle ) , starlight at some wavelengths will not enter the narrow slit , while starlight at other wavelengths will pass into the instrument relatively unhindered . during our observations
, the slit was actually oriented along the major axis of the target object ( position angle ) instead of the parallactic angle , and the object was centered in the slit at @xmath35 . therefore , at nonzero zenith angles , object light from either wavelength end ( blue or red ) may be selectively lost . for example , as seen in table [ tbl-2 ] , the spectrum of agc 191770 at airmass 1.08 was obtained with a 128@xmath36 2@xmath29 slit on a night when the seeing was 1.5@xmath29 .
then , following the calculations of @xcite , if agc 191770 was centered in the slit at a wavelength of 6000 , it would be off center by 0.61@xmath29 at 3600 @xmath37 and by 0.05@xmath29 in the opposite direction at 6500 @xmath38 .
the parallactic angle was @xmath39 east of north , but the slit was actually aligned along a position angle of @xmath40 east of north .
so , if agc 191770 was a star with a gaussian point - spread function , the percentage of light entering the slit at 3600 @xmath41 , 6000 @xmath42 , and 6500 @xmath43 would be ( respectively ) 83.5 , 88.3 , and 88.3 .
thus , the light loss at 3600 would be 4.8% greater than that at 6000 .
differential losses are a bigger concern for point sources than for extended objects such as galaxies . for extended sources ,
especially if the surface brightness is roughly constant over an area larger than the slit width , blue - end light loss from the position being observed can be partly gained back from neighboring positions within the source . the effect on the relative flux calibration
will not be very large for our targets because of the following reasons .
( 1 ) most spectra of our targets were observed at small airmasses ( table [ tbl-2 ] ) . ( 2 ) our targets are edge - on galaxies with extended areas larger than the slit width ( see figure [ fig : edgeon - slit ] ) , so the blue and red light losses are partly gained back from the neighboring positions within the objects . for completeness , we do the calculations above for all our targets ( at the midpoint of the exposures ) and list the results in table [ tbl-2 ] .
cccccccc 191770&2015 - 02 - 18 & @xmath44 & 34506450 & & shane kast & 1.080 & 149.97 + & 3.05 & 0.61 & 0.00 & -0.05 & 83.5 & 88.3 & 88.3 + 228010&2015 - 02 - 18 & @xmath45 & 34506450 & & shane kast & 1.170 & 41.12 + & 143.88 & 0.80 & 0.00 & -0.06 & 63.0 & 88.3 & 88.2 + 253926&2015 - 02 - 18 & @xmath46 & 34506450 & & shane kast & 1.170 & 32.85 + & 24.11 & 0.80 & 0.00 & -0.06 & 87.7 & 88.3 & 88.3 + 12249 & 2015 - 09 - 21 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.239 & 53.80 + & 108.52 & 0.91 & 0.00 & -0.08 & 65.8 & 95.0 & 94.9 + 321228&2015 - 09 - 21 & @xmath48 & 36716727 & 62239577 & hale dbsp & 1.027 & 42.00 + & 125.96 & 0.97 & 0.00 & -0.08 & 52.7 & 95.0 & 94.8 + 332804&2015 - 09 - 21 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.048 & 80.07 + & 177.59 & 0.42 & 0.00 & -0.03 & 87.0 & 95.0 & 94.9 + 332806&2015 - 09 - 21 & @xmath48 & 36716727 & 62239577 & hale dbsp & 1.055 & 95.95 + & 18.51&0.47 & 0.00 & -0.04 & 85.6 & 95.0 & 94.9 + 100811&2015 - 09 - 21 & @xmath48 & 36716727 & 62239577 & hale dbsp & 1.117 & 57.34 + & 45.41 & 0.67 & 0.00 & -0.05 & 94.2 & 95.0 & 95.0 + 123165&2015 - 09 - 21 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.016 & 157.24 + & 61.13 & 0.27 & 0.00 & -0.02 & 91.9 & 95.0 & 95.0 + 101807&2015 - 09 - 23 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.057 & 175.54 + & 0.38 & 0.47 & 0.00 & -0.04 & 94.9 & 95.0 & 95.0 + 1181 & 2015 - 09 - 23 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.062 & 130.95 + & 16.35 & 0.47 & 0.00 & -0.04 & 86.7 & 95.0 & 94.9 + 748835&2015 - 09 - 23 & @xmath47 & 36716727 & 62239577 & hale dbsp & 1.088 & 55.53 + & 37.02 & 0.57 & 0.00 & -0.04 & 93.6 & 95.0 & 95.0 +
data reduction was carried out using standard iraf procedures . after correction for overscan , bias , and flat field
, we reject cosmic rays using the iraf task l.a.cosmic @xcite , which identifies and removes cosmic rays robustly from astronomical images via a laplacian algorithm .
then , we performed wavelength calibration on the two - dimensional ( 2d ) spectral images of objects and standard stars .
sky subtraction will be described more detailedly in next subsection .
after sky subtraction , we defined the appropriate aperture , traced the flux peaks along the dispersion direction , and interactively extracted one - dimensional ( 1d ) spectrum by using the apall task of iraf .
finally , we averagely combined spectra from multiple exposures for an object into one final spectrum . in figure
[ fig : spectra2 ] , we show the combined blue - side spectra for the three galaxies observed with the lick shane telescope ( they have no red - side observations ) . for the nine galaxies observed with the palomar hale telescope , both the blue- and red - side spectra are shown ( with blue and red colors , respectively ) , after scaling the red side to match the flux level of the blue side in the small overlap region .
strong emission lines are marked .
generally , the night - sky lines are not clearly subtracted in the faint lsbg spectra , so we simply mask four wavelength regions having the strongest night - sky lines ( green ) .
sky subtraction quality is very important for lsbgs as they are naturally fainter than the sky background in surface brightness .
so , we describe the three methods of sky subtraction we have tried in data reduction for our observations below . the first one is called the `` median - filter method . '' before performing this method , we make the spatial and dispersion directions perpendicular to each other through rectification , such as by performing 2d wavelength calibration on spectral images of objects .
then , on the rectified images , assuming the direction of night - sky lines is along the column , we perform a median filter represented by a @xmath49 array along each column to generate a sky image . here , @xmath50 must be larger than at least two times the width ( major axis in pixels ) of the object projected on the images .
this method is good at removing sky lines but may overestimate the sky continuum blended in the object .
the second one is the sky - fitting method adopted in the apall task of iraf , which uses a second - order polynomial to fit the background from the substantial sky region along the slit .
this method can do well in estimating sky continuum blended in the object but does not do very well in removing sky lines .
the third one is the @xcite method which makes full use of the raw image data ( before the optical distortions and spectral line curvatures have been rectified ) to model the sky background .
this method is not very good for our faint objects , because it either underestimates the sky continuum or leaves sky - line residuals in our images .
none of the three methods above is sufficiently good at removing both sky continuum and sky lines clearly from spectra of our faint lsbgs .
however , we have no other good ideas to implement it ideally .
so , for lick data which have no obvious distortions of raw images , the first median - filter " method has been used to simultaneously subtract the sky lines and sky continuum .
however , for hale data which have relatively serious distortions of raw images , this median - filter " method does not do well in sky continuum subtraction , so we finally used the second method to improve the sky continuum subtraction a little better , but we warn the reader that the final spectral continuum shapes for hale spectra are still probably not accurate . examining the extracted spectra
, it appears that the reductions of the hale spectra are not good at wavelengths @xmath51 .
unexpectedly , many absorption lines appear in the observed spectra of our standard stars in this blue - end region due to unresolved reasons , so we are not able to determine an accurate continuum shape for fitting the flux - calibration curve in this blue - end region .
fortunately , except for agc 123165 , our lsbg galaxies all have redshift @xmath52 , so the bluest useful line ( [ o ] @xmath533727 ) falls at @xmath54 .
although we have extracted the red - side palomar hale spectra , their continuum s / n is quite low , so we abandoned using the red - side spectral continuum .
however , the emission lines in the red spectra , such as [ n ] @xmath536584 , h@xmath0 , and [ s ] @xmath556716 , 6731 , are reasonably strong , and they could therefore be used for dust - extinction corrections and oxygen - abundance measurements . moreover , as mentioned above , the spectral continuum shapes of the nine galaxies observed with the hale telescope are not reliable , so we will not use them to draw any firm conclusions .
the gas - phase metallicity ( oxygen abundance , @xmath56 log(o / h ) ) of a galaxy is a powerful probe of the past star formation and growth .
it can also constrain the current stage of galaxy evolution as it is intimately connected to the fueling and regulation of star formation .
usually , flux ratios of strong optical emission lines are used as the abundance - sensitive diagnostics , such as @xmath1 , [ n ] /[o ] , [ n ] /h@xmath0 , [ n ] /[s ] , and so on @xcite . based
on extracted spectra of our lsbgs , we measure the observed fluxes of strong emission lines in the blue - side ( [ o ] @xmath533727 , h@xmath57 , [ o ] @xmath554959 , 5007 ) and red - side ( [ n ] @xmath556548 , 6584 , h@xmath0 , [ s ] @xmath556716 , 6731 ) . an intrinsic h@xmath0/h@xmath57 flux ratio of 2.86 is assumed for case b recombination at @xmath58 k @xcite because the relative intensities of the hydrogen balmer lines are nearly independent of density and temperature .
therefore , for observed line fluxes measured from the 9 hale spectra which have both blue- and red - side , we use the balmer decrement ( h@xmath0/h@xmath57 ) and the ccm extinction law with @xmath59 @xcite to correct dust extinction . based on these intrinsic fluxes ,
we compute several useful flux ratios of [ n ] /[o ] , [ n ] /[o ] ,
[ n ] /h@xmath0 , and @xmath1 [ where @xmath60 ( [ o ] @xmath533727 + [ o ] @xmath554959 , 5007)/h@xmath57 ] , among which [ n ] /h@xmath0 and
@xmath1 are individually used as metallicity diagnostics .
for the [ n ] /h@xmath0 diagnostic , both the pp04 @xcite and d02 @xcite calibrations are used to measure the oxygen abundances . for the @xmath1 diagnostic , both the m91 @xcite and
z94 @xcite calibrations are employed to measure the oxygen abundances .
thus , we have four sets of oxygen abundances for each of the 9 hale - observed galaxies except for agc 123165 which have no @xmath1based metallicities because its spectra has some problem in the blue - end wavelength range covering [ o ] @xmath533727 ( see 4.3 ) .
it should be noted that when using the m91 calibration , we need to determine which branch is appropriate for each galaxy because the @xmath1 versus o / h relation is double - valued @xcite . distinguishing a branch ( high-@xmath62 or low-@xmath62 ) can be accomplished with the [ n ] @xmath536584 line : strong ( weak ) nitrogen emission indicates that the high-@xmath62 ( low-@xmath62 ) branch is appropriate .
the dividing line is at log([n ] /[o ] ) @xmath63 @xcite .
however , for observed line fluxes measured from the 3 lick spectra which have no observed red - side spectra ( neither h@xmath0 nor [ n ] ) , no reddening correction is performed and only @xmath1 diagnostic and @xmath1based metallicities are measured . when using the m91 calibration , no available log([n ] /[o
] ) can be used for distinguishing branches , so we empirically adopt the low-@xmath62 branch to measure metallicities as lsbgs are generally acknowledged to be metal - poor . in table [ tbl-3 ] , we report measurement results above and also show the average ( @xmath64 ) of both @xmath1-based metallicities ( @xmath65 and @xmath66 ) and the average ( @xmath67 ) of both [ n ] /h@xmath0-based metallicities ( @xmath68 and @xmath69 ) for each lsbg . in statistic , our measurements show that the @xmath1-based metallicities ( @xmath64 ) are 8.008.62 dex , with a mean of 8.30 dex and a median of 8.26 dex .
the [ n ] /h@xmath0-based metallicities ( @xmath67 ) are 7.678.54 dex , with a mean of 8.20 dex and a median of 8.26 dex .
generally , the two sets of metallicities are consistent with each other and are both subsolar , with the @xmath70 log(o / h ) value being 8.87 dex for the sun @xcite .
our metallicities for the lsbg sample agree with the results for lsbgs from @xcite ( see figure [ fig : lumi_metal ] ) .
ccccccccccc 191770 & & & 5.93 & & 8.04 & 8.68 & & & 8.36 & + 228010 & & & 6.64 & & 7.80 & 8.59 & & & 8.195 & + 253926 & & & 4.16 & & 7.78 & 8.89 & & & 8.335 & + 12249 & 3.42 & -0.67 & 5.98 & -0.70 & 8.56 & 8.67 & 8.46 & 8.61 & 8.615 & 8.535 + 321228 & 4.67 & -1.10 & 11.44&-1.13 & 8.00 & 8.01 & 8.22 & 8.30 & 8.005 & 8.26 + 332804 & 5.79 & -1.30 & 13.00 & -1.36 &
8.79 & 7.85 & 8.14 & 8.13 & 8.32 & 8.135 + 332806 & 3.53 & -0.98 & 7.63 & -2.07 & 8.53 & 8.47 & 7.72 & 7.61 & 8.5 & 7.665 + 100811 & 4.19 & -0.92 & 9.37 & -0.94 & 8.22 & 8.27 & 8.31 & 8.44 & 8.245 & 8.375 + 123165 & 6.24 & & & -1.08 & & & 8.24 & 8.33 & & 8.285 + 101807 & 4.55 & -1.05 & 9.35 & -1.51 & 8.24 & 8.27 & 8.08 & 8.02 & 8.255 & 8.05 + 1181 & 5.30 & -1.01 & 9.89 & -1.14 & 8.15 & 8.21 & 8.22 & 8.29 & 8.18 & 8.255 + 748835 & 4.76 & -1.04 & 9.37 & -1.18 & 8.20 & 8.27 & 8.20 & 8.26 & 8.235 & 8.23 + the strong correlation between galaxy luminosity ( @xmath71 ) and galaxy metallicity ( @xmath62 ) is one of the most significant phenomenological results in chemical evolution studies of galaxies .
previous work found that the correlations between @xmath71 and @xmath62 for irregular galaxies @xcite , disk galaxies ( predominantly high surface brightness galaxies , hsbgs ; @xcite ) , and elliptical galaxies @xcite are very similar .
thus , it is instructive to plot our lsbg sample onto the familiar @xmath71@xmath62 relation defined by previous work on other galaxy types to validate the consistency of our measurements of lsbgs . in the @xmath71@xmath62 plot in figure [ fig : lumi_metal ] , we show both @xmath1-based ( red filled circles ) and the [ n ] /h@xmath0-based ( black filled circles ) metallicities for our lsbgs . here , the @xmath18-band absolute magnitude for our lsbgs are computed from the galaxy distances ( provided by @xmath7240 catalog ) and the @xmath18-band apparent magnitudes ( which are transformed from magnitudes of @xmath12- and @xmath13-bands measured by our own surface photometry @xcite ) .
data of lsbgs from @xcite ( purple triangles ) and dwarf irregulars ( dirrs ; dark - green squares ) from @xcite are also plotted in this @xmath71@xmath62 figure for comparisons . besides
, we show a variety of recent fits to the @xmath71@xmath62 relation determined respectively for the dirr sample of @xcite ( green line ) , a large sample of sdss star - forming galaxies from @xcite ( orange line ) , a large sample of starbursting emission - line galaxies drawn from the kiss ( kpno international spectroscopic survey ) sample from @xcite ( cyan line ) , and a large sample of 2dfgrs ( two - degree field galaxy redshift survey ) star - forming galaxies from @xcite ( blue line ) from the literature . among these relations
, it seems that there is some disparities , which are probably caused by the utilization of different samples and different diagnostics and calibrations for measuring abundances .
it is not easy to constrain a quantitatively reliable @xmath71@xmath62 relation for our lsbg data ; they do not have sufficient leverage because of their small coverage in luminosity space .
as shown in figure [ fig : lumi_metal ] , if all 12 galaxies ( red filled circles ) are considered , our sample covers only @xmath73 mag of absolute magnitude , and if only the 9 galaxies ( black filled circles ) having both blue- and red - side spectra are considered , it covers only @xmath74 mag . instead of being quantitative ,
we thus carry out our studies qualitatively . from figure
[ fig : lumi_metal ] , it appears that our lsbg sample with both sets of metallicities , especially with the [ n ] /h@xmath0based metallicities , seems to generally follow the similar distribution as the lsbgs from @xcite in their common covering region in luminosity and metallicity space .
both lsbg samples are consistent with the general distribution trend of other samples ( normal galaxies , dirrs ) in this @xmath71@xmath62 plot , although with some degree of scatter .
additionally , we checked for quantitative consistency between our sample and the lsbg sample from @xcite .
we did the test 50 times .
each time , we randomly picked a subsample of size equal to our small lsbg sample with [ n ] /h@xmath0-based metallicities from the lsbg sample of @xcite , and then checked to see whether this subsample gives a distribution consistent with our small lsbg sample by using a k - s test .
our criteria for each k - s test to decide whether the subsample is consistent with being drawn from the same distribution as our small lsbg sample is the probability value being larger than 0.90 ( [email protected] ) .
the result is that in the total of 50 trials , 32% of the subsamples from the @xcite lsbg sample give distributions that are consistent with our limited lsbg sample in the @xmath71@xmath62 plot .
any fraction more than 10% and less than 90% is basically inconclusive .
so , such a fraction shows that the samples are too small to provide a conclusive result .
if our sample could be improved to a larger size and wider luminosity coverage , such quantitative tests might give a more conclusive statistical fraction for consistency .
for our lsbgs , we derived two sets of stellar masses .
one set is the `` @xmath18-based '' stellar masses which are derived by using their @xmath18-band luminosity and @xmath76 color transformed from @xmath25 measured in @xcite .
the other set is the `` @xmath2-based '' stellar masses in the sdss mpa - jhu catalog which are computed from the sdss @xmath2-band luminosities in synergy with the @xmath2-band stellar @xmath77 derived by the sdss multi - band sed fitting method @xcite .
resultantly for our lsbgs , the `` @xmath18-based '' stellar masses are @xmath78@xmath79 m@xmath80 , with a mean of @xmath81 m@xmath80 and a median of @xmath82 m@xmath80 . the `` @xmath2-based '' stellar masses are @xmath83@xmath84 m@xmath80 , with a mean of @xmath85 m@xmath80 and a median of @xmath86 m@xmath80 .
we report the results in table [ tbl-4 ] .
apparently , the set of @xmath18-based " stellar masses is on average @xmath87 order of magnitude lower than the set of @xmath2-based " stellar masses . in this paper , we prefer the @xmath2-based " stellar masses for our lsbgs , and the reason will be given in 6.2 .
ccccc 191770 & 7.89 & 8.39 & 0.96 & 0.87 + 228010 & 7.95 & 8.37&0.98 & 0.94 + 253926 & 9.11 & 9.96 & 0.95 & 0.73 + 12249 & 9.21 & & 0.82 & + 321228 & 8.82 & & 0.79 & + 332804 & 9.20 & 9.88&0.90 & 0.64 + 332806 & 8.70 & 9.46 & 0.95 & 0.76 + 100811 & 9.09 & 9.79 & 0.87 & 0.57 + 123165 & 8.64 & & 0.74 & + 101807 & 9.11 & 9.88 & 0.93 & 0.70 + 1181 & 9.00 & 9.68 & 0.92 & 0.71 + 748835 & 9.09 & & 0.91 & + we show our lsbg data in the mass - metallicity ( @xmath88@xmath62 ) plot in figure [ fig : mass_metal ] . for metallicity , both @xmath1-based ( black ) and [ n ] /h@xmath0-based ( red ) oxygen abundances are displayed . for stellar mass ,
both the @xmath18-based " ( open circles ) and @xmath2-based " ( filled circles ) stellar masses are shown .
so , in this @xmath88@xmath62 plot , our lsbg sample appears as four individual sets of distributions , among which the two sets represented by red and black open circles are systematically biased to lower stellar mass because the @xmath18-based " stellar masses may be underestimated ( as mentioned in 5.3.1 and discussed in 6.2 ) .
so , comparing with distributions represented by the red and black open circles , we prefer distributions based on the @xmath2-based " stellar mass estimations ( red and black filled circles ) .
for comparison , the lsbg sample ( purple triangles ) of @xcite , the dirr sample ( dark - green squares ) of @xcite , and the massive spiral galaxy sample ( brown asterisks ) of @xcite in the very local universe ( including m101 , m31 , ngc 1365 , m81 , ngc 2997 , and so on ) are also included in this @xmath88@xmath62 plot . additionally , the @xmath88@xmath62 relation determined for @xmath89 sdss star - forming galaxies of @xcite is shown .
the black open diamonds represent the median in bins of 0.1 dex in stellar mass that include at least 100 data points .
the magenta line shows the polynomial fit of @xcite to the median data .
the cyan blue dashed and solid lines are the contours that enclose 68% and 95% of the data .
obviously , the distributions of our lsbg sample with @xmath2-based " stellar mass are generally consistent with the lsbg sample from @xcite at their common stellar mass space .
additionally , for normal star - forming galaxies and other types , there is a tendency that galaxies with lower stellar masses generally have lower metallicities . specifically , for normal star - forming galaxies with stellar masses between @xmath90 and @xmath91 m@xmath80 , @xcite find the correlation between stellar mass and metallicity to be roughly linear .
owing to the small coverage in stellar mass space , our lsbg sample is not sufficient for fitting a certain @xmath88@xmath62 relation , but galaxies in our sample are consistent with the relatively low abundances for their stellar masses .
the gas consisting mostly of hydrogen and helium is the raw material out of which stars and metals are formed . in lsbgs , the largest constituent of gas is assumed to be neutral atomic hydrogen .
the molecular gas remains unknown in quantity but , fortunately , it is not expected to contribute greatly to the total mass of gas in lsbgs due to their low metallicities and low dust content @xcite . as for ionized gas , it contributes negligibly to the mass of lsbgs @xcite .
the gas mass fraction ( @xmath92 ) refers to the fraction of baryons in gaseous form .
it can be simply measured by the ratio of the total mass in gas ( @xmath93 , including molecular , atomic , ionized , and metals ) to the combined total mass in gas and stars ( @xmath94 ) : @xmath95 so , taking the contribution of helium and metals into account , the total gas mass of lsbgs in units of solar mass can be given by @xmath96 where @xmath97 is the fraction of the gas mass in the form of hydrogen , adopted to be 0.733 @xcite .
following the formula above , @xmath93 for our edge - on lsbg sample can be easily computed by multiplying @xmath98 ( table [ tbl-1 ] ) which is available from the @xmath7240 catalog @xcite by 1/0.733 . for @xmath94,we
have previously derived two sets of stellar masses for our lsbg sample in 5.3.1 ( table [ tbl-4 ] ) .
consequently , we have two sets of gas mass fraction , @xmath92 , for our sample . in figure [
fig : gas_metal ] , we show oxygen abundance versus inverse gas mass fraction for our lsbg sample . both sets of gas mass fractions are shown for our sample .
the filled circles represent for@xmath92 being computed from the `` @xmath2-based ' ' @xmath94 while the open circles are for @xmath92 being computed from the `` @xmath18-based '' @xmath94 . besides , two sets of metallicities are also presented with red circles showing @xmath1-based metallicities and black circles showing [ n ] /h@xmath0-based metallicities . for comparisons , the lsbg sample from @xcite ( purple triangles ) and the dirr sample from @xcite ( green squares ) are also plotted . for our lsbg sample , data points with `` @xmath18-based '' stellar mass ( red and black open circles ) or with `` @xmath2-based '' stellar mass but @xmath1-based metallicities ( red filled circles )
have systematically biased to much larger gas mass fraction , compared with other samples ( e.g. , the lsbg sample from @xcite ; the dirr sample from @xcite ) .
however , data points with `` @xmath2-based '' stellar mass and [ n ] /h@xmath0-based metallicities ( black filled circles ) show general consistency with other samples in the common high-@xmath92 region .
the systematically larger gas mass fractions are more likely to be caused by different methods used for estimating stellar mass and metallicity .
again , we prefer the `` @xmath2-based '' stellar masses for our sample because lsbgs are acknowledged to be dominated by low - mass stars that would radiate light predominantly at red and near - infrared wavelengths . using @xmath18-band luminosity method is very likely to underestimate stellar masses of lsbgs and then overestimate their gas mass fractions . more details on this issue will be discussed in 6.2 .
stellar populations are the fossils of the past star formation and evolution of galaxies
. however , the sky background makes it very difficult to study the faint stellar continuum and absorption lines in the entire lsbg unless very large telescopes and extraordinarily long exposure times are used . because of such difficulties
, most spectroscopic studies of lsbg stellar populations have been based on spectra of the brighter bulges ( e.g. , morelli et al .
2012 ) or h regions .
a slit of appropriate width and length positioned along the major axis can generally capture light from almost all regions of an edge - on galaxy , especially of a super - thin galaxy .
thus , the spectral properties may in principle be much better determined for edge - on lsbgs compared with face - on ones .
if we expect to derive reliable stellar populations with full - spectrum fitting stellar population synthesis method , we must use high - quality spectra , especially with well - calibrated spectral continuum shapes . given the uncertainties associated with continuum shapes of the palomar hale spectra
, we decided to use only the three lick shane spectra ( of agc 191770 , agc 228010 , and agc 253926 ) for our stellar - population study of lsbgs .
all these three spectra have s / n @xmath11 in the continuum wavelength interval 47304780 and meet our requirements . for stellar population synthesis
, starlight @xcite is a widely used tool .
it can derive the age and metallicity of a galaxy , as well as its extinction and velocity dispersion , by fitting its observed spectrum with a superposition of populations of various ages and metallicities from stellar population models such as starlight ( * ? ? ?
* hereafter bc03 ) .
we used starlight to perform stellar population synthesis for the integrated spectra of the three edge - on lsbgs observed at lick . for the reddening , both the ( * ? ? ?
* hereafter cal ) law and the ( * ? ? ?
* hereafter ccm ) law were separately used during the fit , although the problem of dust attenuation is minimal in lsbgs because they are acknowledged to be dust poor .
we show the fit plot in figure [ fig : starlight ] .
as said by @xcite , the individual stellar population components from starlight are very uncertain because the existence of multiple solutions is an old known problem in stellar population synthesis .
thus , we should avoid focusing in detail on any individual stellar populations .
instead , we jump straight to results based on more robust descriptions of the resultant stellar populations .
as suggested in the starlight paper @xcite , a coarse but robust description of the star - formation history of our galaxies may be obtained by binning the resultant individual stellar populations onto young " ( y ; age @xmath99 gyr ) and `` old '' ( o ; age @xmath100 gyr ) components . in order to show the respective contributions of the two components , we provide the `` light fraction '' ( lf ) out of the total light and the `` mass fraction '' ( mf ) out of the total mass for y and o components in table [ tbl-5 ] .
results from both reddening laws ( cal and ccm ) for each lsbg seem to be consistent .
qualitatively , for both light and mass , the contributions from the o components seem to be much larger than those from the y components for these three lsbgs .
thus , rather than a quantitative conclusion , we can draw a qualitative conclusion that these three lsbgs may have a potential dearth of stars with ages @xmath3 gyr . instead , they are more likely to be dominated by stars with ages @xmath4 gyr .
this supports the general notion that lsbgs have a very low level of current star - formation activity . comparing with the previously measured subsolar gas - phase metallicities for the first three galaxies in table [ tbl-3 ] , their stellar metallicities ( from stellar population synthesis ; table [ tbl-5 ] )
are about solar . on one hand
, this inconsistency is likely to be caused mainly by the large uncertainties of the stellar metallicities . as can be deduced from the amount of work devoted to this topic of stellar population synthesis , recovering the stellar content of a galaxy with
accurately quantitative ages and metallicities from its observed integrated spectrum is not an easy task .
it needs a high - quality integrated spectrum , as well as models which are made and calibrated to mimic nature to the greatest degree .
any stellar population model , such as starlight ( bc03 ) , is based on a stellar library .
usually , the stellar library does not sample a sufficiently wide range in metallicity , making stellar metallicity notoriously more difficult to access than other properties .
also , to get the metallicity correct , one must have correct ages , as they affect one another in these integrated spectral fits .
since our derived stellar ages have large uncertainties , the stellar metallicities are necessarily also poorly constrained .
so , stellar metallicities determined here from stellar population synthesis ( table [ tbl-5 ] ) carry large uncertainties , and are not reliable as specific values for comparisons and scientific studies ; the gas - phase metallicities ( table [ tbl-3 ] ) from the emission - line flux ratios are much more reliable .
we use the blue - side spectra for studying stellar populations by full - spectrum fitting , so the accuracy of the relative flux calibration becomes an important factor for the reliability of the stellar population results .
if the slit is not positioned along the atmospheric dispersion , the light loss at blue - end wavelengths would be greater than that at red - end wavelengths . as shown in table [ tbl-2 ] in section 3.2 ,
if a star was being centered at 6000 and observed with the same slit positioning schemes and observing configurations as those in our program , it will be off center by 0.61@xmath29 at 3600 and 0.05 @xmath29 in the opposite direction at 6500 for agc 191770 , by 0.8@xmath29 at 3600 and 0.06@xmath29 in the opposite direction at 6500 for agc 228010 , and by 0.8@xmath29 at 3600 and 0.06@xmath29 in the opposite direction at 6500 for agc 253926 .
the light loss at 3600 is ( respectively ) 4.8% , 15.2% , and 0.6% greater than that at 6500 for agc 191770 , agc 228010 , and agc 253926 .
obviously , the relative blue - light loss of agc 228010 is the largest among the three galaxies .
however , such an amount of light loss is calculated under the assumption that our targets were stars .
actually , as seen in figure [ fig : edgeon - slit ] , they are galaxies with extended areas larger than the slit width .
so , the lost blue light is partly gained back from the neighboring pixels , and the effect on relative flux calibration of blue - end spectra is not that significant , although it still exists .
the probable light loss at blue - end wavelengths ( @xmath101 ) would cause an underestimate of late a - type stars . even with some blue - light loss
, we could still obtain the same qualitative conclusion that these three lsbgs have a potential dearth of stars with ages @xmath3 gyr because late a - type stars typically have ages older than @xmath74 gyr . limited by the reliability of a spectral continuum , the number of spectra that we can use for a possibly reliable conclusion from full - spectrum fitting is only three .
undoubtedly , this is from a representative sample of the whole edge - on lsbg sample , and not even representative of our final 12 observing targets .
thus , conclusions from the three galaxies may not apply more generally to lsbgs . to get more universal conclusions on stellar populations of lsbgs
, we will seek sufficient observing time in the future , obtaining stellar continuum spectra of good quality for a larger and more representative sample of edge - on lsbgs .
ccccccc log(age@xmath102 ) & 8.8 & 8.7 & 9.0 & 9.0 & 8.9 & 8.7 + log(age@xmath103 ) & 9.1 & 9.1 & 9.0 & 9.0 & 9.3 & 9.0 + @xmath104&0.030 & 0.030 & 0.050 & 0.050 & 0.037 & 0.040 + @xmath105&0.045 & 0.047 & 0.050 & 0.050 & 0.040 & 0.049 + @xmath106 ( mag ) & 0.18 & 0.24 & 0.00 & 0.00 & 0.91 & 0.99 + lf@xmath107 & 0.0 & 0.0 & 0.0 & 0.0 & 9.4 & 11.2 + lf@xmath108 & 100.0 & 100.0 & 100.0 & 100.0 & 90.6 & 88.8 + mf@xmath109 & 0.0 & 0.0 & 0.0 & 0.0 & 0.04 & 0.1 + mf@xmath108 & 100.0 & 100.0 & 100.0 & 100.0 & 99.96 & 99.9 + in figure [ fig : gas_metal ] ,
the gas mass fractions computed from two different sets of stellar masses for our sample have obvious discrepancies .
the gas mass fractions from the `` @xmath18-based '' stellar masses are apparently higher than those from the @xmath2-based " stellar masses .
which set of stellar mass is more appropriate for lsbgs ? here , we prefer the `` @xmath2-based '' stellar masses because of the following reasons : ( 1 ) the luminosity at relatively longer wavelengths ( the red or near - infrared bands ) is more ideal for tracing the stellar mass because it is less affected by extinction and other factors that affect shorter - wavelength bands . ( 2 )
lsbgs have few blue , massive stars formed in the past 1 gyr ; the stellar mass is predominantly in low mass stars which radiate most of their light at red and near - infrared wavelengths .
we checked the _ wise _ satellite 3.4 @xmath110 m and 4.6 @xmath110 m images of our lsbgs and found that all of them are detected in both bands , especially in the bulge regions .
thus , the stellar masses of lsbgs would be underestimated from their @xmath18-band luminosity .
( 3 ) the @xmath2-based " stellar masses of our lsbg sample
are @xmath111@xmath112 m@xmath80 , which agrees with the conclusion of other studies that lsbgs are mostly comparable in mass with prominent spirals that define the hubble sequence ( e.g. , * ? ? ?
* ) . in the closed - box model of the chemical evolution of galaxies , the initially metal - free gas is permanently locked into stars with neither inflow nor outflow and only evolves `` in isolation '' with time .
other assumptions include an invariant stellar initial mass function and instantaneous recycling .
the relation between the metallicity ( @xmath62 ) and the gas mass fraction ( @xmath92 ) for a system that evolves in isolation is @xmath113 where @xmath114 is the `` yield , '' the ratio of the mass of newly formed metals to the mass of gas permanently locked in stars . here
, @xmath62 is best expressed as an elemental abundance in logarithmic form . for oxygen , following @xcite
, this equation can be rewritten as @xmath115 in a plot of oxygen abundance , 12 + log(o / h ) , against inverse gas mass fraction , conveyed by loglog(@xmath116 ) , the closed - box model predicts a slope of unity . deviations from this slope may be an indication for the inflow or outflow of gas .
the oxygen yield , @xmath117 , can be derived from the intercept of the plot .
following @xcite , for the field dirrs , there is an excellent correlation represented by a line with a slope of unity ( green solid line in figure [ fig : gas_metal ] ) which is consistent with the closed - box model of the chemical evolution of galaxies .
the standard deviation of the fit is 0.126 dex in log(o / h ) . from the intercept of the line ,
the effective oxygen yield by mass ( @xmath117 ) for the field dirrs from @xcite is 2.22@xmath11810@xmath119 .
obviously , the lsbg data ( purple open triangles ) from @xcite are generally consistent with the field dirr data ( green open squares ) and can also be fitted by the closed - box model fit ( green solid line ) of @xcite , albeit with more scatter .
we can not tell for sure whether our lsbg sample can be explained with the closed - box model by fitting ; our sample is too small and also too narrow in the coverage of both gas mass fraction and gas - phase metallicity to show a true relation , even if some essential relations do indeed exist .
however , to compare with the field dirrs from @xcite and the lsbgs from @xcite , we still forced our lsbg data ( all filled circles with `` @xmath2-based '' stellar mass ) to be fitted by a line with a slope of unity ( red solid line ) .
the standard deviation of this fit is 0.27 dex in log(o / h ) . from this artificial fitting ,
the effective oxygen yield by mass for our lsbg sample is @xmath120 under the assumption of a closed - box model .
the value of true yield is often considered to be the solar oxygen abundance by mass . for the sun , the oxygen abundance is @xmath70 log(o / h ) @xmath121 dex @xcite , which is @xmath122 by mass .
apparently , the effective oxygen yield for our lsbg sample is larger than that for the dirr sample , and is 69% of the value for the true oxygen yield .
the difference between the effective and true yields roughly represents the fraction of metal lost ( up to 30% ) up to now @xcite .
such a value of metal loss can be reasonable since it is often argued that lsbgs have shallow potential wells , making it easier for mass loss to occur .
however , given the substantial uncertainties in both metallicity and gas mass fraction for our lsbg sample , this mass - loss value of 30% is only a rough estimate , and we can not completely rule out a closed - box scenario .
additionally , judging from the offset between the two closed - box lines in figure [ fig : gas_metal ] , we find evidence that our lsbgs retain up to @xmath123 times as much of their metals compared with dwarf irregulars ( green squares ) , consistent with metal retention being related to galaxy mass . in any case , although the closed - box scenario can not be completely ruled out for our lsbg sample , there is still substantial room for considering other , more complicated chemical evolution models with outflow or even infall ( e.g. , * ? ? ? * ; * ? ? ?
* ) for our lsbg sample . in the @xmath2-based stellar masses ( table [ tbl-4 ] )
, our lsbg sample is estimated to be @xmath124@xmath112 m@xmath80 , showing no fundamental differences from prominent spiral galaxies .
furthermore , comparing with previously published luminosity - metallicity plots for lsbgs and other galaxy types ( figure [ fig : lumi_metal ] ) , our lsbg sample agrees well with the lsbg sample of @xcite and also generally the sample of normal galaxies . in the mass - metallicity plot ( figure [ fig : mass_metal ] ) ,
our lsbg sample is also consistent with the lsbg sample of @xcite and with the relatively low abundance for their stellar mass , which is appropriate for normal galaxies .
therefore , from the behavior in the @xmath71@xmath62 and @xmath88@xmath62 plots , our lsbg sample does not seem to be fundamentally different from normal disk galaxies . regarding the plot of gas mass fraction versus oxygen abundance ( figure [ fig : gas_metal ] ) , we prefer the @xmath2-based stellar masses ( black or red filled circles ) .
our lsbg sample is not likely to be explained by simple closed - box models , although the galaxies evolve in relative isolation .
systematically biased to larger gas mass fractions , our lsbg sample has higher oxygen abundances for smaller gas mass fractions , which is also shown by other samples . such qualitative agreement in gas mass fraction versus
metallicity distribution makes the lsbgs seem consistent with other galaxy types .
we therefore suggest that lsbgs are probably not fundamentally different from normal galaxies , but are just the continuous extension to the lower surface - brightness end of normal disk galaxies .
we have obtained long - slit spectra of 12 edge - on lsbgs selected from the egis-@xmath7240 sample .
observations were made with the kast spectrograph on the 3-m shane telescope at lick observatory for 3 galaxies and with the double spectrograph on the 5-m hale telescope at palomar observatory for 9 galaxies .
we adopted a median - filter method to estimate the sky background of objects observed at lick observatory and a second - order polynomial to fit the sky background of objects observed at palomar observatory , which can be comparable to or even exceed the galaxy surface brightnesses .
the stellar populations of three of the lsbgs were studied with the program starlight , as their spectral quality is sufficiently high .
we find that our lsbgs are dominated by stellar populations with ages @xmath4 gyr and have a potential dearth of stars with ages @xmath3 gyr .
this indicates that lsbgs probably have a very low level of star - formation activity over the past 1 gyr , which may be the most likely explanation for their low surface brightness .
we have measured oxygen abundances ( 12 + log(o / h ) ) using both the @xmath1 and [ n ] /h@xmath0 diagnostics after performing reddening corrections for the observed fluxes of emission lines .
the @xmath1-based metallicities are 8.008.62 , with a mean of 8.30 dex and a median of 8.26 dex .
the [ n ] /h@xmath0-based metallicities are 7.678.54 , with a mean of 8.20 dex and a median of 8.26 dex .
the two sets of oxygen abundances are consistent with each other .
comparing with the luminosity - metallicity relation established in previous work on lsbgs , normal spirals , and other galaxy types , our lsbgs tend to be consistent with normal galaxies .
we derive two sets of stellar masses for our lsbg sample .
one set is the @xmath2-based stellar masses which are @xmath125@xmath112 m@xmath80 , and the other set is the @xmath18-based stellar masses which are @xmath126@xmath127 m@xmath80 , 0.5 order of magnitude lower .
we argue that the @xmath2-based stellar masses are better for our lsbg sample because such galaxies primarily consist of low mass stars which radiate their light mostly at red and infrared wavelengths .
stellar masses derived from the @xmath18 band might be underestimated . according to the @xmath2-based stellar masses , our lsbgs are comparable to normal spiral galaxies in stellar mass , but they have considerably lower abundances . in the plot of gas mass fraction versus oxygen abundance
, we find that like other galaxy types , our lsbgs with relatively high `` @xmath2-based '' stellar masses have lower metallicities for higher gas mass fractions .
regarding the chemical evolution of our sample , the lsbg data appear to allow for up to 30% metal loss , but we can not completely rule out the closed - box model . additionally , we find evidence that our galaxies retain up to about 3 times as much of their metals compared with dwarfs , consistent with metal retention being related to galaxy mass . combining the above conclusions , our lsbg sample is generally consistent with the luminosity - metallicity , mass - metallicity , and gas mass fraction - metallicity plots of normal spiral galaxies and other galaxy types ( e.g. , dirrs , starbursts ) .
thus , we speculate that lsbgs are probably not fundamentally different from normal galaxies , but rather are just the continuous extension in surface brightness to the low end .
we acknowledge the anonymous referee for whose detailed comments helped strengthen this paper .
this research uses data obtained through the telescope access program ( tap ) , which has been funded by the national astronomical observatories of china , the chinese academy of sciences ( the strategic priority research program `` the emergence of cosmological structures '' grant no .
xdb09000000 ) , and the special fund for astronomy from the ministry of finance . three observational nights ( 2022 sep
. 2015 ) with the double spectrograph on the 5-m hale telescope were distributed to us for scientific studies of lsbgs via the tap .
observations obtained with the hale telescope at palomar observatory were obtained as part of an agreement between the national astronomical observatories , chinese academy of sciences , and the california institute of technology .
additionally , we are grateful to the staffs of lick and palomar observatories for their assistance with the observations .
research at lick observatory is partially supported by a generous gift from google .
w. du thanks cheng cheng , zhimin zhou and fan yang in naoc for discussions concerning data reduction , and thanks ruixiang chang in shao for helpful suggestions during revision .
this work is supported by the national natural science foundation of china ( grant nos .
11403037 , 11225316 , and 11173030 ) , the strategic priority research program
the emergence of cosmological structures " of the chinese academy of sciences ( grant no .
xdb09000000 ) , and the key laboratory of optical astronomy , naoc .
has received generous financial assistance from the christopher r. redlich fund , the tabasgo foundation , and nsf grant ast-1211916 .
abazajian k. n. , adelman - mccarthy j. k. , agueros m. a. et al .
2009 , , 182 , 543 bell e. f. , mcintosh d. h. , katz n. , weinberg m. d. 2003 , , 149 , 289 bizyaev d. v. , kautsch s. j. , mosenkov a. v. et al .
2014 , , 787 , 24 bothun g. d. , impey c. d. , malin d. f. , mould j. r. 1987 , , 94 , 23 bothun g. d. et al .
1997 , , 109 , 745 brodie j. p. , huchra j. p. 1991
, , 379 , 157 bruzual g. , charlot s. 2003 , , 344 , 1000 burkholder v. , impey c. , sprayberry d. 2001 , , 122 , 2318 calzetti d. , armus l. , bohlin r. c. et al .
2000 , , 533 , 682 cardelli j. a. , clayton g. c. , mathis j. s. 1989 , , 345 , 245 ceccarelli l. , herrera - camus r. , lambas d. g. , galaz g. , padilla n. d. 2012 , , 426 , 6 chang r. x. , hou j. l. , shen s. y. , shu c. g. 2010 , , 722 , 380 cid fernandes r. , mateus a. , sodre l. , stasinska g. , gomes j. m. 2005 , mnras , 358 , 363 cole s. , norberg p. , baugh c. m. et al .
2001 , , 326 , 255 de grijs r. 1998 , mnras , 299 , 595 de blok w. j. g. , mcgaugh s. s. 1998 , , 508 , 132 de blok w. j. g. , van der hulst j. m. 1998 , a@xmath128a , 336 , 49 de naray r. , mcgaugh s. , de blok w. 2004 , , 355 , 887 das m. et al .
2009 , , 693 , 1300 denicolo g. , terlevich r. , terlevich e. , 2002 , , 330 , 69 disney m. j. 1976 , nature , 263 , 573 du w. , wu h. , lam m .-
2015 , , 149 , 199 edmunds m. g. 1990 , , 246 , 678 edmunds m. g. 1999 , in davies j. i. , impey c. , phillips s. , ed .
, asp conf .
vol.170 , the low surface brightness universe .
pac . , san francisco , p. 383
faber s. m. 1973 , , 179 , 731 filippenko a. v. 1982 , , 94 , 715 galaz g. et al .
2011 , , 728 , 74 garnett d. r. , shield g. a. 1987 , , 317 , 82 garnett r. d. 2002 , , 581 , 1019 grevesse n. , noels a. , sauval a. j. 1996 , in asp conf . ser .
99 , cosmic abundances : proceedings of the 6th annual october astrophysics conference , ed . s. s. holt @xmath128 g. sonneborn
( san francisco : asp ) , 117 haberzettl l. et al . 2007 ,
a@xmath128a , 471 , 787 haynes m. p. , giovanelli r. , martin a. m. et al .
2011 , , 142 , 170 impey c. @xmath128 bothun g. 1997 , ar@xmath128aa , 35 , 267 impey c. , burkholder v. , sprayberry d. 2001 , , 122 , 2341 kauffmann g. , heckman t. m. , white s. d. m. et al .
2003 , , 314 , 33 kelson d. d. 2003 , , 155 , 688 kewley l. j. & dopita m. a. 2002 , , 142 , 35 kochanek c. s. , pahre m. a. , falco e. e. et al .
2001 , , 560 , 566 lam m. i. , wu h. , yang m. et al .
2015 , , 446 , 4291 lamareille f. , mouhcine m. , contini t. , lewis i. , maddox s. 2004 , , 350 , 396 lee h. , mccall m. l.,kingsburgh r. l. , ross r. , stevenson c. c. 2003 , , 125 , 146 lequeux j. , peimbert m. , rayo j. f. , serrano a. , torres - peimbert s. 1979 , a@xmath128a , 80 , 155 liang y. c. et al .
2010 , , 409 , 213 matthews l. d. @xmath128 gallagher j. s. , 114 , 1899 matthews l. d. @xmath128 wood k. 2001 , , 548 , 150 mcgaugh s. s. , 1991 , , 380 , 140 mcgaugh s. s. 1994 , , 426 , 135 mcgaugh s. s. 1996 , , 280 , 337 mcgaugh s. s. , de blok w. j. g. 1997 , , 481 , 689 mcgaugh s. s. @xmath128 bothun g. d. 1994 , , 107 , 530 melbourne j. , salzer j. j. 2002 , , 123 , 2302 j. s. & stone r. p. s. 1993 , lick obs . tech . rep .
66 santa cruz : lick obs .
minchin r. f. et al .
2004 , , 355 , 1303 morelli l. , corsini e. m. , pizzella a. et al . , 2012 ,
, 423 , 962 oke j. b. & gunn j. e. 1982 , pasp , 94 , 586 oneil k. et al .
2000 , , 119 , 136 osterbrock d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( mill valley , ca : university science books ) pagel b. e. j. et al .
1979 , , 189 , 95 pettini m. @xmath128 pagel b. e. j. , 2004 , , 348 , l95 schombert j. m. , bothun g. d. , impey c. d. , mundy l. g. 1990 , , 100 , 1523 schombert j. , mcgaugh s. s. , eder j. a. 2001 , , 121 , 2420 schombert j. et al .
1992 , , 103 , 1107 schombert j. et al .
2013 , , 146 , 41 skillman e. d. , kennicutt r. c. , hodge p. w. 1989 , , 347 , 875 smith j. a. , tucker d. l. , kent s. , richmond m. w. et al .
2002 , , 123 , 2121 trachternach c. , bomans d. j. , haberzettl l. , dettmar r. j. 2006 , a@xmath128a , 458 , 341 tremonti c. a. , heckman t. m. , kauffmann g. et al .
2004 , , 613 , 898 van dokkum p. g. 2001 , , 113 , 1420 van der hulst j. m. et al .
1993 , , 106 , 548 van der kruit p. c. 1981 , a@xmath128a , 99 , 298 van der kruit p. c. 1988 , a@xmath128a , 192 , 117 van der kruit p. c. , kregel m. , freeman c. k. 2001 , a@xmath128a , 379 , 374 wen x. q. , wu h. , zhu y. n. et al .
2013 , , 433 , 2946 wu h. , cao c. , hao c. n. et al .
2005 , , 632 , 79 zaritsky d. , kennicutt jr .
r. c. , huchra j. p. 1994
, , 420 , 87 zhu y. n. , wu h. , li h. n. , cao c. 2010 , raa , 10 , 329 | we present long - slit optical spectra of 12 edge - on low surface brightness galaxies ( lsbgs ) positioned along their major axes . after performing reddening corrections for the emission - line fluxes measured from the extracted integrated spectra , we measured the gas - phase metallicities of our lsbg sample using both the [ n ] /h@xmath0 and the @xmath1 diagnostics .
both sets of oxygen abundances show good agreement with each other , giving a median value of 12 + log(o / h ) = 8.26 dex . in the luminosity - metallicity plot ,
our lsbg sample is consistent with the behavior of normal galaxies . in the mass - metallicity diagram , our lsbg sample has lower metallicities for lower stellar mass , similar to normal galaxies .
the stellar masses estimated from @xmath2-band luminosities are comparable to those of prominent spirals . in a plot of the gas mass fraction versus metallicity
, our lsbg sample generally agrees with other samples in the high gas mass fraction space .
additionally , we have studied stellar populations of 3 lsbgs which have relatively reliable spectral continua and high signal - to - noise ratios , and qualitatively conclude that they have a potential dearth of stars with ages @xmath3 gyr instead of being dominated by stellar populations with ages @xmath4 gyr . regarding the chemical evolution of our sample , the lsbg data appear to allow for up to 30% metal loss , but we can not completely rule out the closed - box model . additionally , we find evidence that our galaxies retain up to about 3 times as much of their metals compared with dwarfs , consistent with metal retention being related to galaxy mass . in conclusion ,
our data support the view that lsbgs are probably just normal disk galaxies continuously extending to the low end of surface brightness . |
magnetic reconnection is thought to be one of the most important mechanisms by which magnetic energy is converted to kinetic energy in astrophysical plasmas ( for a review , see e.g. * ? ? ?
this conversion may produce a tail in the particle energy distribution including randomly oriented relativistic particles and high - frequency radiation , coherent bulk outflows , or both . in the relativistic reconnection regime ,
in which the ratio of the magnetic energy to the total enthalpy of the particles ( the magnetization @xmath3 ) is much larger than 1 , magnetic reconnection has been hypothesised to be responsible for observed high - energy emission in the prompt phase of gamma - ray bursts ( grbs ) @xcite , emission from pulsar wind nebulae @xcite , and active galactic nucleus ( agn ) jets @xcite .
relativistic jets launched by the blandford - znajek mechanism @xcite are expected to be dominated by magnetic energy , so reconnection is a primary mechanism invoked for the conversion of these fields to particles and radiation . since
relativistic bulk outflows have been predicted to occur in relativistic reconnection @xcite , it has been hypothesized that the variability of these sources may arise from the direction of these outflows swinging past the observer s line of sight .
the fast variability model of blazar jets @xcite , and the jet - within - a - jet model for the prompt emission in grbs ( assuming that the jet is magnetically dominated ) @xcite both rely on this beaming to produce the fast variability observed , so the amount of beaming produced in magnetic reconnection is critical for evaluating their viability .
numerical investigations of particle acceleration in relativistic magnetic reconnection are generally carried out using particle - in - cell ( pic ) simulations that include all kinetic effects required to probe the non thermal spectrum of high - energy particles and the possible anisotropy in the particle distribution .
these simulations , which have mostly focused on the pair - plasma case which is easier to simulate , have shown that reconnection can produce non thermal tails of highly accelerated particles in both two and three dimensions , and results are similar when electron - ion plasmas are studied @xcite . in most cases ,
the spectrum of the highly accelerated particles has been consistent with a power law , although analytical models by @xcite and @xcite indicate that the distribution may be different .
this distribution can be hard enough , especially at large magnetisation , that the particles at the highest lorentz factors dominate the energy distribution @xcite .
it has been suggested that the drift - kink instability ( dki ) may pnt the development of fast reconnection and particle acceleration in three - dimensional simulations without a guide field perpendicular to the reversing field of reconnection @xcite , but more recent work has shown that on longer timescales fast reconnection and particle acceleration indeed take place , just as in the two - dimensional case @xcite . in this work ,
we therefore carry out two dimensional simulations with the expectation that the results will apply to three - dimensional reconnection as well .
we do note that the details of reconnection in 2d are likely to be different from those of the reconnection on long timescales that follows the initial effects of the drift - kink instability , but the overall physics is likely to be the same .
the relativistic beaming of particles and the resulting synchrotron radiation resulting from coherent bulk outflows in reconnection has been investigated to a lesser degree .
the evidence for the presence of relativistic bulk flows in simulations is uncertain , with @xcite finding fast relativistic flows , and @xcite finding in the case of electron - ion plasmas that only mildly relativistic outflows are present . in relativistic mhd simulations of forced reconnection with moderate magnetization
, @xcite found that near - alfvnic outflows were produced , although it is uncertain whether this would continue at high magnetization .
it should be noted that the bulk outflows may not be a good probe of the beaming of the outgoing radiation , since the hard power laws found in relativistic reconnection indicate that the output flux will be dominated by higher - energy particles , which may not be beamed in the same way as the bulk plasma .
@xcite have found in both two and three - dimensional simulations that substantial beaming of particles and radiation results from magnetic reconnection ; the radiation can be beamed within @xmath4 of the sky , however , these simulations focus on the radiation of particles close to the synchrotron burnoff limit , where they radiate in a time comparable to their period of gyration .
@xcite carries out simulations of explosive reconnection including radiative feedback , finding significant beaming and variability in both the fast and slow cooling cases , but low radiative efficiency in the slow cooling case . in this work ,
we investigate the beaming of particles under the assumption that they do not radiate their energy quickly .
relatively little numerical investigation has been done of relativistic reconnection at high magnetisations , but there is significant evidence that reconnection physics , and thus , beaming , will be significantly affected . @xcite and @xcite , but not @xcite , find that at high magnetisations the velocity of inflows into the current sheet and the normalised rate of reconnection of magnetic fields becomes extremely high , of order unity as opposed to the typical values of @xmath5 found for both nonrelativistic @xcite and relativistic ( e. g. * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) reconnection at lower magnetisation .
it is possible that at high magnetisations the density of background plasmas is very low . while magnetospheric space - plasma measurements have indicated that density contrasts in nonrelativistic reconnection are typically modest @xcite , relativistic reconnection has never been observed directly and it is uncertain whether this result will apply in the relativistic case .
@xcite finds that in this low - density case , the current sheets became very thick , and there is a deficit of particles at intermediate lorentz factors between the beginning of the nonthermal tail and the highest - energy particles .
our simulations investigate the physics of relativistic reconnection at high magnetisation and low background density , as well as extending the investigation of beaming to these relatively unexplored regimes .
this paper is organised as follows . in section
[ sec : methodology ] we discuss our method of simulation and calculation of synchrotron radiation . in section [ sec : results ] , we present the results of our simulations , including our calculations of the beaming of both particles and synchrotron radiation .
finally , in section [ sec : conclusions ] , we discuss our conclusions .
we use the relativistic particle - in - cell ( pic ) plasma code tristan - mp @xcite to simulate the evolution of relativistic reconnection in two dimensions in a pair plasma .
particle - in - cell simulations include the kinetic effects of individual particle motions by evolving the discretised versions of the exact equations of electrodynamics - maxwell s equations and the lorentz force law .
the fields are calculated on the vertices of a grid , while the particles that are evolved using the lorentz force law ( interpolated from the grid to their positions ) are macroparticles , each of which represent many physical particles .
this section describes the initial conditions , parameter choices , and resolution requirements in these simulations .
we use a rectangular spatial domain of the form @xmath6 , @xmath7 , and the boundary conditions are periodic in all directions .
the initial configuration contains two harris current sheets @xcite without guide field at @xmath8 and @xmath9 with antiparallel currents .
it is characterised by the magnetic field profile @xmath10\hat{{\mathbf y } } , \label{eq : harris_field}\ ] ] where @xmath11 is the half - thickness of each current sheet .
the density profile , which is defined including both species , consists of a specially varying , drifting current population entered at each current sheet , plus a background population of stationary particles
. @xmath12+n_{\rm b}.\ ] ] pressure equilibrium between the background and the current sheet requires that @xmath13 , where @xmath14 is the temperature of the particles ( in units of @xmath15 ) in the drifting current sheet population in the simulation frame .
the density in the middle of each current sheet in the lab frame is @xmath16 , and the density in the background plasma is @xmath17 .
the drift velocity @xmath18 of the positively charged and negatively charged particles in the current sheet population is given by the relation @xmath19 where the sign is positive for particles in the current sheet at @xmath8 and negative for particles in the current sheet at @xmath20 .
we do not add any initial perturbation to this equilibrium , but instead allow instabilities to grow from noise .
one of the important parameters determining the physics of reconnection is the value of the magnetization @xmath3 , which in a pair plasma is given by @xmath21 where @xmath22 is the average enthalpy of particles , @xmath0 is the mean particle lorentz factor and @xmath23 is the particle pressure .
we investigate the physics of reconnection at three values of the magnetisation in the background plasma @xmath244 , 40 , and 400 .
we initialize the drifting and stationary populations of both species in a relativistic maxwellian at temperature @xmath25 , which corresponds to @xmath26 and @xmath27 .
the value of @xmath28 used to initialise each simulation may be found using the equation , @xmath29 where we define @xmath30 as the magnetisation in the background plasma .
this equation can be derived by combining the equation of pressure equilibrium above with the definition of @xmath3 ( [ eq : sigmadef ] ) .
the other important parameter for the physics of reconnection is the ratio @xmath31 , where @xmath32 is the plasma skin depth of particles in the center of the current sheet , given by @xmath33 this ratio sets the drift velocity of the plasma , which can be found by combining equations ( [ eq : drift ] ) , ( [ eq : nratio ] ) , and ( [ eq : plasma ] ) to yield @xmath34 we choose to set @xmath35 , which corresponds to a mildly relativistic drift velocity in the current sheet @xmath36 .
this current sheet width is small enough that fast reconnection can occur .
our choices for the parameters set the physical system up to the choice of fiducial magnetic field @xmath37 , so that it can be scaled to model many physical systems . in a pic simulation ,
the number of macroparticles of each species located in each grid cell must be large enough to resolve variations in the current density and limit high - frequency particle noise .
tristan - mp uses a current filtering algorithm to reduce high - frequency particle noise , substantially reducing the required number of macroparticles per cell per species .
a convergence test described in the work of @xcite has demonstrated that the evolution of reconnection configurations with this filtering in tristan - mp is unaffected by particle noise for densities as low as 4 particles / cell / species .
we choose to initialize the simulations with 8 macroparticles / cell / species , but we have carried out convergence tests for particle densities up to @xmath38 macroparticles / cell / species , finding similar results .
density gradients in the initial configuration are accounted for without computational cost by varying the mass of macroparticles while keeping the charge - to - mass ratio the same . in all simulations , the fraction of particles with mass significantly different than the average is small , and the presence of these varied macroparticles has no significant effect on the dynamics . to resolve the kinetic length scales of reconnection , which automatically resolves modes on the scale of the current sheet width
, we set the grid spacing @xmath39 , which ensures that the kinetic spatial and temporal scales are adequately resolved .
the convergence tests by @xcite discussed earlier in this section found that the evolution is insensitive to increases in resolution beyond this value , but we also carry out separate convergence tests that indicate similar physics up to @xmath40 . finally , to ensure that we are able to probe nonlinear reconnection in a quasi - steady state
, the size of the simulation must be much larger than the fastest - growing linear tearing mode wavelength of order @xmath41 , and the duration @xmath42 of the simulation must be significantly larger than the time required for an alfvn wave to cross the box ( see @xcite for a detailed discussion ) .
we choose @xmath43 , @xmath44 , which is much larger than the tearing length scale of @xmath45 .
these large spatial scales for the box make adding 3d effects impractical because it would require larger computational resources by a factor of approximately 5000 . for our simulations , in which @xmath46 ,
the alfvn speed is close to the speed of light , so the duration of the simulations must satisfy @xmath47 , where @xmath48 is the plasma oscillation frequency .
we choose to run our simulations for a total duration of at least @xmath49 , which is significantly longer than this required duration .
energy is conserved within @xmath50 in all of our simulations . in this paper
, we calculate the radiation spectrum assuming that particles radiate solely through synchrotron radiation in the limit where radiative feedback is negligible ; because the presence of electric fields means that the synchrotron radiation formulae can not be used in the lab frame , and there may be no frame in which the electric field vanishes in a reconnection region we use the prescription of @xcite to calculate an effective magnetic field @xmath51 that would produce the instantaneous radius of curvature of the particle ; this replaces all factors of @xmath52 , where @xmath53 is the angle between the particle direction of motion and the magnetic field direction , in synchrotron formulae . assuming that the force during a single tilmestep is adequately approximated by the lorentz force at the particle s position , the effective magnetic field is @xmath54 where @xmath55 is the instantaneous lorentz force resulting from the electromagnetic field at the particle location .
then the total synchrotron power @xmath23 is given by @xmath56 and the synchrotron spectrum for each particle is given by @xmath57 where @xmath58 is the radiation frequency and @xmath59 is the peak radiation frequency , @xmath60 and @xmath61 is the synchrotron function @xmath62 where k represents a modified bessel function of the second kind . in calculating the synchrotron spectrum as a function of direction , we do not use the fully general formula in which radiation at different frequencies has different angular distribution ; instead , we assume that all radiation is distributed according to the angular distribution of total power per unit solid angle @xcite @xmath63 where @xmath64 is the angle between the particle direction and the observer direction and @xmath65 is the power per unit solid angle of radiation emitted in the direction of particle motion @xmath66 .
we then use a kernel to allocate the total radiation in a discrete grid of observer directions .
our calculations of synchrotron power are in arbitrary units , and our calculated radiation frequencies are normalized to ( within order unity ) @xmath67 , so our synchrotron radiation calculations , like our particle - in - cell calculations , do not depend on the absolute value of the magnetic field @xmath68 .
the evolution of the current sheet in all of our simulations is similar to that found in previous 2d simulations of magnetic reconnection .
the initial configuration is unstable to the tearing instability seeded by noise , and this instability grows producing chains of alternating x - points where reconnection occurs and dense magnetic islands where the outflows from reconnection meet .
the instability becomes nonlinear at @xmath69 in all simulations , and the islands begin to merge in hierarchical fashion until only one is left at typical time @xmath70 , at which point fast energy transfer stops .
when the x - points and islands become large enough , secondary islands and secondary x - points in between them can form due to secondary tearing instabilities in the large x - points or in the current sheet produced between merging magnetic islands .
figure [ fig : schematic ] shows schematically these structural elements of the reconnecting current sheet in our simulation with @xmath71 during this nonlinear stage . in the x - points the particles are accelerated by the electric field in the @xmath72 direction . while at the edges of the x - point , particles flow towards the islands in the @xmath73 direction as part of the outflow from the x - point . as we show later in section [ sec : velmapping ]
, particles in the x - point tend to have most of their motion in this plane , with different species having opposite motions in the @xmath74 direction but similar motions in the @xmath75 directions at a given location . in this paper
, we focus on the configuration of reconnection midway through this process at @xmath76 , a time when the rate of conversion of magnetic energy to kinetic energy is near its peak and the simulation is not yet affected by the boundary conditions because there has not been time for for a light wave to cross the box . in this way
, we ensure that our results are not sensitive to the size of the simulations . because the initiation of reconnection occurs at slightly different rates in each simulation we choose a slightly different time in each simulation for detailed analysis .
our results , which are discussed throughout this section , are summarised in table [ tab : overview ] .
cccccc & 0.15&1.55&1.65&25&no + s40&0.20&1.82 & 1.5&130&yes + s400&0.17&2.04&1.4&180&yes while many aspects of the simulations do not depend qualitatively on @xmath30 , significant differences are found between simulation s4 with @xmath77 and simulations s40 and s400 with @xmath78 . in simulation s4 ,
the current sheets remain thin throughout the evolution , consistent with the results of @xcite . as a result , the aspect ratios of the large - scale x - points are large , and secondary islands and x - points are produced within them by secondary tearing instabilities .
in contrast , for simulations s40 and s400 we find that the current sheet width increases as the structures become larger , and aspect ratios remain small , of order 10 . as a result
, no substructure is formed in these current sheets .
this is likely a result of the initial conditions in our simulations , in which density contrast is used in the initial conditions to maintain pressure balance .
@xcite found in their two - dimensional relativistic simulations that when the density contrast in the current sheet , which in our simulations with @xmath79 is approximately @xmath80 , is significantly larger than 10 , a phase change in reconnection occurs which leads to widening of the current sheet and produces significantly different reconnection physics . in this paper
, we examine what aspects of reconnection are significantly affected by this phase change .
this section is organized as follows . in section [ sec : overallevol ] we discuss the overall evolution of the reconnection configuration in the three simulations . in section [ sec : reconnectionrate ]
we discuss methods of measurement of the reconnection rate in our simulations and compare our results to those found in other simulations . in section [ sec : partaccel ]
we investigate the particle energy spectrum produced by acceleration in the overall simulation and the local particle energy distribution at various locations in the current sheet . in section [ sec : velmapping ] we calculate the spatial distribution of velocities for particles of different energies throughout the current sheet . in section [ sec : beaming ] , we calculate the angular distributions of particle velocities and the resulting synchrotron radiation , and investigate whether strong beaming of either is produced in our simulation . figure [ fig : overallevol ] shows the reconnection configuration for @xmath71 at time @xmath81 ( left column ) , @xmath82 at time @xmath83 ( middle column ) , and @xmath84 at time @xmath85 ( right column ) .
the top three rows of panels show the basic characteristics of relativistic reconnection in these simulations .
the tearing instability has produced alternating x - points and magnetic islands , with particles entering the current sheet at x - points and flowing out into the magnetic islands flanking the x - points in both directions .
secondary current sheet structures are identifiable at the locations of island mergers .
although the limited island size at the chosen time of investigation makes the detailed structure of these current sheets unclear , examination of similar secondary current sheets at later times indicates that they have the same x - point and magnetic island structures found in the primary current sheets .
the plot of density @xmath86 in the first row of figure [ fig : overallevol ] shows that in all of our simulations most of particles that enter the current sheet are eventually concentrated in the centers of magnetic islands .
however , in contrast to the simulations of @xcite , we find that these particles are not highly accelerated .
this difference likely arises from the differing initial conditions ; in our simulations , all particles begin at the same temperature , while in those of @xcite , the current sheet population begins with lorentz factors similar to @xmath30 .
particles in the center of the islands are likely to be those that initially entered the x - points ; such particles have had time to isotropize and are more likely to remain in the center of the islands .
particles with high energies that enter the magnetic islands at later times go into the current sheet with a significant initial momentum in the @xmath87 direction and therefore oscillate across the islands , spending most of the time at island edges where their direction reverses .
therefore , the different temperatures found at island centers are a result of the initial conditions , rather than differing reconnection physics . the second row of figure [ fig : overallevol ] , which shows the magnetic energy to particle rest mass ratio @xmath88 , confirms that both the centers of islands and the x - point regions are dominated by particle energy as expected in magnetic reconnection .
detailed examination of the third row of figure [ fig : overallevol ] indicates that particles entering the current sheet at x - points where reconnection is taking place are initially mildly relativistic with @xmath89 , but are accelerated to a typical lorentz factor larger than 10 in the reconnection region .
conservation of energy implies that the typical lorentz factor in the current sheet measured in the lab frame should be approximately equal to @xmath30 if the background particles are cold . for simulation s40 with @xmath82 , and simulation s400 with @xmath84 ,
the results shown in the third row of figure [ fig : overallevol ] are broadly consistent with this expectation ( note that the values of @xmath90 in the center of the current sheet for simulation s400 can be as large as @xmath91 ) .
in contrast , for simulation s4 , the typical lorentz factor in the current sheet is significantly larger than @xmath71 , @xmath92 , but this is consistent with the maximum @xmath93 of the power law discussed in the next section .
the fourth row of figure [ fig : overallevol ] shows the bulk momenta of the plasma in the simulations .
it indicates that there are three locations in which significant bulk flows are present .
the first location is the inflow region of x - points , which has only mildly relativistic flows of approximately @xmath94 in simulation s4 but highly relativistic inflows with @xmath95 in simulations s40 and s400 ; these high velocities are consistent with those found by @xcite in their simulations of low - density plasmas . however , these flows are intermittent and we show later that they do not imply an extremely high rate of energy transfer overall .
the second location of significant flow is in apparent shock waves produced by the collisions of magnetic islands , which produce disturbances in the background material that increase in importance with @xmath30 .
however , these disturbances are not associated with significant energy transfer or radiation .
it should be noted that individual species of particles tend to have oppositely oriented fast flows in the @xmath72 direction in the center of the x - point which are not shown in calculations of bulk motions ; we discuss these in section [ sec : velmapping ] .
the third location of significant velocity flow is the outflow regions of x - points as particles move towards magnetic islands .
these locations are are typically thought to contain the fastest flows produced in relativistic reconnection , with typical flow momenta of @xmath96 found in the simple model of @xcite as well as the simulations of @xcite and @xcite .
figure [ fig : bulkflow ] shows the outflow bulk momentum distribution in the current sheet . in all runs , this distribution is flat up to approximately @xmath97 , and that the cutoff above this value does not depend strongly on @xmath30 . specifically , we find that the typical bulk momentum in the outflow direction @xmath98 defined in the caption to figure 3 increases only slowly with @xmath30 from @xmath99 ( see table 1 ) , which in the highest - magnetisation case of simulation s400 is much smaller than the prediction of @xmath100 .
this result is resolution - independent , varying by at most @xmath101 with macroparticle density and cells / skin depth for all values of @xmath30 , and it is consistent with the finding by @xcite in the case of electron - ion plasmas that the vast majority of the kinetic energy produced in high-@xmath3 reconnection is converted to random motion , rather than ordered bulk outflows .
finally , the fifth row of figure [ fig : overallevol ] shows the total synchrotron power emitted at each location in arbitrary units .
the regions of significant emission are the outflow regions of x - points and the outlying areas of islands , with particularly high emission located at the points where small islands meet larger islands , such as the location @xmath102 , @xmath103 in simulation s4 . unlike the case of the bulk momenta , contributions to the synchrotron radiation from the inflow region of x - points in simulations s40 and s400 are relatively small , indicating that any beaming that may occur there does not have a strong effect on the radiated emission .
this indicates that that the locations of synchrotron emission do not depend qualitatively on the value of @xmath30 .
the oppositely oriented flows of individual species at the center of x - points also have little effect on the overall emission , because they are located in a region of small magnetic field . in our later calculations of synchrotron emission and beaming ,
we focus on simulation s40 in our analysis but confirm that there is no significant qualitative dependence on @xmath30 in our results .
the rate of conversion of magnetic energy to kinetic energy in reconnection is of paramount importance for the fast production of possibly beamed radiation in relativistic reconnection sites . in general ,
reconnection rates are defined by assuming a steady - state reconnection equilibrium with inflow velocity @xmath104 and outflow velocity @xmath105 .
then the normalised reconnection rate is given by @xmath106 , which implicitly assumes that the efficiency of conversion of magnetic to kinetic energy is very high once plasma enters the current sheet .
this assumption is indeed justified by the fact that the current sheet is dominated by the energy of the particles in numerical simulations , as shown in the the plot of @xmath88 in figure [ fig : overallevol ] , but this is not a direct measurement . due to the difficulty involved in calculating the typical velocity of inflow , which varies strongly with position , one instead typically measures the nearly uniform electric field @xmath107 , which gives a normalised reconnection rate of @xcite @xmath108 where the second part of the equation reflects the common assumption that the outflow speed is equal to the alfvn speed in the inflow region @xmath109 which is given in the relativistic case by @xcite @xmath110 this corresponds to an outflow bulk momentum of @xmath111 which is consistent with the predictions of @xcite and the simulations of @xcite which have reconnection outflow lorentz factors @xmath96 but not with our simulations or those of @xcite , which find moderate outflow lorentz factors of @xmath112 which do not depend strongly on @xmath30 .
a somewhat more complex local definition for reconnection rate has been proposed by @xcite , which similarly predicts high outflow lorentz factors at high @xmath30 .
the reconnection rate calculated will not be strongly affected by this disagreement because @xmath113 does not depend strongly on @xmath30 , but the low values of @xmath114 found in some simulations cast doubt on the physical realism of the models .
in addition to this shortcoming , the local reconnection rate may not be a good proxy for the global rate of energy transfer because inflows only occur at x - points rather than throughout the whole current sheet . in the case of low - density plasma ,
@xcite find that the local reconnection approaches 1 at high @xmath30 while remaining close to the typical value of @xmath5 found in most simulations at low @xmath30 , and our results are similar .
but the normalised global rate of energy transfer ( discussed below ) is similar at all values of @xmath30 in our simulations , so the local reconnection rate is not an adequate measure of the global energy transformation rate .
@xcite define a global measure of reconnection rate by choosing @xmath104 as the average velocity of inflow far from the current sheet .
however , this definition is somewhat difficult to apply because the distance from the current sheet must be chosen in a somewhat arbitrary manner and flows resulting from magnetic island collisions far from the current sheet may bias the result .
instead , we propose a definition of reconnection rate based directly on the global characteristics of energy transfer . the maximum rate at which energy transfer can take place by reconnection occurs when at all locations along the current sheet the inflow speed is @xmath115 and all magnetic energy is converted to kinetic energy .
the energy gained in this case accounting for the two directions of inflow into each current sheet and the two current sheets present in the simulation is approximately @xmath116 where @xmath117 is the total energy in the magnetic field at the beginning of the simulation .
then , we can define a dimensionless reconnection rate by normalising the rate of change of total kinetic energy to this value , yielding @xmath118 the generalization of this definition to three dimensions yields the same equation .
we find that the reconnection rate calculated in this way does not vary strongly with time during the period of fast reconnection .
table 1 shows the average value of this reconnection rate in each simulation from the beginning of nonlinear reconnection at time @xmath119 until the end of the reconnection phase , chosen by the time when significant energy transfer stops . using this definition , we find that the reconnection rate does not depend strongly on @xmath30 , but instead occurs at a universal rate of rate @xmath120 which is approximately consistent with the results of past two - dimensional simulations of relativistic reconnection at lower magnetisation ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
we find that despite the significant differences in the current sheet thickness and the relativistic inflow speeds found in our simulations at different @xmath30 , the global reconnection rate does not vary greatly .
this conclusion needs testing in 3d simulations , in which the reconnection rate can be significantly smaller @xcite , but if it holds in that case our finding that the reconnection rate is independent of @xmath30 in relativistic reconnection will be confirmed .
we now consider the particle energy spectrum resulting from relativistic reconnection . in all simulations
, reconnection produces strong particle acceleration , and at around @xmath121 approximately @xmath122 of the magnetic energy in each simulation has been converted to kinetic energy . the resulting spectrum for all simulations is a maxwellian plus a power law , with a cutoff at high energy .
table 1 shows that the power law becomes harder at larger values of @xmath30 , with the index @xmath53 of the power law spectrum @xmath123 changing from 1.65 at @xmath71 to 1.3 at @xmath124 .
this relation is in approximate agreement with the results of previous simulations @xcite as well as the detailed study of @xcite .
the hard power laws found in our simulations indicate that the maximum @xmath0 should reach saturation at late times as a result of energy conservation .
the maximum possible value of the average lorentz factor of the particles is @xmath125 , which yields @xmath126 , where @xmath127 , which is expected to be of order unity , is the fraction of the magnetic energy that is converted into kinetic energy when a part of the plasma undergoes reconnection .
if all of the particles are located in a power law with @xmath128 beginning at @xmath129 and ending at @xmath93 , with @xmath130 , the average lorentz factor is @xmath131 then , we have @xmath132 for @xmath133 as applicable to our simulations , @xmath134 , and @xmath135 , which is consistent with the measured particle energy spectra at all values of @xmath30 , we find that as an upper limit , @xmath136 .
@xcite found in their simulations that the actual saturation value is close to @xmath137 .
we find that the power law spectra in our simulations indeed reach saturation at late times at lorentz factors shown in table 1 , and that the maximum lorentz factor at saturation @xmath93 increases with @xmath30 .
for @xmath71 , the value of @xmath138 is comparable to both predictions ( which are similar ) , for @xmath82 , the value of @xmath93 we find is compatible with the predictions of @xcite . and
finally , for @xmath84 , the value of @xmath93 is much smaller than either prediction .
these results may be explained by the fact that @xmath93 is also limited by the total size and timescale of the simulation . specifically , the maximum possible energy gain for a particle accelerated over a time @xmath139 is given by @xmath140 , which may be combined with the definitions of @xmath30 and @xmath141 to yield @xmath142 ; the local value of @xmath143 in our low density simulations s40 and s400 .
because fast reconnection in our simulations extends over a maximum period of @xmath144 , it is clear that the simulation size is limiting acceleration for simulation s400 , but not for the other simulations .
physically , the fact that @xmath93 is significantly smaller than predicted by equation [ eq : gammap ] is an indication that particles in the background of the simulation and those originally located in the current sheet have received some of the dissipated magnetic energy .
the plot of @xmath90 in figure 2 indicates that in simulation s400 , the whole background plasma has been heated by the reconnection in the current sheet from the initial value of @xmath145 at @xmath146 to @xmath147 at @xmath148 .
this strong heating is likely a result of the boundary conditions becoming important in this simulation . for simulation s40 ,
the disagreement of the value of @xmath93 with the prediction of equation ( [ eq : gammap ] ) arises because most ( @xmath149 ) of the energy converted is transferred to the particles initially located in the current sheet ; thus , less energy is provided to the inflowing particles and they do not reach high @xmath93 .
we now investigate in detail the particle energy spectra for simulation s40 with @xmath82 , including an investigation of its variation with location .
the only significant qualitative difference between the spectra at different values of @xmath30 is the presence of a `` trough '' in the x - point spectrum at intermediate energy for @xmath82 and @xmath84 and the absence of such a trough for @xmath71 , which will be discussed later in this section .
this qualitative difference is present in the simulations of @xcite for low - density plasmas that correspond to high @xmath30 , as well . in this analysis , and in most of our later results ,
we focus on the characteristics of the positively charged ( macro)particles in order to display the differing motions of each species that produce the antiparallel flows that constitute the current in each current sheet
. results for the negatively charged particles are always very similar to the results for positively - charged particles , except that motions in the @xmath72 directions are reversed for a given current sheet .
since we are simulating a pair plasma , we occasionally refer to the positively charged particles as positrons .
the top panel of figure [ fig : partspec ] shows energy spectra in the overall box and in various regions of the box , while the bottom panel of the figure shows several regions of interest for which we take particle energy spectra , which correspond to the center of the current sheet in an x - point ( region a ) , the entire x - point ( region b ) , and a magnetic island ( region c ) .
the top panel of figure [ fig : partspec ] shows that the overall energy spectrum has a relatively weak thermal peak at @xmath89 with a hard power law tail that contains a large proportion of the total energy . the initial energy spectrum at the beginning of the simulation
is also shown , illustrating that most particles with @xmath150 have been significantly accelerated during the simulation .
region a , which corresponds to the center of the current sheet in an x - point , has a peculiar spectrum , consisting of a moderately accelerated population at @xmath151 , which may correspond to the initial acceleration of particles flowing in to the current sheet , and another population at very high energies close to the maximum found in the whole box , with almost no particles found in the trough at intermediate energies .
there is another trough present at @xmath152 , indicating that multiple features of this type exist .
this spectrum may be understood by considering that the particles being accelerated in an x - point are moving in speiser orbits @xcite in which they oscillate across the current sheet while being accelerated by the electric field in the x - point . because the initial inflow is relativistic , the particles that flow into the current sheet at a given time tend to have similar and large momenta in the @xmath87 direction with low random velocity and the particles will undergo few oscillations before leaving the x - point .
thus , particles in the x - point with a given value of @xmath0 will all tend to be at a similar distance @xmath153 from the center of the x - point .
if the distance is high , a trough will be present and the particles will tend to have little momentum in the @xmath87 direction , while if the distance is low the trough will be absent and the particles will have large momenta in the @xmath154 direction .
over longer timescales in which many speiser oscillations can occur , we expect these features to disappear as the particles no longer move coherently .
the spectrum in region b indicates that when a larger portion of the current sheet is included , the trough disappears because the particles with high oscillation amplitude are now included in the spectrum .
the spectrum in the magnetic island ( region c ) is very different than that found in the x - points . instead of having a thermal peak at the background average energy of @xmath155 ,
the peak is at a significantly larger value of @xmath156 , which is similar to the average kinetic energy in the center of the island where the density is extremely high . above this peak ,
the particle energy spectrum declines faster than the overall power law of @xmath157 , but bumps are present at @xmath158 and at @xmath159 which correspond to locations in region c with similar values of @xmath160 , corresponding approximately to the colors yellow and red in the bottom panel of figure [ fig : partspec ] .
the magnetic island appears to have relatively few particles with @xmath161 , indicating that the highest - energy particles are found mainly in the x - points .
as long as fast reconnection is still taking place and the highest - energy particles have not reached saturation , the particles remaining in the x - points will always have a higher energy than those in the islands because they have spent a longer time being accelerated .
overall , these spectra indicate that most particle acceleration takes place in reconnection regions , in which hard power laws are produced .
the magnetic islands mostly contain particles that have been only slightly accelerated in the x - points , which are in an approximately thermal distribution peaked at @xmath156 , and high - energy particles contribute less of the total particle energy in these regions .
the overall energy spectrum of particles that have entered the current sheet and been accelerated will be some combination of the spectra shown for regions b and c. this spectrum peaks at around @xmath162 , which is around @xmath163 times the initial average particle energy , and has a hard power law spectrum that extends to high energies ; the exact index of the power law depends on whether more particles are located in the x - points or the islands .
we expect these accelerated particles to dominate the high - energy synchrotron emission , so we study in detail the particle distribution and the resulting radiation from these two regions in the subsequent sections . at late times in the simulation ,
as energy conversion through reconnection becomes slower , the islands become the dominant reservoir of the highest - energy particles due to either saturation of @xmath93 or the influence of the boundary conditions , with most of these particles being located where outflows from x - points meet the outer layers of the island .
nevertheless , particles in the x - points still have a higher average energy than those in the islands do at late times . because this saturation may be a result of artificial boundary conditions , we assume that the particle distributions at the earlier time focused on in this work are a more accurate representation of particle energy spectra in large - scale reconnection . although we have not found bulk beaming in our simulations , it is still possible that strong energy - dependent kinetic beaming @xcite will be present . in this section ,
we discuss the average momentum and the spread in the direction of the momentum ( i.e. , the beaming ) as a function of energy for particles at various locations in the simulation .
we exclude particles that began in the current sheet from our calculations , just as in section [ sec : partaccel ] .
particles that are not significantly accelerated can have strong momentum flows in three locations , as explained in section [ sec : overallevol ] .
however , particles that are significantly accelerated with @xmath164 only have significant bulk momentum , and significant beaming , in the x - point and in the outflow regions from those x - points in the current sheet .
although bulk flows are not present in magnetic islands .
we investigate whether there is beaming of high - energy particles in these locations , because they dominate the bolometric synchrotron flux .
we first study the beaming of particles in the x - points and their adjacent outflow regions .
we focus on the beaming of particles in region a to include in the analysis the spread in the @xmath87 direction resulting from particles entering the x - point in the opposing @xmath154 directions . in the x - point outside of this region , there is a net momentum in the @xmath154 direction and very little spread in that direction , which is unrepresentative of the particle distribution in the x - points .
figure [ fig : velmapping ] shows the characteristics of the momentum distributions of the positrons ( positively charged macroparticles ) as a function of @xmath0 for particles located in the middle x - point in the left current sheet in region a. the top panel of the figure shows the direction of bulk momentum in the @xmath165 plane ; the average momentum of particles in the @xmath87 direction is close to 0 everywhere in region a because the inflows are of similar strength on both sides of the x - point .
it indicates that the direction of the bulk momentum changes gradually from the direction of the acceleration provided by the electric field ( @xmath166 ) towards the direction of the outflow ( @xmath73 ) as one moves from the center of the x - point towards the magnetic islands .
positrons with higher energy are deflected somewhat less towards the outflow direction as they move away from the center of the x - point .
this makes sense because such particles have significantly higher inertia following acceleration than lower energy particles do .
the second panel of figure [ fig : velmapping ] shows the ratio of the bulk momentum of particles falling within a given energy band to the central lorentz factor of that band . at all locations ,
the bulk momentum is similar to the total lorentz factor , indicating that almost all of the particles are moving in the same direction at a given location .
note that the points with @xmath167 indicate that much of the bulk momentum is in the higher - energy particles within the band , and as a result the maximum possible bulk momentum is closer to @xmath168 where @xmath169 is the maximum value of the lorentz factor in the energy band . in all cases,@xmath170 , so this high degree of alignment is indeed physically possible .
the third panel of figure [ fig : velmapping ] shows the ratio between the standard deviation @xmath171 of the direction of positron momentum to @xmath172 , the minimum opening angle of radiation arising from a distribution of particles with lorentz factor @xmath0 .
we discuss this calculation in an appendix .
this panel shows that accelerated particles in the x - point typically are beamed within @xmath173 , with no significant dependence of the normalised value on @xmath0 .
a somewhat higher spread is found for the case @xmath174 , which also has a significantly lower normalized bulk momentum in the middle panel of figure [ fig : velmapping ] , but this is a discrepancy of only a factor of @xmath112 .
the increased spread may occur because there are few particles in region a at this value of @xmath0 , which is in the trough discussed in the previous section .
these results indicate that accelerated particles located in the reconnection region are highly beamed , but that the direction of that beaming changes greatly as one moves from the center of the x - point to its edge . thus , the overall particle distribution and the resulting energy distribution should be strongly kinetically beamed in the plane of the current sheet ( the @xmath165 plane ) , but not in other directions .
we now investigate beaming in magnetic islands , focusing on region c at @xmath175 , @xmath176 ; we include particles that began in the current sheet , because they have a significant contribution to the total particle density at the island centers .
unlike the case in the x - points , the direction of the average particle velocity does not vary in a continuous fashion throughout the region .
flows in the @xmath177 plane tend to be towards the center of the island , but counterflows are often present as well .
similarly , the @xmath74 component of the average velocity tends to be oriented in the direction of particle acceleration in the x - points ( the @xmath166 direction for the positrons studied here ) , but it is reversed in some locations in the island . in general , particles are oriented relatively close to the @xmath166 direction in the outlying areas of the islands that are not in the path of outflows from the x - points . while the direction of beaming of particles in the magnetic island varies in irregular fashion , the magnitude of beaming varies with the radius from from the center of the island in a manner consistent with the increase in the magnitude of the overall bulk flows with radius ( figure [ fig : overallevol ] , middle column , fourth row ) .
the top panel of figure [ fig : islandradius ] shows that the normalized flow velocity increases with distance from the island center , but it is generally significantly smaller than the maximum values found in the x - point .
the bottom panel shows that the normalised beaming angle decreases with radius , but it is still generally quite large at all radii .
the outer regions in which this stronger beaming is present also contain higher typical particle energy ( figure [ fig : partspec ] , bottom panel ) , so this result is consistent with earlier simulations that indicate that particles that spend a long time in the island are thermalized and isotropized @xcite , while particles that spend more time being accelerated in the x - point and enter the island later do not have time to undergo this process .
figure [ fig : islandgamma ] shows the variation of the two parameters averaged over the whole island ( region c ) with @xmath0 .
it indicates that the normalised beaming angle @xmath178 increases quickly with @xmath0 , so that high - energy particles are not more strongly beamed than low - energy particles .
the normalised bulk momentum increases slowly with @xmath0 , but does not go above 0.8 ; smaller beaming angles correspond to a normalised momentum very close to 1 , as shown in figure [ fig : velmapping ] .
the change in behavior of both parameters above @xmath179 is a result of the fact that the highest - energy particles in the islands are those that began in the current sheet .
such particles could began their acceleration earlier than those that began in the background plasma , but still have time to be isotropized in the islands .
alternatively , they might have been reaccelerated during an island merger @xcite or as result of island contraction @xcite . overall
, we do not find strong beaming of the particles anywhere in the magnetic islands except for the highest - energy particles in the regions where the outermost part of the island meets the reconnection outflows .
instead , we find a very moderate average beaming in the @xmath166 direction , the original direction of acceleration in the x - point .
however , it is uncertain if this will mean there is no significant beaming of the resulting synchrotron radiation . unlike the case of the x - points , where the majority of radiation is likely
to be emitted where the magnetic field is @xmath180 and the field varies slowly with @xmath75 , the field in the island varies quickly with location . at @xmath181
the field can be as high as @xmath182 , while the field is typically @xmath183 at the center of the island .
thus , the radiation from the island may be highly beamed if specific locations on the outskirts of the island dominate the overall emission , which can not be ruled out . to clarify the conclusions of this section , we sketch a schematic particle trajectory and explain how it is consistent with the above conclusions .
figure [ fig : reconschema ] shows the trajectory of an electron that enters the current sheet at large @xmath74 .
it oscillates across the current sheet in the @xmath154 direction in a speiser orbit @xcite as it is accelerated by the electric field in the @xmath184 direction .
eventually , it moves in the @xmath185 direction towards a magnetic island and is accelerated in that direction as it moves .
thus , its velocity moves from the @xmath184 direction towards the @xmath185 direction as it moves towards the magnetic island in accordance with the top panel of figure [ fig : velmapping ] , except that the initial direction of acceleration is changed from @xmath166 to @xmath184 because the charge of the electron is negative .
once the electron reaches the island , it is moving mostly in the @xmath185 direction , indicating that particles entering the island from the x - point are highly beamed .
afterwards , the electron is deflected into a randomised trajectory , indicating that the overall particle distribution in the magnetic island is not highly beamed , as shown in figure [ fig : islandgamma ] .
it remains mostly on the outskirts of the island , consistent with the finding in figure [ fig : overallevol ] that accelerated particles are mostly found in current sheets and the outskirts of magnetic islands .
this section discusses the beaming of particles and radiation at various energies .
some differences exist between the beaming results in simulation s4 with @xmath71 and those in the other two simulations due to the presence of the oscillations in the energy spectrum in the x - point at higher @xmath30 discussed in [ sec : partaccel ] .
however , the oscillations do not have strong qualitative effects , so we focus on the results in simulation s40 with @xmath82 .
this section is organized as follows .
we first show in section [ sec : regimes ] that the beaming in fast and slow cooling regimes can be probed by calculating the beaming of particles and the resulting radiation within an x - point and an island , respectively .
then , in sections [ sec : xpointmaps ] and [ sec : islandmaps ] we discuss the qualitative features of beaming maps in the x - point and the island .
section [ sec : quantbeam ] compares quantitative measures of the beaming of particles and radiation for both regions with previous work .
finally , section [ sec : schematicbeam ] schematically summarizes the characteristics of radiation produced in the fast and slow cooling regimes . in our simulations , we do not directly determine the effect radiative cooling will have on the particles in a realistic reconnection configuration . however , because particles can be accelerated without bound in the x - point during reconnection due to their focusing toward the center of the current sheet by electric acceleration @xcite , particles are unlikely to lose the bulk of their energy until they reach the edge of the x - point and enter the island .
therefore , the beaming characteristics of the radiation emitted by a particle will be determined by whether it cools before its momentum changes significantly in the magnetic field near the entrance to the island ( fast cooling ) or afterwards ( slow cooling ) .
note that this cooling time is fast or slow compared not to the dynamical time for evolution of the overall reconnection structure but instead to the time for an individual particle to change its properties by motion within that structure .
if the highest - energy particles accelerated in the x - point are in the fast cooling regime , the beaming characteristics of particles and radiation at high energy are the same as those for the instantaneous particle and radiation distributions in region b ( an x - point ) .
otherwise all particles will radiate in the slow cooling regime , and the particles and radiation will be beamed similarly to the instantaneous distributions of the particles and radiation in region c ( a magnetic island ) .
we now calculate the limiting lorentz factor and emission energy at which the transition from slow to fast cooling occurs .
a particle with lorentz factor @xmath0 accelerated in an x - point with electric field @xmath186 gains energy at a rate @xmath187 while the same particle radiating in a magnetic field @xmath68 loses energy at a rate @xmath188 the critical burnoff lorentz factor at which the acceleration and cooling are equal is @xmath189 we can rewrite equation ( [ eq : coolingfirst ] ) in terms of @xmath190 as @xmath191 the cooling time can then be expressed in terms of the relativistic cyclotron frequency @xmath192 as @xmath193 because the time required for a particle s momentum to be modified significantly is just the particle s gyration time @xmath194 , a particle will be in the fast cooling regime if @xmath195 the corresponding fast cooling limit for the characteristic photon energy may be found by combining this equation with the synchrotron frequency equation ( [ eq : critomega ] ) , setting @xmath196 , and using the relation @xmath197 for photons to obtain @xmath198 where the latter estimate applies for physical electrons .
this indicates that all emission below this energy will come from magnetic islands in the slow cooling regime .
as we see in the following sections , this has major implications for the beaming of this radiation . in both regions ,
we calculate the angular distribution of particles and of the synchrotron radiation in three energy bands .
the angular distribution is given as a function of the longitude @xmath199 ( the angle within the current sheet ) and latitude @xmath200 ( the angle perpendicular to the current sheet ) , where @xmath201 is the unit vector corresponding to a given direction .
the frequency @xmath58 of radiation is normalized to @xmath202 , the fiducial peak synchrotron frequency of a particle with @xmath203 moving perpendicular to the background magnetic field @xmath37 .
thus , the radiation spectrum from a particle with lorentz factor @xmath0 is expected to peak at @xmath204 and the particles in each interval of lorentz factors are expected to emit in the corresponding radiation energy band .
however , particles at higher energy in each range often contribute more synchrotron radiation in each band than those at low energy do , and many particles emit at significantly higher or lower magnetic field than @xmath37 , so that the energy ranges do not precisely correspond .
therefore , we also calculate the angular distribution of the total synchrotron flux as a function of direction from particles in each interval .
figure [ fig : xpointmaps ] shows aitoff projections of the angular distribution of the positrons in three energy bands , the total synchrotron flux from those particles , and the synchrotron flux from all particles @xmath205 integrated over three logarithmic energy bands in region b , corresponding to an x - point .
this radiation corresponds to a fast cooling regime . in the x - point , the particle energy distribution @xmath206 shown in figure [ fig : partspec ] is nearly flat above @xmath207 , so that the synchrotron energy per logarithmic interval in the two lower frequency ranges is dominated by the emission from particles at significantly higher @xmath0 .
therefore , the particle and radiation angular distributions are significantly different for the two lowest energy bins .
the particle distribution in the lowest energy band ( top row , left column ) , corresponding to @xmath208 ( and the synchrotron radiation from particles in that range , middle column ) is highly anisotropic , but the percentage of particles that are at such a small @xmath0 in the x - point is extremely low ( figure [ fig : partspec ] ) , and they do not contribute significantly to the synchrotron radiation at any frequency .
comparison of the second and third rows of figure [ fig : xpointmaps ] indicates that the particles ( left column ) and the resulting radiation ( middle column ) become more anisotropic as the energy increases , with all of the distributions becoming focused towards the equator at @xmath66 .
the particles are concentrated in two populations .
the first is located at @xmath209 and located approximately at the value of @xmath210 predicted from @xmath211 calculated at the edge of the x - point . for the range @xmath212 ( @xmath213 ) these particles
are located at @xmath214 , which is close to the value of @xmath215 for @xmath216 found in figure [ fig : velmapping ] for particles in the outflow regions at the edge of the x - point .
the second large group of particles has a significant velocity in the @xmath154 direction but is mostly moving in the @xmath166 direction in the @xmath165 plane , with its motion characteristic of speiser orbits in the center of an x - point . for @xmath213 ,
the particles moving in the @xmath166 direction are also in speiser orbits , but they have a much smaller momentum in the @xmath154 direction because they have spent more time being accelerated @xcite .
these particles are expected to produce little radiation because they spend much of their time in the current sheet where the magnetic field is low , and indeed the expected radiation for this population ( middle column , bottom row ) is small .
the synchrotron flux distributions in the three frequency bands are significantly different than those of the particles as a result of the sharply peaked energy distribution of the particles .
the angular distribution in the frequency band @xmath217 is clearly similar to that of the particles in the next higher energy band @xmath212 . for @xmath218 two components
are present .
one has relatively low amplitude and is similar to the distribution of the corresponding particles at @xmath212 .
the second is concentrated only along the equator with a smaller range of @xmath210 and has a significantly higher amplitude .
this radiation is likely produced by higher - energy particles moving in speiser orbits , which peak in this frequency band because they radiate in significantly lower magnetic field in the x - point .
finally , for @xmath219 the radiation is extremely beamed on the equator at @xmath220 .
this beaming is even stronger than that expected from the distribution of high - energy particles , but can be explained by considering that slightly lower - energy particles will only contribute to this frequency band if they are at the edge of the x - point where the magnetic field is close to @xmath37 and the beaming is in this direction .
overall , in the fast cooling case in which most emission comes from the edge of the x - point , we expect the observed radiation to be highly beamed . figure [ fig : islandmap ] shows aitoff projections of the angular distribution of the positrons in three energy bands , the total synchrotron flux from those particles , and the synchrotron flux from all particles @xmath205 integrated over three logarithmic energy bands in region c , corresponding to a magnetic island .
we include particles that began the simulation in the current sheet in our analysis , because they can have a significant portion of the total energy at the center of the island .
this radiation corresponds to a slow cooling regime in which radiation is emitted well after particles exit the x - point and enter the island .
the distributions are significantly simpler than those found in the x - point . at low energies ,
all three distributions are nearly isotropic , although there is a slight concentration of synchrotron flux in the @xmath166 direction . for the middle energy band , the particle distribution for @xmath212
is concentrated so that most particles have no significant momentum in the @xmath72 directions .
it is uncertain why this is the case , but the concentration is in any case fairly small and a significant number of particles are oriented in the @xmath166 direction .
the total synchrotron radiation from these particles and the synchrotron radiation in the corresponding energy band @xmath218 are more isotropic , with a slight concentration in the @xmath166 direction that may be explained by the fact that the locations where most particles are oriented in the @xmath166 direction are generally in the outlying areas of the island , where the magnetic field is strong .
the particles in the highest energy range @xmath213 are highly anisotropic and oriented in the @xmath73 direction , and the radiation from these particles has an identical distribution .
these are clearly the highest - energy particles from the population entering the island from the outflow regions of the x - points , which are shown in figure [ fig : partspec ] to have the highest average energy in the islands .
note , however , that these particles are not as concentrated in the @xmath64 direction as the highest - energy particles in the x - point .
the synchrotron radiation in the corresponding frequency band @xmath221 is considerably more spread out in both @xmath210 and @xmath64 , probably as a result of the radiation of lower - energy particles in locations with large magnetic field @xmath222 that contribute significantly to the observed distribution .
overall , we expect the emission from islands to be relatively unbeamed except at the highest energies .
this high - energy radiation comes from particles located where outflows from x - points meet the magnetic islands .
however , in the physical slow cooling case this high - energy particle population will probably not radiate until its momentum has been significantly altered by the field of the magnetic island .
if this is true , we expect no significant beaming at any frequency for this slow - cooling case .
because agn and grbs emit at energies significantly below @xmath223 mev in the slow cooling regime , models of these systems that require strong beaming may need to be reconsidered .
however , this conclusion holds only if acceleration is a single - step process .
if significant reacceleration that produces beaming occurs in magnetic islands , radiation may be strongly beamed below this limit . because particles in the slow cooling regime defined in this paper can still cool quickly compared to the dynamical time of the overall reconnecting region
, there is a regime in which reaccelerated particles radiate efficiently below 160 mev .
however , it is not clear that acceleration mechanisms that work in magnetic islands , such as island contraction @xcite and island coalescence @xcite are likely to produce strong beaming .
we now make more quantitative comparisons of the beaming of particles and radiation in both regions . because the angular distributions of the particles in a given energy band and the total emission from those particles ( left and middle rows of the previous figures ) are very similar
, we only compare the particle beaming as a function of @xmath0 to the synchrotron distribution from all particles in a region as a function of @xmath224 . because all of the distributions appear to have much larger spread in @xmath210 than in @xmath64 at high energies , we calculate two standard deviations @xmath225 and
@xmath226 of the angular distributions about their means @xmath227 and @xmath228 instead of using a single parameter ; we discuss the method of calculation in an appendix .
to correct for the presence of double peaks in the angular distributions for opposite signs of @xmath210 , we restrict our calculation to the region @xmath229 .
figure [ fig : spreads ] shows the variation of the standard deviations @xmath225 and @xmath226 of the angular distributions about their means @xmath227 and @xmath228 as a function of energy and the spectral energy distribution for both positrons and their synchrotron radiation in both regions .
the left column of the figure shows that the value of @xmath225 for the particles is insensitive to @xmath0 in both regions ( although there is a slight decrease in this parameter in the island at large @xmath0 ) , and that @xmath225 for the radiation is also insensitive to @xmath224 below the peak of the synchrotron spectrum , which is typically at @xmath230 . above this peak , @xmath225 declines with increased frequency @xmath224 in both regions , but only becomes comparable to the small values of @xmath231 rad found for @xmath226 in the case of the x - point . in the x - point
( region b ) , the reason for this decline is probably that particles at high energies that are moving in the @xmath166 direction can no longer contribute significantly above the peak because they are in regions of low magnetic field . in the island ( region c ) ,
the reason for this decline and the slight decline in the particle beaming is probably that the particle population that is entering the island from the adjacent x - points becomes more prominent , and there is less contribution from particles oriented in the @xmath166 direction .
@xmath226 depends much more strongly on energy for both the particles and the radiation . in the x - point ,
@xmath226 declines with @xmath0 for @xmath232 , but an oscillatory signature is present in which less spread is present at locations corresponding to the troughs present in the particle energy distribution for region a in figure [ fig : partspec ] , and more spread is present at other locations .
the oscillatory signature is also present in the radiation from the x - point , producing a slight rise in spread from @xmath233 , but it is much less significant because the particle energy distribution in the x - point is strongly peaked at @xmath152 and most of the radiation comes from there . in the absence of the oscillation , we expect the dips in spread of the particles to disappear , but it is difficult to calculate the expected trends . for particles moving on speiser orbits with amplitude larger than the current sheet width , @xcite predicts that the spread should decrease as @xmath234 , which is roughly consistent with the decline from @xmath235 at @xmath236 to @xmath237 at the peak in spread at @xmath238 . in the case of the radiation ,
the value of @xmath226 decreases approximately as @xmath239 for @xmath240 , which is again roughly consistent with the expected results from speiser orbits because @xmath241 .
a decline in @xmath226 with @xmath0 also occurs in the island , but only for @xmath161 above the peak of the particle energy distribution , and it only decreases to @xmath242 rad at @xmath243 ; the steep decline at @xmath244 may be a result of small number statistics .
the corresponding decline in @xmath226 with @xmath58 , which is approximately proportional to @xmath245 , begins above the peak of the radiation distribution at @xmath246 and continues to high energies , but it ls much less steep than the @xmath244 decline in the particle energy distribution . again , sharp features in the particle beaming are smoothed out in the radiation beaming .
we now compare our results for the beaming of radiation with other work . because we are interested in high energy radiation that contains significant portion of the total energy , we focus on the values of @xmath226 and @xmath225 one decade above the peak of the distribution in each region .
which is approximately at @xmath247 for region b ( the x - point ) , and @xmath246 for region c ( the island ) .
the beaming in the simulations of @xcite without radiative feedback , and @xcite , which include radiative feedback , indicates that high energy radiation within one decade in frequency above the peak in @xmath248 is beamed within a solid angle of @xmath249 , the solid angle within which the particles with @xmath229 are beamed in our simulations may be estimated as @xmath250 , so the beaming solid angle for all @xmath210 in our simulation s40 is @xmath251 , where the standard deviations are expressed in radians in the equation . in region b ( the x - point ) , at @xmath247 , @xmath252 rad , @xmath253 rad , and @xmath254 . in region
c ( the island ) at @xmath246 , @xmath255 rad , @xmath256 rad , and @xmath257 .
thus , our results indicate that strong beaming is present at high energies in both cooling regimes , and our results are consistent with previous work but indicate slightly stronger beaming .
however , in the slow cooling case the narrowing of the spread at high energies due to particles entering the island from the x - points may not occur because radiation is produced well after these particles leave the x - points and are deflected by the island s magnetic field . in this case
, the spread at high energies will remain similar to that at low energies with @xmath258 rad , and @xmath259 .
thus , it possible that high - energy radiation produced in relativistic magnetic reconnection is strongly beamed only in the fast - cooling case . to clarify the conclusions of this section , we schematically show in figure [ fig : beamingschema ] the magnitude and opening angle of the radiation coming from particles that are accelerated to @xmath260 in the fast cooling and slow cooling cases , although . in both cases ,
the opening angle of the radiation distribution from these particles in the x - point is @xmath261 rad .
the figure for the fast cooling case also illustrates that the amount of radiation increases and the opening angle decreases as @xmath0 is increased . in the x - point , the direction of beaming varies from the direction of acceleration ( @xmath166 ) towards the direction of the outflow towards the magnetic island ( @xmath73 ) as found in our analysis of particle beaming in figure [ fig : velmapping ] .
the amount of radiation emitted by each particle increases with distance from the center of the x - point due to the variation of the @xmath87 component of the magnetic field .
in contrast , in the magnetic islands , the opening angle of the radiation distribution is very wide , approximately 1 rad as seen in figure [ fig : islandgamma ] , and the direction of beaming varies greatly with location .
we show four representative beaming directions at various locations in the islands . in the fast cooling case ( left )
, each particle loses most of its energy at the edge of the x - point in a very short time , and emits little energy thereafter .
the blue cones in the islands correspond to emission from particles with @xmath0 of order unity , which is relatively unimportant .
this means the high - energy radiation in the fast cooling case is similar to the instantaneous emission from the x - point , which is dominated by particles at its edge . in the slow cooling case ( right ) ,
particles lose very little energy before reaching the island .
because they spend most of their time in the island and the magnetic field is high there , the radiation emitted from the island dominates the observed radiation .
this is again in agreement with our previous statements that the islands are responsible for most emission in the slow cooling case .
note that in the dynamical simulations of @xcite , the low radiative efficiency in the slow cooling case is likely because the emission of the particles during the dynamical phase is much less than that which will occur afterwards in the quiescent phase , consistent with this conclusion .
we have carried out particle - in - cell simulations of relativistic magnetic reconnection at three values of the background magnetisation @xmath30= 4 , 40 , and 400 , and calculated the particle and synchrotron radiation beaming resulting from the reconnection process . given the limits of our simulations not all these results are necessarily generic to systems of larger size that exist over longer durations or for systems with larger @xmath3 .
therefore we first summarize our findings in these specific simulations we carried out and then we identify the results that we expect to be generic and relevant for astrophysical systems such as grbs and agns . *
the reconnection rate and the characteristics of the angular distributions of particles and radiation do not change significantly with the background magnetisation @xmath30 , despite the significantly thicker current sheet structure and the much higher inflow velocities in low density plasmas with high values of @xmath30 that are produced for our choice of initial conditions .
this indicates that most of the important aspects of relativistic reconnection do not depend strongly on the background magnetisation so long as it is larger than 1 , and that the choice of an initial equilibrium with high density contrast at high magnetization does not change these aspects of reconnection either .
* particle acceleration in our simulations is efficient , producing a hard power - law tail in the particle energy spectrum of the form @xmath262 - with index @xmath263 that becomes harder as @xmath30 increases .
the high - energy cutoff @xmath93 of the power law distribution also increases with @xmath30 in accordance with energetic constraints .
* particles are accelerated into this power - law distribution primarily in x - points , with no significant acceleration taking place within the islands .
once particles are accelerated they leave the x - point and move towards the islands . in our simulations , which are limited in size and time , the highest energy particles still reside at the end of the simulations in the x - points , while particles with @xmath264 are just entering the islands . *
the outflow of accelerated plasma from the x - points has a complicated beaming pattern and can not be regarded simply as a relativistic bulk flow - as commonly envisaged in idealized analytic models .
in fact the bulk outflow , @xmath265 , is mildly relativistic for all values of @xmath30 we studied .
instead the beaming size is @xmath0 dependent and beaming directions vary from one position to another .
particles with lorentz factor @xmath0 are typically beamed within @xmath266 .
the direction of the beam rotates with location from the direction of the electric field in the center of the x - point to the direction of outflow towards the island at the edge of the island .
the result of this rotation is that the integrated population of particles with a lorentz factor @xmath0 that are still in the x - points ( i.e. , havent entered the islands yet ) forms a fan narrowly confined to the plane of the current sheet ( the y - z plane ) .
once entering the islands , the accelerated particles gradually lose their beaming , so the island population is mostly unbeamed . *
the resulting synchrotron radiation pattern follows suit .
radiation from the x - points is highly beamed .
since the magnetic field in the center of the x - point is weak radiation is significant only from particles that are leaving the x - points towards the island . as a result the narrow fan of particles produces a radiation pattern that is dark in the center and becomes brighter towards the edges of the fan . because the fan width depends on @xmath0 higher energy radiation is more narrowly beamed . in our simulation
the total beaming at the highest synchrotron frequencies of radiation coming from the x - points is @xmath267 of the sky . in magnetic islands , high - energy particles
are unbeamed of so is the radiation they generate .
an exception is the radiation from particles that are just entering the islands from the x - points and did not have time to isotropise . in our simulation , where the highest energy particles did not have time to isotropise in the islands ,
their radiation in the strong magnetic field of the islands produces a strongly beamed synchrotron component from the islands themselves .
based on these results we can draw some generic conclusions that are relevant for astrophysical settings . most and possibly all particles are accelerated in the x - points and flow into the islands where they spend most of their life .
exceptions are particles that cool down radiatively within the x - points .
the radiation from the x - points is beamed into a narrow fan while radiation from the islands is not strongly beamed .
the combination of the beaming pattern from different regions and the fact that most particles spend most of their lifetime in islands implies that synchrotron radiation from reconnection sites is generally rather isotropic .
highly beamed radiation may be generated only by particles that cool down rapidly before reaching the islands . as we have shown
these particles emit around or above the burn - off limit , which is @xmath268 mev in the rest frame of the reconnection site .
therefore , models of grbs and agn that require strong beaming at lower energies may need to be reevaluated .
we thank l. sironi for useful discussions of the overall evolution of these simulations in comparison to his , m. hoshino for discussions of the possible effects of 3d physics on our simulations without guide field and suggestions regarding the physics of fast inflows at high magnetization , and p. binyamini and j. granot for discussions of the applications of magnetic reconnection to the jet - within - a - jet model of grbs .
tp was supported by the i - core center for excellence in research in astrophysics , and a cnsf - isf grant .
dk and en were partially supported by an erc starting grant ( grb / sn ) , isf grant ( 1277/13 ) and an isa grant . here ,
we discuss our calculation of the directional standard deviation @xmath171 of the particles in the x - points and islands in section [ sec : velmapping ] , and the two - dimensional directional standard deviations @xmath226 and @xmath225 for both particles and radiation in section [ sec : beaming ] . to calculate @xmath171
, we assume that the directions of the particle momenta are distributed in a von mises - fisher distribution @xcite of the form @xmath269 , where @xmath270 is the concentration parameter that represents how clustered the directions are , and @xmath271 is the spherical angle relative to the mean direction .
this distribution is the spherical equivalent of the gaussian distribution . treating the direction of motion as a vector @xmath201 on the unit sphere , the confidence interval within which the mean direction is @xmath272 likely to fall
is given by @xmath273 here @xmath276 is the ratio of the summed vector to the number of vectors , which is equivalent to the normalised momentum calculated above but for directions only , and @xmath277 indicates a summation over all particles .
we take this value of @xmath171 to be an accurate estimate of the spread of the particle direction about the mean . under the assumption that the angular distribution is a bivariate von mises - fisher distribution in the @xmath64 and @xmath210 directions , accurate confidence intervals for the uncertainty in the mean direction
may be calculated merely by calculating the standard deviations in the corresponding cartesian coordinates @xmath278 and @xmath279 @xcite where @xmath61 is the number of particles or the amount of synchrotron flux emitted in a given direction within the energy band , and @xmath281 is the surface area element .
we again take the confidence intervals to be accurate estimates of the spread in direction about the mean . | relativistic reconnection has been invoked as a mechanism for particle acceleration in numerous astrophysical systems .
according to idealised analytical models reconnection produces a bulk relativistic outflow emerging from the reconnection sites ( x - points ) .
the resulting radiation is therefore highly beamed . using two - dimensional particle - in - cell ( pic ) simulations , we investigate particle and radiation beaming , finding a very different picture .
instead of having a relativistic average bulk motion with isotropic electron velocity distribution in its rest frame , we find that the bulk motion of particles in x - points is similar to their lorentz factor @xmath0 , and the particles are beamed within @xmath1 . on the way from the x - point to the magnetic islands , particles turn in the magnetic field , forming a fan confined to the current sheet .
once they reach the islands they isotropise after completing a full larmor gyration and their radiation is not strongly beamed anymore .
the radiation pattern at a given frequency depends on where the corresponding emitting electrons radiate their energy .
lower energy particles that cool slowly spend most of their time in the islands , and their radiation is not highly beamed . only particles that quickly cool at the edge of the x - points generate a highly beamed fan - like radiation pattern .
the radiation emerging from these fast cooling particles is above the burn - off limit ( @xmath2 mev in the overall rest frame of the reconnecting plasma . )
this has significant implications for models of grbs and agns that invoke beaming in that frame at much lower energies . |
chemotaxis @xcite and brownian motion @xcite belong to the key processes which govern the motility of microorganisms ( e.g. bacteria , amoeba and endothelial cells ) @xcite . in the simplest approach ,
the microbe smells " a chemical and moves along the gradient of the concentration field of the chemoattractant in order to reach efficiently the secretion source of the chemical . the opposite case of negative chemotaxis is realized in case the microbe intends to avoid another object which is secreting the chemical @xcite .
this chemotactic drift is superimposed to stochastic motion due to fluctuations associated with the active process of self - propulsion of the microorganism @xcite .
chemotaxis can lead to clusters of aggregated bacteria @xcite which are still emitting chemoattractant .
here we study the self - coupled situation where the microorganism smells " itself , i.e. it reacts chemotactically to its own secreted chemical .
autochemotaxis " is realized in aggregated clusters of different bacteria if the aggregate is considered as a net particle .
another realization is a single bacterium which has both an emitter and a sensor of the same chemical .
tsori and de gennes @xcite have studied a simple model for this situation in different spatial dimensions @xmath1 and find self - trapping of the bacterium in its own chemoattractant cloud for @xmath2 but not for @xmath3 .
this means that for low dimensionality the bacterium is fooled by its own secretion such that it is getting localized for long times . in a subsequent numerical study of a model microbe coupled to its own
chemoattractant secreted at constant rate , grima @xcite calculated the long - time dynamics in various dimensions @xmath1 and found long - time diffusive behavior even for @xmath2 in contradiction to ref .
grima also studied the case of negative chemotaxis and finds in all dimensions long - time diffusion or ballistic motion depending on the strength @xmath4 which couples the microbe s driving force to the gradient of the chemical concentration field .
grima predicts that if @xmath5 exceeds a critical value @xmath6 , then the long - time motion is super - diffusive . in this paper
we revisit autochemotaxis by studying a model which is similar but not identical to that proposed by grima @xcite . in the model of grima ,
a finite global extinction rate ( @xmath7 ) of the secreted chemical is present .
here we focus on the case @xmath8 , which can be realized in experiments ( see e.g. @xcite ) . by extensive computer simulations ,
we study different spatial dimensions @xmath0 and both cases of positive and negative chemotaxis .
our results are summarized as follows : consistent with grima @xcite , we find for positive chemotaxis ( i.e. for a coupling parameter @xmath9 ) long - time diffusive motion .
in addition we find _ dynamical transients _ before reaching the long - time diffusive limit . during the transients , the dynamics of the microorganism
is strongly reduced resulting in almost dynamical arrest . the crossover time from intermediate arrest to long - time diffusion grows strongly with the coupling @xmath10 .
therefore the idealized analysis of tsori and de gennes who predicted self - trapping ( i.e. a complete dynamical arrest ) is manifested by long transients for very strong couplings @xcite .
the averaged mean - square displacement as a function of time does not exhibit a universal slope in this transient regime , but the actual mean slope is decreasing with an increase in the coupling @xmath10 . the transient behavior is most pronounced in one dimension but weakened considerably in three dimensions . for negative chemotaxis ( i.e. dynamical self - avoiding of the microbe ) , on the other hand , we find transient ballistic motion in all dimensions . in @xmath11
we observe a long - time diffusion for all coupling strength . such long - time diffusive motion in @xmath12 , though not directly observed in the simulation , is possible by finite probability of changing the direction of motion ( left to right ) at long time scales for non - zero temperature .
according to refs .
@xcite , the critical coupling @xmath6 above which ballistic long - time behavior is found depends on the global extinction rate @xmath7 but stays finite when @xmath13 .
one reason for the discrepancy is because noise has not been completely included in the earlier treatments @xcite , while solving for the integrals concerning the non - markovian chemotactic force . we take note of the effect of noise on the microbe s trajectory in an appropriate place in this paper ( for another example demonstrating
the importance of noise , see ref .
again we address the transient behavior and find an intermediate time window where the super - diffusive motion is found between a short - time and long - time diffusive behavior .
this motion we observe is similar to persistent random walk of a microbe which can be mapped to the worm - like chain model @xcite . in the particular case we consider ,
the persistence length is seen to depend on the microbe s coupling with the repellent .
our predictions can in principle be verified in experiments on aggregates of bacteria .
for many bacteria our model reduces to particles interacting via gravitation - kind potentials for @xmath3 .
therefore our analysis might have applications for brownian dynamics of gravitational matter @xcite .
further generalizations of our model are to predator and prey models possibly leading to interesting spatiotemporal delay effects , see e.g. @xcite . our paper is organized as follows : in section ii , we propose the model of a microorganism coupled to its own chemoattractant / chemorepellent , provide the simulation details and point out experimental situations to compare typical estimates of parameters used in the model . in section iii we present the results of our investigation . in section
iv we explain our findings with simple theoretical analysis .
we conclude the paper in section v by discussing the main points of our findings , comparisons of our results with relevant practical cases of self - propulsion , and future directions of our research .
_ model : _ the microorganism is modeled as a point particle which undergoes completely overdamped brownian motion with an effective temperature @xmath14 in a medium with viscosity coefficient @xmath15 .
this ` particle ' is assumed to emit a chemical , with which it is self - coupled , continuously in time . in this context , the word ` particle ' is thereby taken to represent the chemotactic agent of interest - an idealized microorganism coupled with a self - emitted chemical field .
fluctuations associated with the active process of self - propulsion of the microbe is modeled by the effective temperature parameter @xmath14 in our system .
the time evolution of the density field @xmath16 of the continuously emitted chemical is thus governed by a diffusion equation with a source term which depends upon the instantaneous position @xmath17 of the moving microbe : @xmath18 here , the constants @xmath19 and @xmath20 are the rate of emission of the chemical and the diffusion constant of the chemical in the medium , respectively . in the absence of chemical ,
the microbe diffuses non - chemotactically in the medium with an effective free diffusion constant @xmath21
. however , with the presence of the emitted chemical the resulting ` chemotactic ' behavior depends on the nature of the self - coupling of the microorganism to its chemical field , i.e. whether it moves ` up ' or ` down ' the chemical density gradient .
we study both cases by simply modeling the self - coupling ` force ' to be proportional to the gradient of the chemical density field @xmath22 , and the proportionality constant @xmath10 determines the strength as well as the nature of the coupling . in reality , of course , chemotaxis can be more complex involving temporally sampling of the concentration field and a biased random walk @xcite . in this context , we note that the ` chemotactic force ' imitates the net effect of chemotactic movements on a phenomenological level .
the details of the actual propulsion mechanism @xcite of the chemotactic agent and the effect of the solvent flow field on the diffusing chemical are not taken into account . within our simple model , positive and negative @xmath10
naturally generate the cases of positive and negative chemotaxes , respectively .
we are thus lead to the following idealized model of a chemotactic agent @xmath23 here , @xmath24@xmath25 is an effective noise specified by @xmath26 and @xmath27 , with @xmath28 and @xmath29 referring to the spatial components of the noise vector .
this noise term is assumed to effectively take care of all non - equilibrium fluctuations that may be associated with the _ active _ process @xcite of self - propulsion , in absence of the chemical .
@xmath30 denotes the model ` chemotactic force ' taken to imitate the systematics of the effective chemotactic movement of the microbe at the position @xmath31 at time @xmath32 due to the chemical secreted all along the trajectory traversed in the past .
it is obtained by analytically solving eq .
( [ eq : rd ] ) for the density field @xmath33 by the method of green s function , and subsequently calculating the gradient @xmath22 . the ` force ' at a time instant is dependent on the entire previous path history of the microorganism , thereby generating a strongly non - markovian dynamics . however , owing to a physical _ memory time _ ( @xmath34 ) associated with the microbe to sense its chemical , the part of the trajectory in this most recent time @xmath34 , i.e. for all @xmath35 with @xmath36 , does not contribute .
the physical import of this is that there is a finite time delay @xmath34 , however small , between the act of secreting chemical by the microorganism and the act of responding to it , during which the sensor gets to activate . with the introduction of the memory time , @xmath34 , the ` chemotactic force ' at time @xmath32 and at position
@xmath37 becomes @xmath38}{(4\pi d_{c}|t - t^{\prime}|)^{d/2 } } , \label{eq : gradrho1}\ ] ] where @xmath1 is the dimensionality of the embedding space .
evidently form eq .
( [ eq : gradrho1 ] ) , for the mathematical case of @xmath39 , the ` force ' becomes divergent .
_ simulation details : _ we performed extensive brownian dynamics simulation @xcite for this non - markovian process of a microorganism moving by autochemotaxis .
we measured time in units of @xmath40 , all lengths in units of @xmath41 and energies in units of @xmath14 .
the coupling strength @xmath10 is measured in units of @xmath42 .
thus , we set @xmath43 , @xmath44 and @xmath45 for convenience .
further we considered the physical situation when the microorganism diffuses at a much slower rate compared to the emitted chemical in the medium @xcite , taking @xmath46 and fixing the ratio @xmath47 . in our brownian dynamics simulations we used @xmath48 .
the langevin equation [ eq . ( [ eq : bd ] ) ] is solved with a discrete time step @xmath49 .
space is , however , continuous .
_ connection to experiments : _ in order to get an estimate for the coupling strength @xmath10 in our units , we note that the typical value of the ejection rate of chemical from a microorganism is @xmath50 @xcite , and usually @xmath51 @xcite . in three spatial dimensions , for example , the chemotaxis of _ dictyostelium _ to shallow camp gradients @xcite with typical values of @xmath52 , @xmath53 , @xmath54 , and moving with steady state velocity @xmath55 , yields @xmath56 .
chemotactic _ microglia _ cells @xcite move at a speed of @xmath57 in a spatial gradient @xmath58 of a chemoattractant ( @xmath59 ) which is secreted at a rate of @xmath60 and diffuses with @xmath61 .
the coupling strength in this case is @xmath62 for an effective non - chemotactic diffusion constant @xmath63 of microglial cells , due to random motility in the tissue .
the chemoattractant has a low decay rate of @xmath64 .
microglial cells are also known to respond to a chemorepellent ( @xmath65 ) they produce , with similar production , diffusion and decay rates as the chemoattractant .
our estimate of the chemoattractant gradient was based on the value of the chemotactic coefficient @xmath66 , a ratio between cell velocity and chemical gradient , used in ref .
the corresponding value for the repellent is not known .
again , for _ e. coli _ of size @xmath67 , swimming at @xmath68 in the spatial gradient @xmath69 of a chemoattractant diffusing with @xmath70 , in a medium of viscosity @xmath71 @xcite ; the coupling strength is @xmath72 .
the non - chemotactic diffusion coefficient of the bacterium is @xmath73 .
the time required by the chemical in this case , to diffuse a length equal to the size of the bacterium , is of the order of @xmath74 .
the memory time , @xmath34 , needed by the bacterium to respond to the chemical stimulus can be much smaller than this time .
in all the above calculations , the effective non - chemotactic diffusivity @xmath75 was used to express the energy unit @xmath76 .
we now investigate the nature of the dynamics in all the dimensions and for both cases of positive and negative @xmath10 , examining the model microorganism from some initial reference point taken as the origin , i.e. @xmath77 . for this purpose
we computed the mean square displacement of the microbe as a function of time and averaged over @xmath78 realizations for each case .
we checked that the system is in a steady state , and have also performed a steady state averaging of the mean square displacement .
we illustrate our findings below . upon examining the motion of the microbe in a chemoattractant ( @xmath79 ) in one dimension ( @xmath12 )
, we found signatures of long - time diffusion ( @xmath80 ) with a modified diffusion constant @xmath81 .
the value of @xmath81 depends on the strength of the coupling @xmath10 , and decreases with increasing @xmath10 .
for very high @xmath10 values it is computationally difficult to obtain the long time diffusive behavior ; but we obtained an upper estimate of the diffusivity for lower @xmath10 values from a fit to the obtained data . in the opposite limit ,
i.e. at very short times , the microbe s dynamics is also diffusive with the non - chemotactic diffusion constant @xmath75 .
the departure from this behavior occurs at times dependent on the coupling strength : the stronger the coupling , the dynamics become markedly history dependent and the microorganism deviates from the non - chemotactic diffusion faster . at intermediate times we find long period of sub - diffusive crossover regime , showing signatures of a transient dynamical arrest at high @xmath10-values .
the crossover time also increases with the coupling strength .
[ figattrr2](a ) shows the mean square displacement as a function of time in @xmath12 , for three values of @xmath10 and compared with the non - chemotactic diffusion . in @xmath82
, we find similar long - time diffusive behavior .
the crossover times from early - time non - chemotactic diffusion to the long - time modified diffusion , is greatly reduced for a given @xmath10-value as compared to the one dimensional case .
this feature is attributed to the effect of fluctuations .
the crossover time to the final long - time diffusion , however , increases with coupling strength .
the intermediate sub - diffusive regime is also diminished in time .
the mean square displacement as a function of time , in this case , is shown in fig .
[ figattrr2](b ) . in fig .
[ figattrr2](c ) we show the case for @xmath3 , and in consistence with our expectation we find similar long - time diffusion with a further reduced crossover time as compared to the low - dimensional cases . thus the transient dynamical arrest becomes more prominent at lower dimensions and at stronger couplings .
the @xmath10-dependence of the modified diffusion constant @xmath81 , is shown in fig .
[ figattrr2](d ) , for all dimensions , relative to the corresponding non - chemotactic diffusion constant @xmath75 . for the repulsive case , when the model microorganism gets repelled by its ejected chemical ( @xmath83 ) , we found a short - time non - chemotactic diffusive motion with a crossover to a ballistic behavior ( @xmath84 ) in one dimension , for all values of the coupling strength ( fig . [ figreplr2](a ) ) .
the time of commencement of the ballistic behavior , however , depends on @xmath10 . for weak coupling ( i.e. low @xmath5 ) ,
the dynamics resembles the non - chemotactic diffusion for longer times before finally going over to the ballistic dynamics .
the velocity of this ballistic motion is given by the time derivative of the root mean square displacement , @xmath85 , and this velocity is seen to increase with increase of the coupling strength @xmath5 .
we argue that the microorganism can change the direction of its motion in one dimension due to a non - vanishing finite barrier crossing probability . under such circumstance
the motion will be diffusive at very long times ( not seen in the simulation ) , with higher diffusion constant . in higher dimensions , @xmath82 ( fig .
[ figreplr2](b ) ) and @xmath3 ( fig . [ figreplr2](c ) ) , we observed a crossover from the non - chemotactic diffusion to a long - time diffusion for all coupling strengths , with a ballistic transient dynamics at intermediate times for high @xmath5 .
the velocity of the transient ballistic motion increase with increasing coupling strength .
the time duration of this ballistic transient as well as the modified long - time diffusion constant are also @xmath10-dependent , both increase with increase in @xmath5 . in fig .
[ figreplr2](d ) we show this modified diffusion constant for the repulsive case relative to the non - chemotactic diffusion constant as a function of @xmath5 , in @xmath86 .
our findings can be understood qualitatively and partly quantitatively by relatively simple theoretical considerations , which are presented in the current section . in subsection
[ subsec : theory_pos ] , a scaling law for the diffusion constant for the case of positive chemotaxis , @xmath87 , is derived based on a simple rate theory .
the same scaling @xmath88 had been predicted for the slightly different model , which includes evaporation , by newman and grima @xcite and grima @xcite . in subsection
[ subsec : theory_neg ] we present a theory , which quantitatively predicts the long - time diffusion constant in the case of negative chemotaxis for all coupling strengths .
both approaches , for negative and positive chemotaxis , attribute the long - time diffusion to small fluctuations about the respective steady states in the case of zero noise . in each subsection
we therefore first present the solutions to the equation of motion ( [ eq : bd ] ) without fluctuations ( i.e. , at zero temperature or infinite coupling strength @xmath10 ) , before discussing the influence of small fluctuations on the long - time dynamics . in the case of strong positive
autochemotaxis the model microorganism is trapped within its own secretion , which effectively provides an attractive external potential at the microbe s location @xmath89 . for zero noise ,
the microbe is at rest , i.e. , @xmath90 , and does not experience any force .
if , on the contrary , the microbe was at time @xmath32 instantaneously placed a distance @xmath91 away from the location , which it occupied at all earlier times @xmath92 , it would feel a force @xmath93 , which is obtained analytically by evaluating eq .
( see also fig .
[ fig : forceattr ] ) : ( color online ) .
the force @xmath94 , which a microorganism would experience if it was at time @xmath32 instantaneously placed a distance @xmath91 away from the location , which it occupied at all earlier times @xmath92 , plotted as a function of @xmath95 for all dimensions @xmath0 .
representative parameters chosen are @xmath96 , @xmath97 , and @xmath98.,width=302 ] @xmath99\frac{{\bf x}}{x^2 } \,,&d=3\ , .
\end{array}\right.\ ] ] here , @xmath100 is the dimensionless distance from the former position @xmath31 of the microbe and @xmath101 is its absolute value .
locally , i.e. , for distances @xmath102 , the force is linear in the distance and given by @xmath103 without fluctuations , the microbe never experiences the described force field .
however , if the noise is nonzero but small , the microbe eventually walks up the locally parabolic walls of its self - generated potential for a short time @xmath104 , which , in turn , might lead to a shift of the position of the minimum @xmath105 of the self - generated attractive potential by a distance @xmath106 . clearly , the minimum is only displaced , if the duration of the excursion @xmath104 is larger than the memory time @xmath34 and if the microbe excurses predominantly in one direction . after time
@xmath107 the microbe position might relax to the new minimum , @xmath108 .
by such a fluctuation , the microorganism effectively manages to move a distance @xmath109 within the time @xmath104 .
most relevant for the long - time diffusion are those fluctuations , which yield a large displacement @xmath109 and still occur at a high rate @xmath110 .
these fluctuations are regarded to constitute the relevant mean steps in an effective continuous time random walk with the desired diffusion constant @xmath111 . in this subsection
, we only attempt to obtain a scaling law for the long - time diffusivity .
we therefore restrict our consideration to a subset of very simple fluctuations , which are believed to be representative for all fluctuations relevant for diffusion .
in particular , we consider a `` jump '' process at time @xmath112 : at all earlier times , @xmath113 , the microbe is resting at the location of the self - generated potential s minimum at the origin , i.e. , @xmath114 , before it undergoes an excursion to a new location @xmath115 within a time @xmath116 ; the latter timescale is assumed to be small with respect to the residence time @xmath104 . during the latter time span @xmath104
the microbe stays at the new position , where it resists the force due to the secretion from earlier times . in the case of a large coupling strength @xmath10 the process of getting to the new location during the time span of duration
@xmath116 is irrelevant .
after time @xmath104 the position of the model microbe is deterministically relaxing towards the new minimum of the chemoattractant at @xmath117 , which is a function of the parameters @xmath104 and @xmath118 , only . summarizing , the simple pathway is described by @xmath119 the probability of the described fluctuation is proportional to the arrhenius factor @xcite @xmath120 \propto \exp\left[-s(\tau,\delta r_b)/4\right]\,,\ ] ] with @xmath121 and @xmath122 the onsager - machlup action @xmath123 neglecting the initial process of moving the distance @xmath124 and also ignoring the relaxation of the chemoattractant during the time @xmath104 the action is approximately given by @xmath125 ^ 2\,,\ ] ] which , together with eqs .
( [ eq : fss ] ) and ( [ eq : fsslimit ] ) yields @xmath126 according to eq .
, a minimum requirement for the described fluctuation to occur frequently is that @xmath122 does not exceed a value of the order of @xmath127 , i.e. , @xmath128 . as eq .
( [ eq : om2approx ] ) is a strictly monotonically increasing function of @xmath104 and @xmath124 , this constraint is equivalent to the equality @xmath129 diffusion is believed to be governed by those random displacements of the microorganism , which maximize the shift of the potential s minimum @xmath130 subject to this constraint .
we will shortly see that for large coupling constants @xmath131 small excursions @xmath132 are most relevant . in this latter limit , @xmath133 is obtained analytically as @xmath134\delta r_b\,.\ ] ] the maximum of @xmath133 subject to the constraint of eq .
( [ eq : constraint_action ] ) is given by @xmath135 at an optimum displacement of the microbe and an optimum residence time @xmath136 with prefactors @xmath137 , @xmath138 , @xmath139 and @xmath140 , @xmath141 , @xmath142 , in one , two , and three dimensions , respectively . clearly , eq . ( [ eq : optimum_displacement_restime ] ) fulfills the before mentioned assumption of small excursions in the limit of large @xmath143 .
( [ eq : optimum_displacement_restime ] ) yields the desired scaling behavior for the long - time diffusivity @xmath144 , @xmath145 in conclusion , the simple theory predicts an inverse quadratic dependence of @xmath81 on @xmath10 in the strong - coupling limit , which has also been observed in the computer simulations [ see fig . [
figattrr2](d ) ] .
the same scaling had already been predicted for the slightly different model , which includes evaporation , by newman and grima @xcite and grima @xcite .
first , we consider the microorganism s motion with no fluctuations present . in this case , the swimmer reaches a steady state at infinite time , which is determined by a constant swimming speed @xmath146 , where the index @xmath147 denotes the steady - state configuration . under this condition
, the drag force induced by the solvent , @xmath148 , equals the driving force due to the chemical , @xmath149 .
figuratively , the microorganism surfs down its own chemorepellent , which it excreted at times @xmath150 .
using eq .
, the steady - state velocity @xmath151 is determined by the self - consistent equation @xmath152}{(4\pi
d_c t')^{d/2}}\nonumber\\ = \left(\frac{v_s}{4\pi^{1/2}d_c}\right)^d\frac{2\lambda}{v_s } \,\gamma\left(1-\frac{d}{2},\frac{v_s^2t_0}{4d_c}\right)\ , , \label{eq : gradrho2}\end{aligned}\ ] ] where @xmath153 denotes the incomplete gamma function .
eq . has non - zero solutions @xmath154 for any value of @xmath10 in one and two spatial dimensions , whereas there is a dynamical phase transition at a lower critical value of @xmath155 in three dimensions , below which the microbe comes to a rest .
the latter is obtained analytically by expanding eq . up to second order in @xmath151 .
for the parameters used in the simulations reported above , @xmath156 , @xmath157 , and @xmath158 , the transition occurs at @xmath159 .
the asymptotic solutions for small @xmath5 in one and two dimensions or for small @xmath160 in three dimensions , respectively , are given by @xmath161\,,&d=2\\ \left(4\pi^2d_c^2\gamma t_0\right)^{-1}\left(|\lambda-\lambda^*|\right)\,,&d=3\ , , \end{array}\right.\ ] ] where @xmath162 is euler s constant .
the steady - state velocity as a function of @xmath5 is shown in fig . [
fig : steadystate ] for all three spatial dimensions . for high coupling constants , @xmath163 , the steady - state velocities @xmath151 agree well with the square root of the slope of the mean - square displacements @xmath164 , as obtained from the simulations with noise in two and three dimensions ( see also fig . [
fig : steadystate ] ) .
( color online ) .
the steady - state velocity @xmath165 as a function of coupling constant @xmath5 for one , two , and three spatial dimensions , determined theoretically ( lines ) , compared to the square root of the slopes of the mean - square displacements at early times , @xmath166 , for two and three dimensions ( symbols ) [ cf .
[ figreplr2 ] ( b , c ) ] .
inset : the asymptotic solution of @xmath151 for small values of @xmath160 , where we set @xmath167 for @xmath168.,width=302 ] if noise is put on , again , the microbe is eventually disturbed in its steady - state motion . in one dimension , the microbe needs to overcome a barrier in order to change the direction of motion from left to right or vice versa , which renders the steady state very stable .
as we were not able to observe long - time diffusive motion in one dimension for any coupling constant but always found ballistic motion within the time window accessible in computer simulations , we do not attempt to give a theoretical estimate of the diffusivity . in two or three dimensions ,
the picture is very different : the microorganism is only constrained in its motion parallel to its current trajectory , whereas it is free to move perpendicular to the same . during such transverse fluctuations
the direction of the microbe s motion changes with the gradient of the chemical field . in the following ,
we give a theoretical estimate of the microbe s long - time diffusivity int @xmath169 under the assumption of a fast relaxation of the chemical field , which is justified for @xmath170 . in this case , the change of orientation is determined by the local , time - independent curvature @xmath171 of the isodensity line ( in 2d ) or surface ( in 3d ) at the steady - state position @xmath172 ; the isodensity planes are defined by @xmath173 . locally , they have the form of a parabola ( in 2d ) or of a paraboloid of revolution ( in 3d ) and they move with the microorganism at the microorganism s velocity ; this can be seen in fig .
[ figrepchemdiffu ] , where a typical trajectory and the according chemical field at time @xmath32 is plotted for a microorganism in two dimensions .
( color online ) .
( a ) snapshot of the instantaneous density profile @xmath174 of the chemorepellent released by the microorganism moving in @xmath175dimensions , obtained from simulation , at time instant @xmath176 ( in units of @xmath177 ) .
( b ) the entire trajectory , shown as the red ( thick ) curve , of the microorganism . the current position of the microorganism @xmath178 is indicated by the blue ( black ) dot in both the figures , and the direction of motion is indicated by arrows along the trajectory . the corresponding coupling strength being @xmath179.,width=302 ] fluctuations transverse to the direction of motion lead to a mean transverse displacement @xmath180 , which , in turn , leads to a change in orientation of the velocity director @xmath181 by a mean - square angle @xmath182 exploiting that @xmath183 is gaussian distributed , the average , normalized , and on the initial orientation @xmath184 projected velocity vector @xmath185 is given by @xmath186\rangle = \exp\left[-\frac{\langle\theta^2(t)\rangle}{2}\right]\,.\ ] ] replacing time by arc length @xmath187
, the problem under study can therefore be mapped to the _ worm - like chain _ ( wlc ) model @xcite of a polymer of a total arc length @xmath188 and a persistence length , which is defined implicitly by @xmath189\rangle=\exp[-v_st / l_p]$ ] . making use of eqs .
( [ eq : meansquareangle ] ) and ( [ eq : averagecostheta ] ) the latter therefore reads @xmath190 $ ] .
the mean square end - to - end distance of the wlc for long chains @xmath191 is well known @xcite to be given by @xmath192 the long - time diffusivity of the microbe is then obtained as the time derivative of the mean square end - to - end distance , @xmath193 where we point out the @xmath10-dependence of the steady - state velocity and the steady - state curvature .
@xmath194 is plotted in fig .
[ figreplr2 ] ( d ) and compared to the results of the simulation .
it is ascertained from fig .
[ figreplr2 ] ( d ) that , in the limit of large @xmath195 , the theory overestimates the diffusivity by a factor of @xmath196 , both in two and three dimensions , for the parameters @xmath197 and @xmath158 , almost independent of @xmath10 .
this discrepancy is attributed to the strong assumption of a constant , non - fluctuating curvature @xmath171 in eqs .
( [ eq : meansquareangle ] ) and ( [ eq : dprime ] ) . relaxing this assumption
to fluctuating curvature with first and second moments @xmath198 and @xmath199 , which is further uncorrelated with the transverse displacement , i.e. , @xmath200 , the diffusion constant reduces to @xmath201 $ ] .
therefore , although we do not attempt to give an estimate of @xmath202 here , the long - time diffusivity is expected to be smaller than in the `` zeroth order '' theory , eq .
( [ eq : dprime ] ) , in agreement with the simulation results .
in conclusion , we have explored the dynamics of autochemotaxis : a model microorganism is smelling " its own secretion which is diffusing away .
the microorganism follows the gradient @xcite of its secreted chemical . for the attractive case
, the mean - square displacement of the microorganism reveals a transient dynamical arrest , most pronounced in low spatial dimensionality . in the opposite case of chemorepulsion
, there is a transient ballistic behavior which crosses over to ultimate diffusion @xcite .
a simple theoretical analysis for large coupling strengths @xmath10 reveals a scaling law for the long - time diffusion in the case of a chemoattractant , by regarding small excursions of the microbe s position about the bottom of an effective , time - dependent trapping potential . in the case of chemorepulsion
, the microbe s trajectory was mapped to the contour of a worm - like chain , which gives a semiquantitative agreement with our computer simulations .
the ranges of the coupling strength reported here are also easily obtainable in real experimental situations ( @xmath203 ) , as estimated earlier in the text .
we note that though the mechanism of temporal sensing @xcite of chemoattractant or chemorepellent which we do not consider here is found in many small , fast moving bacteria like _ e. coli _ , the direct gradient sensing mechanism is also present in microorganisms like the amoeba _ dictyostelium discoideum _ , the yeast cell _ saccharomyces carevisiae _ , lymphocytes , glial cells and myxobacteria @xcite .
chemotaxis in the gliding bacterium _
m. xanthus _ was demonstrated to be in response to self - generated signaling chemicals @xcite . however , the diffusivity of the chemoattractant is significantly lower compared with the bacterial motility in this peculiar case .
a numerical investigation @xcite of a simple model case in one - dimension compared the interplay between chemotaxis and chemokinesis mechanism with a concentration dependent switching rate , showing crossover from suppressed to enhanced diffusion in the parameter space .
this mean - field approach , studied in the limit of vanishing chemical diffusivity and chemical degradation rate , was based on the simplifying assumption that the chemical coupling only affects the frequency of direction reversal of the cell , keeping the cell speed unaltered .
models of active colloids using surface reactions as a potential mechanism for self - propulsion has been proposed ( see e.g. @xcite ) . in ref .
@xcite , the model _ molecular machine _ is a spherical colloid which reacts with the substrate at a specific site on its surface and self - propel by asymmetric release of the reaction product .
the time evolution of the product particles is similar to that of chemical molecules emitted by a microorganism .
though the exact way of imparting a biased motion to the colloid by the emitted particles is very different from chemotactic coupling , such systems may also provide interesting situations to compare with our results .
in fact in ref .
@xcite , active colloids interacting with a self - generated cloud of solute were shown to have distinct propulsive and anomalous super - diffusive regimes preceding a final long - time effective diffusion .
this is worth comparing with the ballistic to a modified diffusion crossover in our study on negative autochemotaxis .
future work should focus on generalizing the model to a collection of model microorganisms @xcite , where hydrodynamic flow effects can play an important role @xcite and steric repulsions can lead to aggregation and clumps @xcite as known from active particles @xcite .
it would further be interesting to study the case of a geometric confinement of the particles @xcite and the chemoattractant / repellent in order to see effects of a dimensional crossover .
finally , improving the theoretical approaches towards full quantitative agreement will be an important task for future research . | the brownian dynamics of a single microorganism coupled by chemotaxis to a diffusing concentration field which is secreted by the microorganism itself is studied by computer simulations in spatial dimensions @xmath0 .
both cases of a chemoattractant and a chemorepellent are discussed . for a chemoattractant
, we find a transient dynamical arrest until the microorganism diffuses for long times . for a chemorepellent
, there is a transient ballistic motion in all dimensions and a long - time diffusion .
these results are interpreted with the help of a theoretical analysis . |
we address the logarithmic korteweg de vries ( log - kdv ) equation derived in the context of solitary waves in granular chains with hertzian interaction forces @xcite : @xmath0 the log kdv equation ( [ logkdv ] ) has a two - parameter family of gaussian solitary waves @xmath1 where @xmath2 is a symmetric standing wave given by @xmath3 global solutions to the log kdv equation ( [ logkdv ] ) were constructed in @xcite in the energy space @xmath4 by a modification of analytic methods available for the log nls equation @xcite ( also reviewed in section 9.3 in @xcite ) . in the energy space @xmath5 , the following quantities for the momentum and energy , @xmath6 and @xmath7 dx\ ] ] are non - increasing functions of time @xmath8 .
uniqueness , continuous dependence , and energy conservation are established in @xcite under the additional condition @xmath9 , which is not satisfied in the neighborhood of the family of gaussian solitary waves given by ( [ soliton - orbit ] ) and ( [ gaussian ] ) . as a result , orbital stability of the gaussian solitary waves was not established for the log kdv equation ( [ logkdv ] ) , in a sharp contrast with that in the log nls equation established in @xcite .
a possible path towards analysis of orbital stability of gaussian solitary waves is to study their linear and spectral stability by using the linearized log kdv equation @xmath10 where @xmath11 is the schrdinger operator with a harmonic potential given by the differential expression @xmath12 the linearized log kdv equation ( [ linlogkdv ] ) arises at the formal linearization of the log kdv equation ( [ logkdv ] ) at the perturbation @xmath13 .
the schrdinger operator @xmath14 is the hessian operator of the second variation of @xmath15 at @xmath16 .
although @xmath15 in ( [ energy ] ) is not a @xmath17 functional at @xmath18 , the second variation of @xmath15 is well defined at @xmath16 by @xmath19 which is formally conserved in the time evolution of ( [ linlogkdv ] ) . with new estimates to be obtained for the linearized log kdv equation ( [ linlogkdv ] ) ,
we may hope to develop an ultimate solution of the outstanding problem on the orbital stability of the gaussian solitary waves . indeed ,
if we set @xmath20 for the solution to the log kdv equation ( [ logkdv ] ) , we obtain an equivalent evolution equation @xmath21 where the linearized part coincides with ( [ linlogkdv ] ) and the nonlinear term @xmath22 is given by @xmath23.\ ] ] it is clear that the nonlinear term @xmath22 does not behave uniformly in @xmath24 unless @xmath25 decays at least as fast as @xmath2 in ( [ gaussian ] ) . on the other hand , if @xmath26 , where @xmath27 is a bounded function in its variables , then @xmath28 , where @xmath29 is analytic in @xmath27 for any @xmath30 .
therefore , obtaining new estimates for the linearized log kdv equation ( [ linlogkdv ] ) in a function space with gaussian weights may be useful in the nonlinear analysis of the log kdv equation ( [ logkdv - w ] ) .
the spectrum of @xmath14 in @xmath31 consists of equally spaced simple eigenvalues @xmath32 which include exactly one negative eigenvalue with the eigenvector @xmath2 ( defined without normalization ) .
therefore , @xmath15 is not convex at @xmath2 in @xmath5 .
nevertheless , @xmath33 is positive in the constrained space @xmath34 which corresponds to the fixed value @xmath35 in ( [ momentum ] ) at the linearized approximation .
several results were obtained for the linearized log kdv equation ( [ linlogkdv ] ) . in @xcite , _ linear orbital stability _ of gaussian solitary waves was obtained in the following sense : for every @xmath36 , there exists a unique global solution @xmath37 of the linearized log kdv equation ( [ linlogkdv ] ) which satisfies the following bound @xmath38 for some @xmath8-independent positive constant @xmath39 .
this result was obtained in @xcite from the conservation of @xmath33 in the time evolution of smooth solutions to the linearized log kdv equation ( [ linlogkdv ] ) , the symplectic decomposition of the solution @xmath37 , @xmath40 into the translational part and the residual part , @xmath41 and the coercivity of @xmath42 in the squared @xmath43 norm in the sense @xmath44 for some positive constant @xmath39 .
the first two facts are rather standard in energy methods for linear pdes , whereas the last fact , that is , the inequality ( [ coercivity - assumed ] ) , should not be taken as granted . in @xcite ,
the nonzero spectrum of the linear operator @xmath45 was studied by using the fourier transform that maps the third - order differential operator in physical space into a second - order differential operator in fourier space .
indeed , the fourier transform @xmath46 applied to the linearized log kdv equation ( [ linlogkdv ] ) yields the time evolution in the form @xmath47 where @xmath48 is the fourier image of operator @xmath11 given by @xmath49 by reducing the eigenvalue problem for @xmath50 to the symmetric sturm
liouville form , it was found in @xcite that the spectrum of @xmath51 in @xmath31 is purely discrete and consists of a double zero eigenvalue and a symmetric sequence of simple purely imaginary eigenvalues @xmath52 such that @xmath53 the double zero eigenvalue corresponds to the jordan block @xmath54 whereas the purely imaginary eigenvalues @xmath55 correspond to the eigenfunctions @xmath56 , which are smooth in @xmath24 but decay algebraically as @xmath57 .
the fourier transform of @xmath58 is supported on the half - line @xmath59 and decays like a gaussian function at infinity .
it follows from the spectrum of @xmath51 in @xmath31 that the gaussian solitary waves are _ spectral stable_. the eigenfunctions of @xmath51 were also used in @xcite for spectral decompositions in the constrained space @xmath60 in order to provide an alternative proof of the _ linear orbital stability _ of the gaussian solitary waves .
this alternative technique still relies on the conjecture of the coercivity of @xmath42 in the squared @xmath43 norm , that is , on the inequality ( [ coercivity - assumed ] ) . because of the algebraic decay of the eigenfunctions of @xmath51 , it is not clear if a function of @xmath24 that decays like the gaussian function as @xmath57 can be represented as series of eigenfunctions .
numerical simulations were undertaken in @xcite to illustrate that solutions to the linearized log kdv equation ( [ linlogkdv ] ) with gaussian initial data did not spread out as the time variable evolves .
nevertheless , the solutions exhibited visible radiation at the left slopes .
the present work is developed to obtain new estimates for the linearized log kdv equation ( [ linlogkdv ] ) . in the first part of this work
, we rely on the basis of hermite functions in @xmath61-based sobolev spaces and analyze the discrete operators that replace the differential operators . in the second part , we obtain dissipative estimates on the evolution of the linearized log kdv equation ( [ linlogkdv ] ) by representing solutions in terms of a convolution with the gaussian solitary wave @xmath2 . the paper is structured as follows .
section 2 sets up the basic formalism of the hermite functions and reports useful technical estimates .
section 3 is devoted to the proof of the coercivity bound ( [ coercivity - assumed ] ) .
as explained above , this coercivity bound implies _ linear orbital stability _ of the gaussian solitary wave in the constrained space @xmath60 and it is assumed to be granted in @xcite .
the proof of coercivity relies on the decomposition of @xmath62 in terms of the hermite functions .
section 4 is devoted to the analysis of linear evolution expressed in terms of the hermite functions .
it is shown that this evolution reduces to the self - adjoint jacobi difference operator with the limit circle behavior at infinity . as a result
, a boundary condition is needed at infinity in order to define the spectrum of the jacobi operator and to obtain the norm - preserving property of the associated semi - group . both _
linear orbital stability _ and _ spectral stability _ of gaussian solitary waves ( [ gaussian ] ) is equivalently proven by using the jacobi difference operator . in section 5 ,
we give numerical approximations of eigenvalues and eigenvectors of the jacobi difference equation .
we show numerically that there exist subtle differences between the representation of eigenvectors of @xmath51 in the physical space and the representation of these eigenvectors by using decomposition in terms of the hermite functions .
section 6 reports weighted estimates for solutions to the linearized log kdv equation ( [ linlogkdv ] ) by using a convolution representation with the gaussian weight .
we show that the convolution representation is invariant under the time evolution of the linearized log kdv equation ( [ linlogkdv ] ) , which is expressed by a dissipative operator on a half - line .
the semi - group of the fundamental solution in the @xmath63 norm decays to zero exponentially fast as time goes to infinity .
section 7 concludes the paper with discussions of further prospects .
* notations : * we denote with @xmath64 the sobolev space of @xmath65-times weakly differentiable functions on the real line whose derivatives up to order @xmath65 are in @xmath31 .
the norm @xmath66 for @xmath67 in the sobolev space @xmath64 is equivalent to the norm @xmath68 in the lebesgue space @xmath31 .
we denote with @xmath69 the space of square integrable functions with the weight @xmath70 .
the set @xmath71 consists of all non - negative integers , whereas the set @xmath72 includes only positive integers .
the sequence space @xmath73 includes squared summable sequences , whereas @xmath74 contains finite ( compactly supported ) sequences .
* acknowledgements . *
the author thanks gerald teschl and thierry gallay for help on obtaining results reported in sections 4 and 6 , respectively .
the research of the author is supported by the nserc discovery grant .
we recall definitions of the hermite functions ( * ? ? ? * chapter 22 ) : @xmath75 where @xmath76 denote the set of hermite polynomials , e.g. , @xmath77 hermite functions satisfy the schrdinger equation for a quantum harmonic oscillator : @xmath78 at equally spaced energy levels . by the sturm
liouville theory @xcite , the set of hermite functions @xmath79 forms an orthogonal and normalized basis in @xmath31 .
in connection to the self - adjoint operator @xmath11 given by ( [ schrodinger ] ) , we obtain the eigenfunctions of @xmath80 , @xmath81 from the correspondence @xmath82 . with proper normalization , we define @xmath83 it follows from the well - known relations for hermite polynomials @xmath84 that functions in the sequence @xmath85 satisfy the differential relations @xmath86 the following elementary result is needed in further estimates .
[ lemma - f - sequence ] let @xmath87 be given by @xmath88 then , there is a positive constant @xmath39 such that @xmath89 we write @xmath90.\ ] ] by taylor series , for every @xmath91 and every @xmath92 , there is @xmath93 such that @xmath94 furthermore , we recall euler s constant @xmath95 given by the limit @xmath96 since @xmath97 is bounded as @xmath98 , the estimate ( [ taylor ] ) and the limit ( [ euler - gamma ] ) yield the bound @xmath99 for some positive constant @xmath39 . substituting ( [ bound - log ] ) into ( [ f - m - exp ] ) proves the desired bound ( [ bound - decay ] ) .
the following technical result is needed for the proof of coercivity of the energy function .
[ lemma - projections ] let @xmath100 by defined by @xmath101 .
then , there is a positive constant @xmath39 such that @xmath102 multiplying the differential relation ( [ diff - hermite ] ) by @xmath103 and integrating by parts , we obtain @xmath104 integrating directly , we compute @xmath105 furthermore , using ( [ difference - integral ] ) at @xmath106 , we also compute @xmath107 thanks to orthogonality of hermite functions , the right - hand side of ( [ difference - integral ] ) is zero for @xmath108 and the numerical sequence @xmath109 satisfies the recurrence equation @xmath110 starting with the initial values for @xmath111 and @xmath112 in ( [ f-0 ] ) and ( [ f-1 ] ) . the recurrence equation ( [ difference - second - order ] ) admits the exact solution @xmath113 applying the bound ( [ bound - decay ] ) of lemma [ lemma - f - sequence ] with @xmath114 , @xmath115 or @xmath116 , @xmath117 yields the bound ( [ projection - decay ] ) .
in order to prove the coercivity bound ( [ coercivity - assumed ] ) , we define the @xmath14-compatible squared norm in space @xmath43 , @xmath118 dx\ ] ] the second variation @xmath33 is defined by ( [ energy - second - variation ] ) . the following theorem yields the coercivity bound for the energy function , which was assumed in @xcite without a proof .
[ theorem - coercivity ] there exists a constant @xmath119 such that for every @xmath120 satisfying the constraints @xmath121 it is true that @xmath122 where @xmath123 .
the upper bound in ( [ coercivity ] ) follows trivially from the identity @xmath124 whereas the lower bound holds if there is a constant @xmath93 such that for every @xmath125 satisfying constraints ( [ constraints ] ) , it is true that @xmath126 by the spectral theorem , we represent every @xmath120 by @xmath127 where the vector @xmath128 belongs to @xmath129 .
it follows from the first constraint in ( [ constraints ] ) that @xmath130 . using the norm in ( [ norm - h-1 ] )
, we obtain @xmath131 therefore , @xmath132 and coercivity ( [ coercivity - l-2 ] ) is proved if we can show that @xmath133 is bounded by @xmath42 up to a multiplicative constant . to show this
, we use the second constraint in ( [ constraints ] ) .
since @xmath134 , as it follows from ( [ f-1 ] ) , we have @xmath135 by lemma [ lemma - projections ] , there is a positive constant @xmath136 such that @xmath137 which follows from convergence of @xmath138 .
hence , by using cauchy
schwarz inequality in ( [ projection-2 ] ) , we obtain @xmath139 so that the bound ( [ coercivity - l-2 ] ) follows .
the statement of the theorem is proven .
the time evolution of the linearized log kdv equation ( [ linlogkdv ] ) is considered in the constrained energy space @xmath60 given by ( [ constrained - space ] ) . for a vector @xmath140
, we use the decomposition involving the hermite functions , @xmath141 by using @xmath142 and the differential relations ( [ diff - hermite ] ) , the evolution problem for the vector @xmath143 is written as the lattice differential equation @xmath144 it follows from ( [ lattice - c ] ) for @xmath106 that if @xmath36 ( so that @xmath145 ) , then @xmath146 and @xmath37 for every @xmath8 . if @xmath146 , then it follows from ( [ lattice - c ] ) for @xmath147 that the time evolution of a projection of @xmath148 to @xmath149 ( which is proportional to the translational mode @xmath150 ) is given by @xmath151 the projection @xmath152 is decoupled from the rest of the system ( [ lattice - c ] ) . therefore , introducing @xmath153 for @xmath108 , we close the evolution system ( [ lattice - c ] ) at the lattice differential equation @xmath154 since @xmath155 , then @xmath156 , so that we can introduce @xmath157 , @xmath108 with the vector @xmath158 . the sequence @xmath159 satisfies the evolution system in the skew - symmetric form @xmath160 the evolution system ( [ lattice - a ] ) can be expressed in the symmetric form by using the transformation @xmath161 the new sequence @xmath162 satisfies the evolution system written in the operator form @xmath163 where @xmath164 is the jacobi operator defined by @xmath165 the jacobi difference equation is @xmath166 . according to the definition in section 2.6 of @xcite , the jacobi operator @xmath164 is said to have a limit circle at infinity if a solution @xmath167 of @xmath168 with @xmath169 is in @xmath73 for some @xmath170 . by lemma 2.15 in @xcite ,
this property remains true for all @xmath170 .
the following lemma shows that this is exactly our case .
[ lemma - circle ] the jacobi operator @xmath164 defined by ( [ jacobi - f ] ) has a limit circle at infinity .
let us consider the case @xmath171 and define a solution of @xmath172 with @xmath173 . the numerical sequence @xmath174 satisfies the recurrence relation @xmath175 starting with @xmath173 .
then , @xmath176 for even @xmath177 , whereas @xmath178 for odd @xmath177 is given by the exact solution @xmath179 by lemma [ lemma - f - sequence ] with @xmath114 and @xmath180 , there exists a positive constant @xmath39 such that @xmath181 . by lemma 2.16 in @xcite ,
the jacobi operator @xmath182 with the domain @xmath183 is self - adjoint if @xmath184 for all @xmath185 , where the discrete wronskian is given by @xmath186 and @xmath187 . in order to define a self - adjoint extension of the jacobi operator @xmath164 with the limit circle at infinity , we need to define a boundary condition as follows : @xmath188 where @xmath189 is real . by lemma 2.17 and theorem 2.18 in @xcite , the operator @xmath190 , where @xmath191 and @xmath192 represents a self - adjoint extension of the jacobi operator @xmath164 with the limit circle at infinity .
moreover , by lemma 2.19 in @xcite , the real spectrum of @xmath193 in @xmath73 is purely discrete . since there is at most one linearly independent solution of the jacobi difference equation @xmath168 thanks to the recurrence relation in ( [ jacobi - f ] ) , each isolated eigenvalue of the real spectrum of @xmath193 is simple . by lemma 2.20 in @xcite , all self - adjoint extensions of @xmath194 are uniquely defined by the choice @xmath191 in ( [ boundary - condition ] ) spanned by a linear combination of two linearly independent solutions of @xmath195 .
since the value of @xmath111 plays no role for the jacobi operator @xmath164 thanks again to the recurrence relation in ( [ jacobi - f ] ) and the value @xmath112 can be uniquely normalized to @xmath169 , we have a unique choice for @xmath189 given by the solution of @xmath172 with @xmath173 . combining these facts together
, we have obtained the following result .
[ lem - jacobi ] let @xmath191 be a unique solution of @xmath172 with @xmath173 determined in the proof of lemma [ lemma - circle ] .
then , @xmath190 with the domain in ( [ domain - condition ] ) is a unique self - adjoint extension of the jacobi operator @xmath164 given by ( [ jacobi - f ] ) .
moreover , the spectrum of @xmath193 consists of a countable set of simple real isolated eigenvalues .
the following theorem and corollary provide the _ linear orbital stability _ of gaussian solitary waves ( [ gaussian ] ) expressed by using the decomposition in terms of hermite functions .
the same result was obtained in @xcite by using alternative techniques involving either the energy method @xcite or the spectral decompositions @xcite .
[ theorem - linear - evolution ] for every @xmath196 , there exists a unique solution @xmath197 to the evolution system ( [ lattice - a ] ) for every @xmath40 satisfying @xmath198 .
the semi - group property of the solution operator @xmath199 both for @xmath200 and @xmath201 associated with the linear system ( [ lattice - f ] ) follows from the result of lemma [ lem - jacobi ] and the classical semi - group theory @xcite .
the result is transferred to the sequence @xmath158 by using the transformation ( [ transformation - a ] ) .
for every @xmath36 given by ( [ constrained - space ] ) , there exists a unique solution @xmath37 to the linearized log kdv equation ( [ linlogkdv ] ) for every @xmath200 satisfying @xmath202 . by using the transformations @xmath157 and @xmath153 for @xmath108 and the decomposition ( [ decomposition - lin - log - kdv ] ) , we obtain @xmath203 where @xmath130 is set uniquely in @xmath60 .
recall that @xmath130 is an invariant reduction of the linearized log kdv equation ( [ linlogkdv ] ) .
the assertion of the corollary follows from theorem [ theorem - linear - evolution ] and the equivalence ( [ equivalence - a - energy ] ) . at the first glance , the detached equation ( [ projection - c-1 ] )
might imply that the projection @xmath152 to the translational mode @xmath149 may grow at most linearly as @xmath204 in the energy space @xmath60 with conserved @xmath202 .
however , it follows directly from the linearized log kdv equation ( [ linlogkdv ] ) @xcite for the solution @xmath205 that @xmath206 therefore , we obtain as in the proof of theorem [ theorem - coercivity ] that @xmath207 where both terms are globally bounded for all @xmath40 .
we discuss here numerical approximations of eigenvalues in the spectrum of the self - adjoint operator @xmath190 constructed in lemma [ lem - jacobi ] .
let @xmath208 denote the simple real eigenvalues of @xmath209 with the ordering @xmath210 these real eigenvalues of @xmath193 transform to eigenvalues @xmath211 of the spectral problem @xmath212 by using the decomposition ( [ decomposition - lin - log - kdv ] ) .
therefore , the result of lemma [ lem - jacobi ] also provides an alternative proof of the _ spectral stability _ of the gaussian solitary waves ( [ gaussian ] ) , which is also established in @xcite .
however , by comparing the eigenvectors obtained in the two alternative approaches , we will see some sharp differences in the definition of function spaces these eigenvectors belong to . to proceed with numerical approximations
, we note that if @xmath213 is a solution of @xmath172 like in the proof of lemma [ lemma - circle ] , then @xmath176 for even @xmath177 .
let us denote @xmath214 for @xmath215 . from the bound ( [ bound - v ] )
, we note that @xmath216 as @xmath98 .
let @xmath217 be a solution of @xmath168 for @xmath218 .
then , we denote @xmath219 and @xmath220 for @xmath215 .
it follows from the definition ( [ jacobi - f ] ) that @xmath221 and @xmath222 satisfy the coupled system of difference equations : @xmath223 starting with @xmath224 .
the discrete wronskian ( [ wronskian ] ) is now explicitly computed as @xmath225 since generally @xmath226 as @xmath98 , by applying lemma [ lemma - f - sequence ] with @xmath227 and @xmath117 to the first equation of system ( [ coupled - difference ] ) , the limit @xmath228 exists and is generally nonzero . moreover
, the sign alternation of @xmath229 and @xmath230 ensures that the sequence @xmath231 is sign - definite for large enough @xmath177 , so that the limit is actually @xmath232 .
this is confirmed in figure [ fig - wronskian](a ) , which shows the wronskian sequence @xmath231 given by ( [ wronskian - numerics ] ) for @xmath233 .
as @xmath234 for @xmath233 .
( b ) oscillatory behavior of @xmath235 versus @xmath236.,title="fig : " ] as @xmath234 for @xmath233 .
( b ) oscillatory behavior of @xmath235 versus @xmath236.,title="fig : " ] computing numerically @xmath235 by truncation of @xmath231 at a sufficiently large @xmath177 , e.g. at @xmath237 , we plot @xmath235 versus @xmath236 on figure [ fig - wronskian](b ) .
oscillations of @xmath235 are observed and the first two zeros of @xmath235 are located at @xmath238 these values are nicely compared to the first two eigenvalues computed in @xcite for @xmath239 : @xmath240 the numerical approximations confirm that the eigenvalues obtained by using the jacobi difference equation are the same as the eigenvalues obtained in @xcite from the sturm
liouville problem derived in the fourier space .
numerically , we find for the first two zeros @xmath241 of the limiting wronskian @xmath235 that the decay rate of the sequence @xmath242 remains generic : @xmath243 but the decay rate of the sequence @xmath230 becomes faster : @xmath244 let us now recall the correspondence of eigenvectors of the jacobi difference equation @xmath168 and eigenvectors of the linearized log kdv operator @xmath245 . from the previous transformations ,
we obtain @xmath246 where @xmath247 are respectively the odd and even components of the eigenvector with respect to @xmath24 .
thanks to the decay of the sequences @xmath242 and @xmath230 , we note that @xmath248 but that @xmath249 .
therefore , generally @xmath250 defined by ( [ linearized - operator ] ) .
thus , the eigenvector @xmath62 given by ( [ decomposition - numerics ] ) does not solve the eigenvalue problem @xmath251 in the classical sense compared to the eigenvectors constructed in @xcite with the fourier transform . in order to clarify the sense for the eigenvectors given by ( [ decomposition - numerics ] )
, we denote @xmath252 and project the eigenvalue problem @xmath212 to @xmath149 and @xmath62 .
the projection @xmath253 is uniquely found by @xmath254 which can also be obtained from the projection equation ( [ projection - c-1 ] ) . the component @xmath62 satisfies formally @xmath255 . after separating the even and odd parts of the eigenvalue problem ,
we obtain the coupled system @xmath256 as we have indicated above , it is difficult to prove that each term of the coupled system ( [ linearized - formal ] ) belongs to @xmath31 if @xmath257 and @xmath258 are given by the decomposition ( [ decomposition - numerics ] ) in terms of the hermite functions . in order to formulate the coupled problem ( [ linearized - formal ] ) rigorously , we would like to show that the components of the eigenvector belong to @xmath259 and satisfy the coupled system @xmath260 where each term of system ( [ linearized - classical ] ) is now defined in @xmath31 . to show ( [ function - space ] ) and ( [ linearized - classical ] ) , we proceed as follows .
according to ( [ y - odd - even ] ) and ( [ y - spaces ] ) , @xmath261 is odd , hence @xmath262 and the first constraint in ( [ function - space ] ) is satisfied . since the kernel of @xmath14 is spanned by the odd function , we have @xmath263 so that @xmath264 and @xmath265 as is given by the first equation in ( [ linearized - classical ] ) . similarly , from ( [ y - odd - even ] ) and ( [ y - spaces ] ) , we have @xmath266 so that the second equation in ( [ linearized - classical ] ) implies that @xmath267 . hence , the second constraint in ( [ function - space ] ) is satisfied .
thus , the coupled system ( [ linearized - classical ] ) is well defined for the eigenvector @xmath268 of the eigenvalue problem @xmath251 defined in the function space ( [ function - space ] ) .
note that the formulation ( [ linearized - classical ] ) also settles the issue of zero eigenvalue @xmath269 , which should not be listed as an eigenvalue of the problem @xmath270 .
indeed , the first equation ( [ linearized - classical ] ) with @xmath269 implies @xmath271 , hence @xmath272 , where @xmath189 is a solution of @xmath172 .
thus , the existence of the eigenvector @xmath273 for the eigenvalue @xmath269 of the jacobi difference equation does not imply the existence of the zero eigenvalue @xmath274 in the proper formulation ( [ function - space])([linearized - classical ] ) of the system @xmath251 .
the same result can be obtained from the projection equation ( [ projection - eigenvalue - c-1 ] ) . if @xmath269 , then @xmath275 , which corresponds to the zero solution @xmath18 of the jacobi difference equation @xmath172 .
delicate analytical issues in the decomposition ( [ decomposition - numerics ] ) involving hermite functions are likely to be related to the fact that eigenvectors @xmath67 of the eigenvalue problem @xmath276 decay algebraically as @xmath57 , while the decay of each hermite function @xmath277 in the decomposition ( [ y - odd - even ] ) is given by a gaussian function .
for the kdv equation with exponentially decaying solitary waves , the exponentially weighted spaces were used to introduce effective dissipation in the long - time behavior of perturbations to the solitary waves and to prove their asymptotic stability @xcite . for the log kdv equation with gaussian solitary waves
, it makes sense to introduce gaussian weights in order to obtain a dissipative evolution of the linear perturbations .
here we show how the gaussian weights can be introduced for the linearized log kdv equation ( [ linlogkdv ] ) .
let us represent a solution to the linearized log kdv equation ( [ linlogkdv ] ) in the following form @xmath278 with @xmath279 where @xmath280 are new variables to be found .
it is clear that the representation ( [ convolution - y ] ) imposes restrictions on the class of functions of @xmath62 in the energy space @xmath60 .
we will show that these restrictions are invariant with respect to the time evolution of the linearized log kdv equation ( [ linlogkdv ] ) .
we assume sufficient smoothness and decay of the variable @xmath25 . by using the explicit computation with @xmath281 and @xmath282 ,
we obtain @xmath283 dz \\ & = & \int_{-\infty}^0 w(t , z ) \left [ -u_0(x - z ) - z u_0'(x - z ) + \frac{1}{4 } z^2 u_0(x - z ) \right ] dz.\end{aligned}\ ] ] integrating by parts , we further obtain @xmath284 dz.\end{aligned}\ ] ] also recall that @xmath285 and @xmath286 . bringing together the left - side and the right - side of the linearized log kdv equation ( [ linlogkdv ] ) under the decomposition ( [ convolution])([convolution - y ] ) , we obtain the system of modulation equations @xmath287 and the evolution problem @xmath288 where the linear operator @xmath289 with @xmath290 is given by @xmath291 since @xmath171 is a regular singular point of the differential operator @xmath289 , no boundary condition is needed to be set at @xmath171 .
we show that the differential operator @xmath292 is dissipative in @xmath293 .
[ lemma - dissipative ] for every @xmath294 , we have @xmath295 ^ 2 + \int_{-\infty}^0 \left [ z ( \partial_z w)^2 + \frac{1}{4 } z w^2 \right ] dz { \leqslant}-\frac{1}{2 } \| w \|_{l^2(\mathbb{r}^-)}^2.\ ] ] the proof of the equality follows from integration by parts for every @xmath296 : @xmath297 dz & = & \left [ -z w w_z - w^2 + \frac{1}{8 } z^2 w^2 \right ] \biggr|_{z
\to -\infty}^{z = 0 } + \int_{-\infty}^0 \left [ z ( \partial_z w)^2 + \frac{1}{4 } z w^2 \right ] dz \\ & = & - [ w(0)]^2 + \int_{-\infty}^0 \left [ z ( \partial_z w)^2 + \frac{1}{4 } z w^2 \right ] dz.\end{aligned}\ ] ] this yields the equality in ( [ h - form ] ) .
the inequality in ( [ h - form ] ) is proved from the younge inequality @xmath298 where @xmath299 is at our disposal .
picking @xmath300 yields the inequality in ( [ h - form ] ) .
the semi - group theory for dissipative operators is fairly standard @xcite , so we assume existence of a strong solution to the evolution problem ( [ heat ] ) for every @xmath301 .
the next result shows that this solution decays exponentially fast in the @xmath293 norm .
[ cor - dissipation ] let @xmath302 be a solution of the evolution problem ( [ heat ] ) .
then , the solution satisfies @xmath303 the decay behavior ( [ decay - l-2 ] ) is obtained from a priori energy estimates .
indeed , it follows from ( [ h - form ] ) that @xmath304 gronwall s inequality yields the bound ( [ decay - l-2 ] ) .
we recall that the solution @xmath305 needs to satisfy the constraint @xmath306 .
the constraint is invariant with respect to the time evolution of the linearized log kdv equation ( [ linlogkdv ] ) .
these properties are equivalently represented in the decomposition ( [ convolution])([convolution - y ] ) , according to the following lemma .
[ lemma - constraint ] for every @xmath294 , we have @xmath307 where @xmath221 is constant in @xmath8 .
moreover , if @xmath36 , then @xmath308 and @xmath309 furthermore , @xmath310 is constant in @xmath8 . as an alternative derivation
, one can compute from the evolution problem ( [ heat ] ) that @xmath311 and then use the first modulation equation in system ( [ modulation - equation ] ) for integration in @xmath8 . if @xmath36 , then @xmath37 and @xmath308 .
this yields @xmath312 uniquely by @xmath313 applying the decay bound ( [ decay - l-2 ] ) and the cauchy
schwarz inequality , we obtain the decay bound ( [ decay - a ] ) . [ cor - modulation ]
if @xmath36 , then there is @xmath314 such that @xmath315 as @xmath204 .
it follows from the bound ( [ decay - a ] ) that if @xmath36 , then @xmath316 as @xmath204
. since @xmath312 decays to zero exponentially fast , the assertion of the corollary follows from integration of the second modulation equation in system ( [ modulation - equation ] ) .
besides scattering to zero in the @xmath293 norm , the global solution of the evolution problem ( [ heat ] ) also scatters to zero in the @xmath317 norm .
the following lemma gives the relevant result based on a priori energy estimates .
[ lemma - scattering ] let @xmath25 be a smooth solution of the evolution problem ( [ heat ] ) in a subset of @xmath318 . then
, there exist positive constants @xmath319 and @xmath39 such that @xmath320 the proof is developed similarly to the estimates in lemma [ lemma - dissipative ] and corollary [ cor - dissipation ] but the estimates are extended for @xmath321 . differentiating ( [ dissipative - operator ] ) in @xmath236 , multiplying by @xmath322 , and integrating by parts , we obtain for smooth solution @xmath25 : @xmath323 \biggr|_{z \to -\infty}^{z = 0 } + \int_{-\infty}^0 z w_{zz}^2 dz + \frac{3}{4 } \int_{-\infty}^0 z w_z^2 dz \\ & = & - \frac{3}{2 } [ \partial_z w(0)]^2 + \frac{1}{4 } [ w(0)]^2 + \int_{-\infty}^0 z w_{zz}^2 dz + \frac{3}{4 } \int_{-\infty}^0 z w_z^2 dz.\end{aligned}\ ] ] as a result , smooth solutions to the evolution problem ( [ heat ] ) satisfy the differential inequality @xmath324 ^ 2 + \int_{-\infty}^0 z w_{zz}^2 dz + \frac{3}{4 } \int_{-\infty}^0 z w_z^2 dz.\end{aligned}\ ] ] by using young s inequality , we estimate @xmath325 ^ 2 = 2 \int_{-\infty}^0 z w w_z
dz { \leqslant}\beta^2 \| w_z\|_{l^2(\mathbb{r}^-)}^2 + \beta^{-2 } \| w\|_{l^2(\mathbb{r}^-)}^2\ ] ] and @xmath326 where @xmath319 and @xmath327 are to our disposal .
picking @xmath300 and assuming @xmath328 , we close the differential inequality as follows @xmath329 thanks to the exponential decay in the bound ( [ decay - l-2 ] ) , we can rewrite the differential inequality in the form @xmath330 { \leqslant}\frac{1}{2 \beta^2 } \| w(0 ) \|_{l^2(\mathbb{r}^-)}^2 e^{-\frac{\beta^2}{2 } t}.\end{aligned}\ ] ] integrating over time , we finally obtain @xmath331 where @xmath328 is fixed arbitrarily .
thus , the @xmath318 norm of the smooth solution to the evolution problem ( [ heat ] ) decays to zero exponentially fast as @xmath204 .
the bound ( [ decay - l - infty ] ) follows by the sobolev embedding of @xmath318 to @xmath317 . combining the results of this section , we summarize the main result on the dissipative properties of the solutions to the linearized log kdv equation ( [ linlogkdv ] ) represented in the convolution form ( [ convolution])([convolution - y ] ) .
[ theorem - convolution ] assume that the initial data @xmath36 is represented by the convolution form ( [ convolution])([convolution - y ] ) with some @xmath332 , @xmath333 , and @xmath334 .
there exists a solution of the linearized log kdv equation ( [ linlogkdv ] ) represented in the convolution form ( [ convolution])([convolution - y ] ) with unique @xmath335 and @xmath302 .
moreover , there is a @xmath314 such that @xmath336 the existence result follows from the existence of the semi - group to the evolution problem ( [ heat ] ) and the ode theory for the system of modulation equations ( [ modulation - equation ] ) . since @xmath337 , the scattering result ( [ scattering - result ] ) follows from the generalized younge inequality , as well as the results of corollary [ cor - dissipation ] , lemma [ lemma - constraint ] , corollary [ cor - modulation ] , and lemma [ lemma - scattering ] .
we have obtained new results for the linearized log kdv equation . by using hermite function decompositions in section 4 ,
we have shown analytically how the semi - group properties of the linear evolution in the energy space can be recovered with the jacobi difference operator .
we have also established numerically in section 5 the equivalence between computing the spectrum of the linearized operator with the jacobi difference equation and that with the differential equation . finally , we have used in section 6 the convolution representation with the gaussian weight to show that the solution to the linearized log kdv equation can decay to zero in the @xmath338 norms .
it may be interesting to compare these results with the fourier transform method used in the previous work @xcite . from analysis of eigenfunctions of the spectral problem @xmath212 , it is known that the eigenfunctions are supported on a half - line in the fourier space . the decomposition ( [ decomposition - lin - log - kdv ] ) in terms of the hermite functions in the physical space can be written equivalently as the decomposition in terms of the hermite functions in the fourier space .
the jacobi difference equation representing the spectral problem does not imply generally that the decomposition in the fourier space returns an eigenfunction supported on a half - line .
this property is not explicitly seen in the computation of eigenvectors with the jacobi difference operator .
another interesting observation is as follows .
the linear evolution of the linearized log kdv equation in the fourier space ( [ linlogkdv - fourier ] ) can be analyzed separately for @xmath339 and @xmath340 . since the time evolution
is given by the linear schrdinger - type equation , the fundamental solution is norm - preserving in the energy space . if the gaussian weight is introduced on the positive half - line as follows : @xmath341 then the time evolution is defined in the fourier space by @xmath342 , where the linear operator @xmath343 is given by @xmath344 if @xmath292 and @xmath345 in ( [ h - form ] ) and ( [ h - operator - fourier ] ) are extended on the entire line , then @xmath292 and @xmath345 are fourier images of each other .
thus , a very similar introduction of the gaussian weights ( except , of course , the domains in the physical and fourier space ) may result in either dissipative or norm - preserving solutions of the linearized log kdv equation .
although the results obtained in this work give new estimates and new tools for analysis of the linearized log kdv equation , it is unclear in the present time how to deal with the main problem of proving orbital stability of the gaussian solitary waves in the nonlinear log kdv equation .
this challenging problem will remain open to new researchers . | the logarithmic kdv ( log kdv ) equation admits global solutions in an energy space and exhibits gaussian solitary waves .
orbital stability of gaussian solitary waves is known to be an open problem .
we address properties of solutions to the linearized log kdv equation at the gaussian solitary waves . by using the decomposition of solutions in the energy space in terms of hermite functions ,
we show that the time evolution is related to a jacobi difference operator with a limit circle at infinity .
this exact reduction allows us to characterize both spectral and linear orbital stability of solitary waves .
we also introduce a convolution representation of solutions to the log kdv equation with the gaussian weight and show that the time evolution in such a weighted space is dissipative with the exponential rate of decay . |
during the past decade , there have been important advances in our understanding of the mathematical foundations of quantum electrodynamics with non - relativistic , quantum - mechanical matter ( `` non - relativistic qed '' ) .
subtle spectral properties of the hamiltonians generating the time evolution of atoms and molecules interacting with the quantized radiation field have been established .
in particular , existence of atomic ground states and absence of stable excited states have been proven , and the energies and life times of resonances have been calculated in a rigorously controlled way , for a variety of models ; see @xcite , and references given there . furthermore , some important steps towards developing the scattering theory for systems of non - relativistic matter interacting with massless bosons , in particular photons , have been taken .
asymptotic electromagnetic field operators have been constructed in @xcite , and wave operators for compton scattering have been shown to exist in @xcite .
rayleigh scattering , i.e. , the scattering of photons at atoms below their ionization threshold , has been analyzed in @xcite for models with an infrared cutoff .
the results in this paper are based on methods developed in @xcite .
some earlier results on rayleigh scattering have been derived in @xcite .
compton scattering in models with an infrared cutoff has been studied in @xcite . while it is understood how to calculate scattering amplitudes for various low - energy scattering processes in models _ without _ an infrared cutoff , all known _ general _ methods to prove _ unitarity of the scattering matrix _ on subspaces of states of sufficiently low energy , i.e. , _ asymptotic completeness _
, require the presence of an ( arbitrarily small , but _ positive _ ) infrared cutoff . in @xcite , rayleigh scattering has only been studied for models of atoms with _ static _
( i.e. , _ infinitely heavy _ ) nuclei . as suggested by this discussion , the main challenge in the scattering theory for non - relativistic qed
presently consists in solving the following problems : 1 . to remove the infrared cutoff in the analysis of rayleigh scattering ; 2 . to remove the infrared cutoff in the treatment of compton scattering of photons at one freely moving electron or ion ; 3 . to prove _ asymptotic completeness _ of rayleigh scattering of photons at atoms or molecules with _
dynamical _ nuclei . in this paper , we solve problem iii ) in the presence of an ( arbitrarily small , but positive ) infrared cutoff . in order to render our analysis , which is quite technical , as simple and transparent as possible , we consider the simplest model exhibiting all typical features and challenges encountered in an analysis of problem iii ) .
we consider a hydrogen atom or a one - electron ion .
the nucleus is described , like the electron , as a non - relativistic quantum - mechanical point particle of _ finite _ mass .
we ignore the spin degrees of freedom of the electron and the nucleus .
the electron and the nucleus interact with each other through an attractive two - body potential @xmath0 , which can be chosen to be the electrostatic coulomb potential .
electron and nucleus are coupled to a quantized radiation field . as in @xcite , the field quanta of the radiation field are massless bosons , which we will call photons .
physically , the radiation field is the quantized electromagnetic field .
however , the helicity of the photons does not play an interesting role in our analysis , and we therefore consider _ scalar _ bosons .
as announced , we focus our attention on rayleigh scattering ; i.e. , we only consider the asymptotic dynamics of states of the atom and the radiation field with energies below the threshold for break - up of the atom into a freely moving nucleus and electron , i.e. , below the ionization threshold .
moreover , we introduce an _ infrared cutoff _ : photons with an energy below a certain arbitrarily small , but positive threshold energy do _ not _ interact with the nucleus and the electron .
while all our other simplifications are purely cosmetic , the presence of an infrared cutoff is crucial in our proof of asymptotic completeness ( but _ not _ for most other results presented in this paper ) . in the model we study , the atom can be located anywhere in physical space and can move around freely , and the hamiltonian of the system is translation - invariant .
this feature suggests to combine and extend the techniques in two previous papers , @xcite ( rayleigh scattering with static nuclei ) and @xcite ( compton scattering of photons at freely moving electrons ) , and this is , in fact , the strategy followed in the present paper .
as in @xcite , to prove asymptotic completeness we must impose an upper bound on the energy of the state of the system that guarantees that the center of mass velocity of the atom does not exceed one third of the velocity of light .
this is a purely technical restriction which we believe can be replaced by one that guarantees that the velocity of the atom is less than the velocity of light .
we note that , for realistic atoms , the condition that the total energy of a state be below the ionization threshold of an atom at rest automatically guarantees that the speed of the atom in an arbitrary internal state is much less than a third of the speed of light .
next , we describe the model studied in this paper more explicitly . the hilbert space of pure states is given by @xmath1 where the variables @xmath2 and @xmath3 are the positions of the nucleus and of the electron , respectively , and @xmath4 is the symmetric fock space over the one - photon hilbert space @xmath5 , where the variable @xmath6 denotes the momentum of a photon .
vectors in @xmath4 describe pure states of the radiation field .
the hamiltonian generating the time evolution of states of the system is given by @xmath7 where @xmath8 is the hamiltonian of the atom decoupled from the radiation field . here
@xmath9 and @xmath10 are the momentum operators of the electron and the nucleus , respectively , and @xmath11 and @xmath12 are their masses .
the term @xmath13 is the potential of an attractive two - body force , e.g. , the electrostatic coulomb force , between the electron and the nucleus ; ( @xmath14 is negative and tends to zero , as @xmath15 , and it is assumed to be such that the spectrum of @xmath16 is bounded from below and has at least one negative eigenvalue ) . the operator @xmath17 on the r.h.s . of ( [ eq : ham1 ] )
is the hamiltonian of the free radiation field .
it is given by latexmath:[\[h_f = \int { { \rm d}}k \ , with momentum @xmath6 , and @xmath19 , @xmath20 are the usual boson creation- and annihilation operators : for every function @xmath21 , @xmath22 are densely defined , closed operators on the fock space @xmath4 , and , for @xmath23 , they satisfy the canonical commutation relations @xmath24 = ( f , h ) , \quad [ a^{\sharp } ( f ) , a^{\sharp } ( h ) ] = 0 , \ ] ] where @xmath25 denotes the scalar product of @xmath26 and @xmath27 . for @xmath28 , a field operator @xmath29 is defined by @xmath30 it is a densely defined , self - adjoint operator on @xmath4 .
the functions ( form factors ) @xmath31 and @xmath32 on the right side of are given by @xmath33 where @xmath34 and @xmath35 belong to the schwartz space , and @xmath36 for some @xmath37 ( infrared cutoff ) .
in many of our results , we could pass to the limit @xmath38 ; but in our proof of asymptotic completeness of rayleigh scattering , the condition that @xmath37 is essential . finally ,
the parameter @xmath39 on r.h.s . of ( [ eq : ham1 ] ) is a coupling constant ; it is assumed to be non - negative and will be chosen sufficiently small in the proofs of our results .
( it should be noted that we are using units such that planck s constant @xmath40 and the speed of light @xmath41 , and we work with dimensionless variables @xmath42 chosen such that @xmath16 and @xmath17 are independent of @xmath39 . ) in the description of the atom , it is natural to use the following variables : @xmath43 here @xmath44 is the position of the center of mass of the atom , and @xmath45 is the position of the electron relative to the one of the nucleus .
then @xmath46 where @xmath47 , @xmath48 , @xmath49 , and @xmath50 ; ( center - of - mass momentum , relative momentum , total mass , reduced mass , respectively ) .
self - adjointness of @xmath51 on @xmath52 ( under appropriate assumptions on @xmath0 ) is a standard result . in this paper
, we study the dynamics generated by @xmath51 on the subspace of states in @xmath52 whose maximal energy is below the ionization threshold @xmath53 where @xmath54 is the subspace of vectors in @xmath52 with the property that the distance between the electron and the nucleus is at least @xmath55 .
vectors in @xmath52 with a maximal total energy _ below _
@xmath56 exhibit exponential decay in @xmath57 , the distance between the electron and the nucleus ; see @xcite . under our assumptions , @xmath58 , for a finite constant @xmath59 depending only on @xmath34 and @xmath35 .
when @xmath60 then @xmath61 , which is the ionization threshold of a one - electron atom or -ion decoupled from the radiation field .
our choice of the hamiltonian @xmath51 , see and , and of the form factors @xmath62 and @xmath63 , see , makes it clear that the dynamics of the system is space - translation invariant : let @xmath64 denote the momentum operator of the radiation field , and let @xmath65 be the total momentum operator .
it is easy to check that @xmath66 = 0\qquad \text{(translation invariance)}.\ ] ] it is then useful to consider direct - integral decompositions of the space @xmath52 and the operator @xmath51 over the spectrum of the total momentum operator @xmath67 ( which , as a set , is @xmath68 ) .
thus @xmath69 and @xmath70 where the fiber hamiltonian @xmath71 is the operator on the fiber hilbert space @xmath72 given by @xmath73 where @xmath74 is the center - of - mass momentum of the atom , and @xmath75 is the hamiltonian describing the relative motion of the electron around the nucleus .
we are now in the position to summarize the main results proven in this paper for the model introduced above . in a first part
, we analyze the energy spectra of the fiber hamiltonians @xmath76 below a certain threshold @xmath77 , where @xmath56 is given in , and @xmath78 ; @xmath79 is the ground state energy of @xmath80 , and @xmath81 is a constant @xmath82 ( @xmath83speed of light , in our units ) .
the condition @xmath84 guarantees that the electron is bound to the nucleus , with exponential decay in @xmath45 , and @xmath85 implies that , for a sufficiently small coupling constant @xmath39 , the center - of - mass velocity of the atom is smaller than the speed of light .
( for center - of - mass velocities @xmath86 , the ground state energy of the atom decoupled from the radiation field is embedded in continuous spectrum , and the ground state becomes unstable when the coupling to the radiation field is turned on . ) for realistic atoms , in particular for hydrogen , @xmath87 , so that @xmath84 is the only relevant condition .
we let @xmath88 denote the ground state energy of @xmath76 , and we define @xmath89 we prove that , for every @xmath90 , @xmath91 is a simple eigenvalue of @xmath76 , i.e. , that the atom has a _ unique ground state _ ,
provided @xmath39 is small enough .
this is a result that is expected to survive the limit @xmath92 , provided the factors @xmath34 and @xmath35 are not too singular at @xmath93 .
for the pauli - fierz model of non - relativistic qed , existence of a ground state can be proven under conditions similar to the ones described above , _ provided _ the total charge of electrons and nucleus vanishes ; see @xcite . by appropriately modifying mourre theory in a form developed in @xcite
, we prove that the spectrum of @xmath76 in the interval @xmath94 is purely continuous .
with relatively little further effort , our methods would also show that @xmath95 is absolutely continuous .
( these results , too , would survive the removal of the infrared cutoff , @xmath96 .
this will not be shown in this paper ; but see @xcite . )
we denote the ground state of @xmath97 , by @xmath98 ; ( @xmath98 is called the _ dressed atom _
( ground ) _ state _ of momentum @xmath67 ) .
the space of wave packets of dressed atom states , @xmath99 , is the subspace of the total hilbert space @xmath52 given by @xmath100 this space is invariant under the time evolution .
in fact , @xmath101 , where , for @xmath102 , @xmath103 , for all times @xmath104 . in a second part of our paper , scattering theory
is developed for the models introduced above .
we first construct asymptotic photon creation- and annihilation operators @xmath105 where @xmath106 is the free time evolution of a one - photon state @xmath107 . to ensure the existence of the strong limit on the r.h.s . of
, we assume that @xmath108 , that @xmath109 belongs to the range of the spectral projection , @xmath110 , of @xmath51 corresponding to the interval @xmath111 $ ] , with @xmath112 , as above , and @xmath113 , and that the coupling constant @xmath39 is so small ( depending on @xmath114 ) that the velocity of the center of mass of the atom is smaller than one .
the last condition ensures that the distance between the atom and a configuration of outgoing photons increases to infinity and hence the interaction between these photons and the atom tends to @xmath115 , as time @xmath104 tends to @xmath116 .
the details of the proof of are very similar to those in @xcite . from and we infer that , for @xmath117 , and under the conditions of existence of the limit in , vectors of the form @xmath118 exist , and their time evolution is the one of freely moving photons and a freely moving atom : @xmath119 as @xmath120 .
furthermore , @xmath121 , under the same assumptions .
equation provides the justification for calling the vectors @xmath122 _ scattering states_. we already know that the atom does _ not _ have any stable excited states .
it is therefore natural to expect that the time evolution of an arbitrary vector in the range of the spectral projection @xmath123 , with @xmath124 as above , approaches a vector describing a configuration of freely moving photons and a freely moving atom in its ground state , as time @xmath104 tends to @xmath125 .
thus , with and , we expect that , for @xmath124 , @xmath126 where @xmath127 denotes the linear subspace spanned by a set , @xmath128 , of vectors in @xmath52 , and @xmath129 denotes the closure of @xmath130 in the norm of @xmath52 .
property is called _
asymptotic completeness of rayleigh scattering_. the _ main result _ of this paper is a proof of under the supplementary condition that @xmath131 for some @xmath132 ( the proof of is the only part of the paper where , for technical reasons , we need to assume @xmath133 ; all other results only require @xmath134 ) . next , we reformulate in a more convenient language .
we define a hilbert space of scattering states as the space @xmath135 , and we introduce an asymptotic hamilton operator , @xmath136 , by setting @xmath137 where @xmath138 for arbitrary @xmath139 , with @xmath140 , as above . on the range of @xmath141 ,
the operators @xmath142 , given by @xmath143 exist ; see .
the vector @xmath144 is the vacuum in the fock - space characterized by the property that @xmath145 , for @xmath146 .
the operators @xmath147 and @xmath148 are called _ wave operators _ , and the _ scattering matrix _ is defined by @xmath149 from eqs .
and we find that @xmath150 and hence the ranges of @xmath147 and @xmath148 are contained in the range of @xmath110 .
using that @xmath151 , for @xmath152 and @xmath153 , one sees that @xmath154 and @xmath155 are isometries from the range of @xmath141 into @xmath52 .
if we succeeded in proving that @xmath156 we would have established the unitarity of the @xmath128-matrix , defined in , on @xmath157 , i.e. , asymptotic completeness of rayleigh scattering . in order to prove ( [ eq22 ] ) , we show that @xmath142 have right inverses defined on @xmath158 .
our proof is inspired by proofs of similar results in @xcite and in @xcite .
it is based on constructing a so called _ asymptotic observable _ @xmath159 and then proving that @xmath159 is _ positive _ on the orthogonal complement of @xmath99 in @xmath123 .
the proof of this last result is , perhaps , the most original accomplishment in this paper and is based on some new ideas . our paper is organized as follows . in section [ sec : model ] , we define our model more precisely , and we state our assumptions on the potential @xmath14 and on the form factors @xmath62 and @xmath63 . in section [ sec : spectrum ] , we study the spectrum of the fiber hamiltonian @xmath76 : in section [ sec : das ] , we prove the existence of dressed atom states , and , in section [ sec : poscomm ] , we prove two positive commutator estimates , from which we conclude that the spectrum of @xmath76 above the ground state energy and below an appropriate threshold is continuous . in section [ sec : scatt ] , we discuss the scattering theory of the system .
first , in section [ sec : waveop ] , we prove the existence of asymptotic field operators , we recall some of their properties , we prove the existence of the wave operators , and we state our main theorem . then , in section [ sec : modham ] , we introduce a modified hamiltonian , @xmath160 , describing `` massive '' photons , and we explain why it is enough to prove asymptotic completeness for @xmath160 instead of @xmath51 . in section [ sec : observable ] , we construct asymptotic observables @xmath159 and inverse wave operators @xmath161 . in section [ sec : pos ] , we prove positivity of our asymptotic observables when restricted to the orthogonal complement of @xmath162 ( the space of wave packets of dressed atom states ) .
finally , in section [ sec : ac ] , we complete the proof of asymptotic completeness . in appendix
[ sec : fock ] , we introduce some notation , used throughout the paper , concerning operators on the bosonic fock space . in appendix
[ sec : bounds ] , we summarize bounds used to control the interaction between the electron ( or the nucleus ) and the radiation field .
we consider a non - relativistic _ atom _ consisting of a _ nucleus _ and an _ electron _ interacting through a two - body potential @xmath14 .
the hamiltonian describing the dynamics of the atom is the self - adjoint operator @xmath163 acting on the hilbert space @xmath164 , where @xmath2 and @xmath3 denote the position of the nucleus and of the electron , respectively , and @xmath165 and @xmath166 are the corresponding momenta .
we assume that the interaction potential @xmath14 satisfies the following assumptions . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * hypothesis ( h0 ) : * the potential @xmath0 is a locally square integrable function , with @xmath167 , and such that @xmath14 is infinitesimally small with respect to the laplace operator @xmath168 , in the sense that , for all @xmath169 there exists a finite constant @xmath170 such that @xmath171 for every @xmath172 .
moreover the potential @xmath0 is such that the ground state energy of the schrdinger operator @xmath173 ( with @xmath174 ) is non - degenerate . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ remarks . _
* hypothesis ( h0 ) is satisfied by the coulomb potential @xmath175 and it is inspired by this potential .
it follows from ( h0 ) that the hamiltonian @xmath16 with domain @xmath176 is a self - adjoint operator on the hilbert space @xmath177 and that it is bounded from below . * with little more effort we could have covered a much larger class of locally square integrable potentials @xmath0 where only the negative part @xmath178 is infinitesimally small with respect to @xmath179 and @xmath16 is self - adjointly realized in terms of a friedrich s extension .
this would allow , e.g. , for confining potentials that tend to @xmath180 as @xmath181 . * note that we are neglecting the degrees of freedom corresponding to the spin of the nucleus and of the electron , because they do not play an interesting role in the scattering process .
next we couple the atom to a quantized scalar radiation field .
we call the particles described by the quantized field _ ( scalar ) photons_. the pure states of the photon field are vectors in the bosonic fock space over the one - particle space @xmath5 , @xmath182 where @xmath183 denotes the subspace of @xmath184 consisting of all functions which are completely symmetric under permutations of the @xmath185 arguments .
the variables @xmath186 denote the momenta of the photons .
the dispersion relation of the photons is given by @xmath187 , denoted by @xmath188 .
formally , @xmath189 where @xmath190 and @xmath191 are the usual creation- and annihilation operators on @xmath4 , satisfying the canonical commutation relations @xmath192 = 0 $ ] , @xmath193 = \delta ( k - k')$ ] .
more notations for operators on fock space that are used throughout the paper are collected in appendix [ sec : fock ] . the total system , atom plus quantized radiation field , has the hilbert space @xmath194 ; its dynamics is generated by the hamiltonian @xmath195 where @xmath39 is a real non - negative coupling constant ( the assumption @xmath196 is not needed , it just makes the notation a little bit simpler ) , and where @xmath197 the form factors @xmath198 and @xmath199 are square integrable functions of @xmath6 with values in the multiplication operators on @xmath200 .
clearly , @xmath201 describes the interaction between the electron and the radiation field , and @xmath202 couples the field to the nucleus .
the next hypothesis specifies our assumptions on the form factors @xmath203 and @xmath204 . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * hypothesis ( h1 ) : * the form factors @xmath203 and @xmath204 have the form @xmath205 where @xmath206 belong to schwartz space @xmath207 , and @xmath208 if @xmath209 , for some @xmath210 . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the particular form of @xmath203 and @xmath204 given in ( [ eq : gx ] ) guarantees the translation invariance of the system ( see the discussion after ( [ eq : trans ] ) ) .
the presence of an infrared cutoff @xmath211 in @xmath34 and @xmath35 is used in the proofs of many of our results ; but it is not necessary for the existence of the asymptotic field operators and for the existence of the wave operator in section [ sec : waveop ] .
notice that , even though our main results require the coupling constant @xmath39 to be sufficiently small , how small @xmath39 has to be does not depend on the infrared cutoff @xmath212 , in the following sense . if we define @xmath213 and @xmath214 , with @xmath215 , and with @xmath216 monotone increasing and such that @xmath217 if @xmath218 and @xmath219 if @xmath220 , then hypothesis ( h1 ) is satisfied for every choice of @xmath210 .
moreover how small @xmath39 has to be is independent of the choice of the parameter @xmath212 . assuming hypotheses ( h0 ) and ( h1 ) , the hamiltonian @xmath51 , defined on the domain @xmath221 , is essentially self - adjoint and bounded from below .
this follows from lemma [ lm : estim ] , which shows , using hypothesis ( h1 ) , that the interaction @xmath222 is infinitesimal with respect to the free hamiltonian @xmath223 . to study the system described by the hamiltonian @xmath51 , it is more convenient to use coordinates describing the center of mass of the atom and the relative position of the nucleus and the electron .
we define @xmath224 then , the atomic hamiltonian @xmath16 becomes @xmath225 where @xmath226 is the center of mass momentum of the atom , and @xmath227 is the momentum conjugate to the relative coordinate @xmath45 .
moreover , @xmath228 is the total atomic mass , and @xmath229 is the reduced mass . expressed in the new coordinates , the total hamiltonian of the system is given by @xmath230 here @xmath231 and @xmath232 , and we use the notation @xmath233 with @xmath234 the fact that the form factors @xmath34 and @xmath35 contain an infrared cutoff ( meaning that @xmath235 , if @xmath209 ) implies that photons with very small momenta do not interact with the atom .
in other words , they decouple from the rest of the system .
we denote by @xmath236 ( the subscript @xmath237 stands for `` interacting '' ) the characteristic function of the set @xmath238 . then the operator @xmath239 , whose action on the @xmath185-particle sector of @xmath4 is given by @xmath240 defines the orthogonal projection onto states without soft bosons .
the fact that soft bosons do not interact with the atom implies that @xmath51 leaves the range of @xmath239 invariant ; @xmath51 commutes with @xmath239 .
another way to isolate the soft , non - interacting , photons from the rest of the system is as follows .
we have that @xmath241 , where @xmath242 is the open ball of radius @xmath212 around the origin and @xmath243 denotes its complement .
hence the fock space can be decomposed as @xmath244 , where @xmath245 is the bosonic fock space over @xmath246 ( describing interacting photons ) , and @xmath247 is the bosonic fock space over @xmath248 ( describing soft , non - interacting , photons ) . accordingly
, the hilbert space @xmath249 can be decomposed as @xmath250 by @xmath251 we denote the unitary map from @xmath52 to @xmath252 .
the action of the hamiltonian @xmath51 on @xmath252 is then given by @xmath253 with @xmath254 note that in the representation of the system on the hilbert space @xmath252 , the projection @xmath239 ( projecting on states without soft bosons ) is simply given by @xmath255 , where @xmath256 denotes the orthogonal projection onto the vacuum @xmath144 in @xmath247 .
one of the most important properties of the hamiltonian @xmath51 is its invariance with respect to translations of the whole system , atom and field .
more precisely , defining the total momentum of the system by @xmath257 we have that @xmath258 = 0 $ ] .
because of this property , it can be useful to rewrite the hilbert space @xmath259 as a direct integral over fibers with fixed total momentum .
specifically , we define the isomorphism @xmath260 as follows . for @xmath261
, we define @xmath262 with @xmath263 where @xmath264 is the fourier transform of @xmath265 with respect to its first variable . because of its translation invariance
, the hamiltonian @xmath51 leaves invariant each fiber with fixed total momentum of the hilbert space @xmath266 .
more precisely , @xmath267 where we put @xmath268 . recall that @xmath269 .
note that , for every fixed @xmath67 , the operator @xmath76 is a self - adjoint operator on the fiber space @xmath270 .
our first results , stated in the next section , describe the structure of the spectrum of the fiber - hamiltonian @xmath76 , for fixed values of @xmath67 .
the first question arising in the analysis of the spectrum of @xmath271 concerns the existence of a ground state of @xmath76 : we wish to know whether or not @xmath272 is an eigenvalue of @xmath76 .
we will answer this question affirmatively , under the assumption that the energy @xmath273 lies below some threshold and that the coupling constant is sufficiently small .
the restriction to small energies is necessary to guarantee that the center of mass of the atom does not move faster than with the speed of light ( @xmath41 in our units ) , and that the atom is not ionized .
the following lemma ( and its corollary ) proves that an upper bound on the total energy is sufficient to bound the momentum of the center of mass of the atom ( provided the coupling constant is sufficiently small ) and to make sure that the electron is exponentially localized near the nucleus .
[ lm : sigma ] assume that hypotheses ( h0 ) and ( h1 ) are satisfied .
* define @xmath274 , and fix @xmath275 .
suppose @xmath276 .
then there is @xmath277 such that @xmath278 for all @xmath279 , and for all @xmath280 . in particular @xmath281 . *
define the ionization threshold @xmath282 with @xmath283 .
let @xmath284 be such that @xmath285 .
then @xmath286 \i ) fix @xmath169 such that @xmath287 . choose @xmath288 such that @xmath219 for @xmath289 , and @xmath290 if @xmath291 .
then we have that @xmath292 where @xmath293 is the non - interacting fiber hamiltonian and where the constant @xmath59 is independent of @xmath67 . to prove the last equation note that , if @xmath294 denotes an almost analytic extension of @xmath295 ( see appendix a in @xcite for a short introduction to the helffer - sjstrand functional calculus )
, we have that @xmath296 and therefore @xmath297 uniformly in @xmath67 .
next , since @xmath298 , @xmath299 , and by the definition of @xmath295 , we have that @xmath300 since @xmath301 , we conclude from ( [ eq : unif ] ) that @xmath302 for @xmath39 sufficiently small ( independently of @xmath67 ) . as for part
ii ) , we use theorem 1 of @xcite and an estimate from its proof . given @xmath303 and @xmath304 , let @xmath305 where @xmath306 and @xmath307 .
suppose for a moment that @xmath308 for all @xmath309 and all @xmath310 .
then @xmath311 and hence @xmath312 by ( * ? ? ?
* theorem 1 ) .
moreover , the value of the parameter @xmath55 in the proof of ( * ? ? ?
* theorem 1 ) , in the case of the hamiltonian @xmath76 , can be chosen independent of @xmath67 thanks to .
it follows that the estimate for @xmath313 from that proof is also independent of @xmath67 .
it thus remains to prove . to this end
we proceed by contradiction , assuming that ( [ eq : uni - pi ] ) is wrong .
then there exist @xmath314 and @xmath169 such that @xmath315 hence , we find @xmath316 with @xmath317 and with @xmath318 moreover , since the map @xmath319 , for fixed @xmath320 , is continuous in @xmath67 ( it is just a quadratic function in @xmath67 ) , there exists @xmath321 such that @xmath322 for all @xmath67 with @xmath323 .
next , we choose @xmath324 ( where @xmath325 denotes the ball of radius @xmath326 around @xmath327 ) with @xmath328 , and we define @xmath329 by @xmath330 from ( [ eq : ph0hph0 ] ) , we obtain that @xmath331 because @xmath332 .
since @xmath333 ( which is clear from the construction of @xmath109 ) , this contradicts the definition of @xmath334 . in the next proposition
we prove the existence of a simple ground state for @xmath76 , provided the energy is lower than a threshold energy @xmath114 and the coupling constant is small enough .
[ prop : gs ] assume hypotheses ( h0 ) and ( h1 ) are satisfied . fix @xmath134 and choose @xmath335 ( see lemma [ lm : sigma ] for the definition of @xmath336 and @xmath56 ) .
then , for @xmath39 sufficiently small ( depending on @xmath81 and @xmath114 ) , @xmath337 is a simple eigenvalue of @xmath76 , provided that @xmath338 .
_ since @xmath339 it suffices that @xmath340 and that @xmath39 is small enough .
the proof is very similar to the proof of theorem 4 in @xcite .
for completeness we repeat the main ideas , but we omit details . in order to prove the proposition , we consider the modified hamiltonian @xmath160 ( defined in section [ sec : modham ] ) given , on the fiber with fixed total momentum @xmath67 , by @xmath341 where the dispersion law @xmath342 ( with @xmath343 if @xmath344 and @xmath345 for all @xmath6 ) is assumed to satisfy hypothesis ( h2 ) of section [ sec : modham ] . set @xmath346 . note that @xmath347 and @xmath76 act identically on the range of @xmath239 , the orthogonal projection onto the subspace of vectors without soft bosons .
the proof of the proposition is divided into four steps .
* suppose @xmath348 . then , for sufficiently small @xmath39 ( depending on @xmath81 and @xmath114 ) , we have that @xmath349 for every @xmath169 .
this inequality follows by perturbation of the free hamiltonian ( see lemma 35 in @xcite for details ) .
* @xmath350 .
moreover , if @xmath351 is an eigenvector of @xmath71 ( or of @xmath347 ) corresponding to the eigenvalue @xmath273 , then @xmath352 . in particular
, @xmath351 is an eigenvector of @xmath76 corresponding to the eigenvalue @xmath273 if and only if @xmath351 is an eigenvector of @xmath347 corresponding to the eigenvalue @xmath353 . in order to prove these statements note that the hamiltonians @xmath76 and @xmath347 act on the fiber space @xmath354 , where @xmath247 is the fock space of the soft bosons , @xmath355
. thus @xmath356 the restriction of @xmath71 to the subspace @xmath357 with exactly @xmath185 soft bosons is given by @xmath358 with @xmath359 in the last inequality we used the result of part ( 1 ) .
this proves that @xmath360 since @xmath361 , we conclude that @xmath362 .
( [ eq : sof ] ) also proves that eigenvectors of @xmath71 corresponding to the energy @xmath363 , if they exist , belong to the range of @xmath239 . that the same is true for eigenvectors of @xmath347 corresponding to the energy @xmath273 follows from an inequality for @xmath347 analogous to ( [ eq : sof ] ) . *
if @xmath348 , and for @xmath39 sufficiently small ( depending on @xmath81 and @xmath114 ) we have that @xmath364 for @xmath365 , this follows from part ( 1 ) ( because @xmath366 ) , while for @xmath367 , this inequality follows from @xmath368 , by ( [ eq : part1 ] ) and by construction of @xmath369 ( see section [ sec : modham ] ) .
* @xmath370 + where @xmath371 is defined like @xmath56 with @xmath51 replaced by @xmath347 .
( recall from 2 ) that @xmath353 ) .
the proof of part 4 ) is very similar to the proof of lemma 36 in @xcite with a small modification at the beginning .
we first need to localize with respect to the relative coordinate @xmath45 .
that is , we choose @xmath372)$ ] with @xmath373 for @xmath374 , @xmath375 for @xmath376 and @xmath377 .
let @xmath378 .
then @xmath379 as @xmath380 . as in @xcite , one shows that @xmath381 with @xmath382 relatively compact w.r.t .
@xmath347 , while @xmath383 as @xmath380 follows from the definition of @xmath371 .
part 4 ) follows from these estimates applied to the r.h.s .
of ( [ eq : zero ] ) .
along with 1 ) and 2 ) , and since @xmath384 for every @xmath67 ( see ( [ eq : uni - pi ] ) and its proof ) , this proves that @xmath273 is an eigenvalue of @xmath385 , provided that @xmath348 and @xmath39 is sufficiently small ( depending on @xmath114 ) .
the proof of the fact that @xmath363 is a simple eigenvalue is given in corollary [ cor : spectrum ] , below . from now on , for fixed @xmath134 and @xmath67 such that @xmath386 , we denote by @xmath98 the unique ( up to a phase ) normalized ground state vector of @xmath76 .
the vector @xmath98 is called a _ dressed atom state _ with fixed total momentum @xmath67 .
the space of dressed atom wave packets , @xmath387 , is defined by @xmath388 where @xmath389 is the one - dimensional space spanned by the vector @xmath98 ; @xmath99 is a closed linear subspace left invariant by the hamiltonian @xmath51 . in fact , @xmath51 commutes with the orthogonal projection @xmath390 , onto @xmath99 .
this follows from @xmath391 . from hypothesis ( h0 ) , and from standard results in the theory of schrdinger operators ( see @xcite )
, it follows that the spectrum of @xmath392 in the negative half - axis @xmath393 is discrete .
we denote the negative eigenvalues of @xmath80 by @xmath394 .
the eigenvalues @xmath395 can accumulate at zero only . if @xmath80 has no eigenvalues ( a possibility which is not excluded by our assumptions ) , then our results ( which only concern states of the system for which the electron is bound to the nucleus ) are trivial .
we denote the ( finite ) multiplicity of the eigenvalue @xmath395 by @xmath396 . for fixed @xmath397 ,
we denote by @xmath398 , @xmath399 , an orthonormal basis of the eigenspace of @xmath80 corresponding to the eigenvalue @xmath395 . by hypothesis ( h0 ) ,
the lowest eigenvalue , @xmath79 , of @xmath80 is simple ( @xmath400 ) .
the unique ( up to a phase ) ground state vector of @xmath80 is denoted by @xmath320 . for every fixed @xmath67 , the free hamiltonian
, @xmath401 has eigenvalues @xmath402 with multiplicity @xmath396 corresponding to the eigenvectors @xmath403 ( where @xmath404 denotes the fock vacuum ) , for every @xmath397 . for @xmath405 , @xmath406 , while all other eigenvalues @xmath407 , @xmath408 , are embedded in the continuous spectrum ( see fig .
[ fig : spectrum ] ) . for @xmath409 ,
_ all _ eigenvalues of @xmath410 are embedded in the continuous spectrum , and the hamiltonian @xmath410 does not have a ground state .
we restrict our attention to the physically more interesting case @xmath411 ; ( this will be ensured by the condition that the total energy of the system is less than @xmath412 and that the coupling constant @xmath39 is sufficiently small ) .
one then expects the embedded eigenvalues @xmath413 , @xmath408 , to dissolve and to turn into resonances when the perturbation @xmath414 is switched on .
we prove that this is indeed the case , provided the lifetime of the resonances , as predicted by fermi s golden rule in second order perturbation theory , is finite . for given @xmath415 and @xmath416
, we define the @xmath417 matrix @xmath418 for given @xmath397 , we then define the resonance matrix by setting @xmath419 where @xmath420 denotes the dirac delta - function .
note that @xmath421 is an @xmath422 matrix . according to second order perturbation theory ( fermi s golden rule ) ,
the @xmath396 eigenvalues of @xmath421 are inverses of the lifetimes of the resonances bifurcating from the unperturbed eigenvalue @xmath423 .
we put @xmath424 instability of the eigenvalue @xmath413 is equivalent , in second order perturbation theory , to the statement that @xmath425 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * hypothesis ( h2 ) : * for fixed @xmath134 and @xmath426 , we assume that @xmath427 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in order to prove the absence of embedded eigenvalues we use the technique of positive commutators .
we prove the positivity of the commutator between the hamiltonian @xmath71 and a suitable conjugate operator @xmath428 .
then the absence of eigenvalues follows with the help of a virial theorem .
we make use of ideas from @xcite , adapting them to our problem . for fixed @xmath408
, we construct a suitable conjugate operator @xmath428 , and , in proposition [ prop : poscomm1 ] , we prove the positivity of the commutator @xmath429 $ ] when restricted to an energy interval @xmath179 containing the unperturbed eigenvalue @xmath413 but no other eigenvalues of @xmath410 . in proposition
[ prop : poscomm2 ] , we then establish a commutator estimate on an energy interval around the ground state energy @xmath430 of @xmath410 . for fixed @xmath408
we define latexmath:[\[p_j = \sum_{\alpha } |{\varphi}_{j,\alpha}\rangle \langle { \varphi}_{j,\alpha}| \otimes p_{\omega } , \quad \quad p_{\omega } = the orthogonal projection onto the eigenspace of @xmath410 corresponding to the eigenvalue @xmath432 .
we also define the symmetric operator @xmath433 where @xmath434 and @xmath435 in ( [ eq : d ] ) , we introduced the notation @xmath436 and we used @xmath437 note that @xmath438 strongly , as @xmath439 .
the real parameters @xmath440 and @xmath441 will be fixed later on .
[ prop : poscomm1 ] we assume that hypotheses ( h0)-(h2 ) are satisfied .
we fix @xmath442 and choose @xmath443 ; ( see lemma [ lm : sigma ] for the definition of @xmath336 and @xmath56 ) .
moreover , we suppose that the interval @xmath444 is such that @xmath445 and @xmath446 then , for @xmath447 sufficiently small ( depending on @xmath81 , @xmath114 and the distance @xmath448 ) , one can choose @xmath441 and @xmath440 such that @xmath449 e_{\delta } ( h_g(\pi ) ) \geq c e_{\delta } ( h_g ( \pi ) ) \ , , \ ] ] with a positive constant @xmath59 .
_ remarks . _ 1 .
the choice of the parameter @xmath450 and of the constant @xmath59 depends on the value of @xmath39 .
we can choose , for example , @xmath451 for @xmath452 .
2 . how small @xmath39 has to be chosen depends on the values of @xmath81 , on @xmath453 , and on the distance @xmath448 in ( [ eq : dist ] ) . we must have that @xmath454 , @xmath455 , and @xmath456 ( for an arbitrary choice of @xmath457 with @xmath458 ) .
moreover , @xmath39 has to be sufficiently small , in order for eq .
( [ eq : sigma1 ] ) to hold true ( and thus @xmath39 depends on the choice of the threshold @xmath114 ) .
3 . in this proposition , we do not need the infrared cutoff in the interaction ( i.e. , we can choose @xmath459 ) .
however , the infrared cutoff is needed in the proof of proposition [ prop : virial ] ( the virial theorem ) , and hence in the proof of the absence of embedded eigenvalues .
we begin with a formal computation of the commutator @xmath460 $ ] : @xmath461 = n - \frac{(\pi -{\mathrm{d}\gamma}(k))}{m } \cdot { \mathrm{d}\gamma}(\hat{k } ) - g \phi ( ia f_x ) - [ h_g(\pi),d]\,.\ ] ] recall that @xmath462 . using eq .
( [ eq : sigma1 ] ) , it is easy to check that , for @xmath179 as above , @xmath463 thus , defining @xmath464\ , , \ ] ] we conclude that @xmath465 e_{\delta } ( h_g(\pi ) ) \geq e_{\delta } ( h_g(\pi ) ) b e_{\delta } ( h_g(\pi)),\ ] ] and it is enough to prove the positivity of the r.h.s . of the last equation to complete the proof .
the advantage of working with @xmath466 , instead of the commutator @xmath467 $ ] , is that @xmath466 is bounded with respect to the hamiltonian @xmath76 while @xmath429 $ ] is not ( since the number operator @xmath468 is not bounded with respect to @xmath76 ) . in order to prove that @xmath469 we first establish the inequality @xmath470 to this end
we may assume that @xmath471 for otherwise ( [ eq : pos2 ] ) holds with @xmath472 .
this assumption will allow us to apply the feshbach map with projection @xmath473 to the operator @xmath474 .
indeed , this operator restricted to @xmath475 is invertible for small @xmath39 and @xmath476 as is shown in step 1 .
step 1 through step 5 prepare the proof of ( [ eq : pos2 ] ) .
_ there exists a constant @xmath477 , independent of @xmath67 and @xmath39 , such that @xmath478 in fact , @xmath479 applying lemma [ lm : phibounds ] of appendix [ sec : bounds ] to bound @xmath480 , using that @xmath481 ( by the choice of the interval @xmath179 ) , and that @xmath482 , inequality ( [ eq : step1 ] ) follows easily .
_ we define @xmath483 where @xmath484 is defined in ( [ eq : deflambda ] ) ( note that , by ( [ eq : deflambda ] ) and by the result of step 1 , the inverse on the r.h.s . of ( [ eq : fesh ] )
is well defined , if @xmath39 and @xmath485 are small enough ) .
then we have @xmath486 the proof of ( [ eq : step2 ] ) relies on the isospectrality of the feshbach map and can be found , for example , in @xcite .
this inequality says that , instead of finding a bound on the operator @xmath466 restricted to the range of @xmath487 , we can study the operator @xmath488 restricted to the much smaller range of the projection @xmath473 ( which is finite - dimensional ) . using the assumption ( [ eq : deflambda ] ) and eq .
( [ eq : step1 ] ) , we see that , if @xmath39 and @xmath489 are sufficiently small , @xmath490 ( later , when we will choose the parameter @xmath440 and @xmath441 , we will make sure that @xmath491 is small enough , if @xmath39 is small enough ) .
this implies that the operator @xmath488 is bounded from below by @xmath492 next , we study the second term on the r.h.s . of inequality ( [ eq : e ] ) .
_ there exists a constant @xmath477 independent of @xmath493 and @xmath441 such that @xmath494 in order to prove this bound , we note that , for any @xmath495 , @xmath496 furthermore @xmath497 p_j \\ = & - g \overline{p}_j \ , \phi ( iaf_x ) p_j + g\theta \overline{p}_j \ , ( h_g ( \pi ) - e_j ( \pi ) ) \overline{p}_j \ , r_{{\varepsilon}}^2 a^ * ( f_x ) p_j \ , , \end{split}\ ] ] because @xmath498 hence we find that @xmath499 using lemma [ lm : phibounds ] and the bound @xmath500 , we conclude that @xmath501 taking the square of this inequality , we obtain ( [ eq : step3 ] ) . _ step 4 .
_ we show that @xmath502 using that @xmath503 and @xmath504 , we easily find that @xmath505 writing @xmath506 , using that @xmath473 commutes with @xmath410 , and that @xmath507 we find @xmath508 eq . ( [ eq : step4 ] ) now follows writing @xmath436 and using that @xmath509 . from step 3 and step 4 and from ( [ eq : e ] ) we get @xmath510 for a constant @xmath59 independent of @xmath493 and @xmath441 .
_ next , we claim that @xmath511 where @xmath512 , as @xmath439 . in order to prove
, we use the pull - through formula for @xmath513 and the fact that @xmath514 .
this yields @xmath515 next , we write @xmath516 , where @xmath517 is the orthogonal projection onto the eigenspace of @xmath80 corresponding to the eigenvalue @xmath395 .
then we obtain that @xmath518 all operators involved in the expression on the r.h.s . of ( [ eq : par ] )
act trivially on the fock space , and we get the lower bound @xmath519 using that @xmath520 , as @xmath439 , and recalling the definition of the matrix @xmath421 , we find that @xmath521 this proves eq .
( [ eq : step5 ] ) .
_ proof of eq . .
_ from ( [ eq : e2 ] ) and
( [ eq : step5 ] ) , we derive that @xmath522 choosing @xmath523 and @xmath524 , with @xmath525 , we get @xmath526 for @xmath39 sufficiently small .
note that , with this choice of @xmath441 and @xmath440 , @xmath527 , and thus ( [ eq : conde ] ) is satisfied , if @xmath39 is small enough . from ( [ eq : deflambda ] ) and (
[ eq : step2 ] ) we then get that @xmath528 which proves ( [ eq : pos2 ] ) .
_ proof of eq . .
_ to prove ( [ eq : pos1 ] ) , and hence complete the proof of the proposition , we must replace @xmath487 by the spectral projection @xmath529 of the full hamiltonian @xmath76 .
given an interval @xmath530 with @xmath531 and @xmath532 we choose an interval @xmath533 such that @xmath534 , and @xmath535 and with @xmath536 .
furthermore , we choose a function @xmath216 with the property that @xmath537 on @xmath179 and @xmath538 on @xmath539 .
we can assume that @xmath540 .
applying ( [ eq : h0 ] ) with @xmath179 replaced by @xmath541 , and multiplying the resulting inequality with @xmath542 we get @xmath543 setting @xmath544 and @xmath545 , we have that @xmath546 using @xmath547 we find that @xmath548 next we use that @xmath549 from the definition of @xmath466 it follows that @xmath550 with the choice @xmath551 , @xmath524 .
this implies that @xmath552 where the constant @xmath59 depends on @xmath326 ( @xmath59 is proportional to @xmath553 ) .
thus , with ( [ eq : chibchi ] ) , @xmath554 since @xmath555 , we have @xmath556 and @xmath557 .
therefore , multiplying from the left and the right with @xmath558 , and choosing @xmath39 small enough , we find that @xmath559 for some positive constant @xmath59 , which , with ( [ eq : commb ] ) , completes the proof of the proposition . proposition [ prop : poscomm1 ] and proposition [ prop : virial ] , below , prove absence of embedded eigenvalues of @xmath71 on @xmath560 with the exception of a small interval around the ground state energy , @xmath561 .
absence of embedded eigenvalues near the ground state energy follows from our next proposition .
recall from ( [ eq : a ] ) the notation @xmath562 with @xmath462 .
[ prop : poscomm2 ] assume hypotheses ( h0)-(h1 ) .
fix @xmath134 , and choose @xmath563 , with @xmath564 .
suppose the interval @xmath565 is such that @xmath566 ( recall that @xmath567 denotes the first excited eigenvalue of the free hamiltonian @xmath410 ) .
then , if @xmath196 is sufficiently small ( depending on @xmath81 , @xmath114 and @xmath448 ) , there exists @xmath568 such that @xmath569 e_{\delta } ( h_g(\pi ) ) \\
\geq ( 1- \beta ) e_{\delta } ( h_g(\pi ) ) \left(1 - p_{{\varphi}_0 \otimes \omega } \right ) e_{\delta } ( h_g(\pi ) ) - c g e_{\delta } ( h_g(\pi))\ , .\end{gathered}\ ] ] a simple calculation shows that @xmath570 = n -\frac{\pi - { \mathrm{d}\gamma}(k)}{m } \cdot { \mathrm{d}\gamma}(\hat k ) + g \phi ( ia f_x ) \ , .\ ] ] hence @xmath571 e_{\delta } ( h_g(\pi ) ) \\ &
\geq ( 1 -\beta ) e_{\delta } ( h_g(\pi ) ) n e_{\delta } ( h_g(\pi ) ) + g e_{\delta } ( h_g(\pi ) ) \phi ( ia f_x ) e_{\delta } ( h_g(\pi ) ) \\ & \geq ( 1 - \beta - c g ) e_{\delta } ( h_g(\pi ) ) - ( 1- \beta ) e_{\delta } ( h_g(\pi ) ) p_{\omega } e_{\delta } ( h_g(\pi ) ) \ , . \end{split}\ ] ] here we use that , by hypothesis ( h1 ) , @xmath572 , @xmath573 .
next , we note that @xmath574 where @xmath320 is the unique ( up to a phase ) ground state vector of the atomic hamiltonian @xmath80 .
next we choose a function @xmath575 with @xmath217 for @xmath576 and @xmath219 for @xmath577 .
( [ eq : pomega ] ) then implies that @xmath578 note that @xmath579 since , by definition of the interval @xmath179 , @xmath580 , we find that @xmath581 with ( [ eq : com ] ) and ( [ eq : pomega2 ] ) , this shows that @xmath582 e_{\delta } ( h_g(\pi ) ) \\ &
\geq ( 1-\beta ) e_{\delta } ( h_g(\pi ) ) \left(1 - p_{{\varphi}_0 \otimes \omega}\right ) e_{\delta } ( h_g(\pi ) ) - c g e_{\delta } ( h_g(\pi ) ) \ , . \end{split}\ ] ] [ prop : virial ] let hypotheses ( h0)-(h1 ) be satisfied .
assume that @xmath583 , for some @xmath584 with @xmath585 , and for an energy @xmath586
. then @xmath587 { \varphi}\rangle$ } } = 0 \quad \text{and } \quad { \mbox{$\langle { \varphi},[h_g(\pi ) , ia ] { \varphi}\rangle$ } } = 0,\ ] ] where @xmath588 is as ( [ eq : a ] ) , and @xmath589 with @xmath590 defined in ( [ eq : d ] ) . to prove the proposition
, we replace the hamiltonian @xmath76 by a modified hamiltonian @xmath591 where the new dispersion law @xmath592 satisfies @xmath343 , for @xmath593 , and @xmath594 for all @xmath6 .
since the two hamiltonian @xmath76 and @xmath347 act identically on states without soft bosons ( in the range of the projection @xmath239 ) , it is enough to prove that @xmath595{\varphi}\rangle$ } } = 0\ ] ] and @xmath596{\varphi}\rangle$ } } = 0\ ] ] where @xmath597 , and @xmath598 equals @xmath428 with @xmath588 replaced by @xmath599 .
the proof of ( [ eq : vir1 ] ) is very similar to the proof of lemma 40 , in @xcite .
the proof of ( [ eq : vir2 ] ) follows easily from ( [ eq : vir1 ] ) , because @xmath600 is a bounded operator on @xmath601 .
[ cor : spectrum ] assume hypotheses ( h0)-(h2 ) .
fix @xmath134 and choose @xmath602 , with @xmath336 and @xmath56 as in lemma [ lm : sigma ] . then , for sufficiently small values of @xmath603 , @xmath604 where @xmath273 is a simple eigenvalue , for all @xmath67 with @xmath338 .
_ how small @xmath39 has to be chosen depends on the choice of @xmath81 ( @xmath605 ) , on the choice of @xmath114 ( we need ( [ eq : sigma1 ] ) to hold true ) , on @xmath453 ( @xmath606 ) , and it also depends on the distances between the eigenvalues of the atomic hamiltonian ( we must require that latexmath:[$g^{1/2 } \ll \min \ { @xmath185 is such that @xmath608 ) . we first prove that if @xmath39 is sufficiently small , then @xmath609 and that @xmath273 is a simple eigenvalue of @xmath610 . to this end , we define @xmath611 .
we define intervals @xmath612 for @xmath613 , where @xmath185 is such that @xmath614 .
each interval @xmath615 contains exactly one eigenvalue of the free hamiltonian @xmath410 , and @xmath616 . the absence of eigenvalues of @xmath617 inside @xmath615 , for @xmath408 and for @xmath39 sufficiently small , follows from propositions [ prop : poscomm1 ] and [ prop : virial ] .
next , suppose that @xmath351 is a normalized eigenvector of @xmath76 corresponding to an eigenvalue @xmath618 . without loss of generality , we can assume that @xmath619 is real ; recall that @xmath320 is the unique ( up to a phase ) normalized ground state vector of @xmath620 .
then , by proposition [ prop : poscomm2 ] and proposition [ prop : virial ] , we have that @xmath621 hence @xmath622 if there were two orthogonal eigenvectors of @xmath610 , @xmath623 and @xmath624 , corresponding to eigenvalues in @xmath625 then both would satisfy inequality ( [ eq : normpsi ] ) , and , thus , we would conclude that @xmath626 but this is impossible if @xmath627 .
so , for @xmath39 small enough , there can only be one eigenvector of @xmath610 corresponding to an eigenvalue in @xmath625 . in proposition
[ prop : gs ] , we have proven that @xmath610 has a ground state vector .
this proves the fact that @xmath273 is a simple eigenvalue of @xmath610 as well as the fact that @xmath617 has no other eigenvalue in @xmath625 .
hence ( [ eq : cor1 ] ) follows . to complete the proof of the corollary
, we need to show that @xmath628 to this end , we decompose @xmath629 , and we write @xmath630 , where @xmath631 is the space of vectors containing exactly @xmath185 soft , non - interacting , bosons .
the hamiltonian leaves each @xmath632 invariant , and the restriction of @xmath76 on @xmath632 is given by @xmath633 here @xmath634 is an operator over @xmath635 , the space of states with no soft bosons .
we know that the only eigenvalue of @xmath636 in @xmath560 is its ground state energy @xmath637 as long as @xmath638 .
in particular the only eigenvalue of @xmath639 in @xmath560 is given by @xmath640 if this number is less than @xmath114 .
thus @xmath641 is an eigenvalue of @xmath642 if and only if there exists a set @xmath643with positive measure , such that @xmath644 for all @xmath645 .
but this is impossible because @xmath646 for every @xmath67 with @xmath647 and @xmath39 small enough .
the proofs of most of the results in this section are similar to those of the corresponding results in @xcite . in order to give an idea of the structure of the proof of our main result ( asymptotic completeness , theorem [ thm : acph ] )
, we repeat here the most important theorems , but we omit most of their proofs ( we refer to the corresponding statements in @xcite ) . the main difference with respect to @xcite
is encountered in the proof of the positivity of the asymptotic observable in section [ sec : pos ] : there , we propose some new ideas to control the internal degrees of freedom of the atom ( which are not present in @xcite , because there we considered free electrons coupled to the quantized radiation field ) .
the first step towards understanding scattering theory for the model studied in this paper consists in the construction of states with asymptotically free photons .
this can be accomplished using asymptotic field operators , which are constructed in the next theorem .
note that in theorems [ thm : asy_fields ] and [ thm : omega_+ ] we do not impose any infrared cutoff on the interaction ; we can take @xmath459 , provided the form factor @xmath648 is smooth at @xmath93 .
we use the notation @xmath649 .
[ thm : asy_fields ] assume hypotheses ( h0)-(h1 ) are satisfied ( but @xmath459 is allowed ! ) .
fix @xmath134 and choose @xmath563 ( with @xmath336 as in lemma [ lm : sigma ] ) .
if @xmath196 is so small that ( [ eq : sigma1 ] ) is true , then the following results hold . 1 .
let @xmath650 and let @xmath651 .
then the limit @xmath652 exists for all @xmath653 .
2 . let @xmath654
. then @xmath655 = ( g , h ) \quad \text{and } \quad [ a^{\sharp}_+ ( g ) , a^{\sharp}_+ ( h ) ] = 0,\ ] ] in the sense of quadratic forms on @xmath158 ( @xmath656 means either @xmath657 or @xmath658 ) .
3 . let @xmath659 , and let @xmath660 if @xmath661 .
4 . let @xmath662 for @xmath663 .
put @xmath664 and assume @xmath665 . then
if @xmath666 we have that @xmath667 , the limits @xmath668 exist , and @xmath669 5 . if @xmath670 and @xmath671 , @xmath672 ( wave packets of dressed atom states are vacua of the asymptotic field operators ) .
the proof of this theorem is very similar to the one of theorem 13 and lemma 14 in @xcite .
it relies on a propagation estimate for the center of mass of the atom ( see proposition 12 in @xcite ) , which guarantees that if the energy is smaller than @xmath336 , then the asymptotic velocity of the atom is bounded above by @xmath81 ( here @xmath113 ) , and it exploits the fact that , because the energy is below the ionization threshold , the electron is exponentially bound to the nucleus .
these two facts and the fact that the propagation speed of photons is the speed of light are sufficient to prove that the interaction between the atom and asymptotically freely propagating photons tends to zero , as @xmath673 . the existence of asymptotic field operators allows us to introduce the wave operator @xmath154 of the system . in order to define @xmath154 , we add a new copy of the fock space @xmath4 describing states of free photons to the physical hilbert space @xmath674 .
we define the extended hamiltonian @xmath675 on the extended hilbert space @xmath676 . in the next theorem
, we establish the existence of the wave operator @xmath154 as an isometry from a subspace of @xmath677 to a subspace of the physical hilbert space @xmath52 .
the `` scattering identification map '' , @xmath678 , used in the definition of the wave operator @xmath154 , is defined in appendix [ sec : scattiden ] .
[ thm : omega_+ ] let hypotheses ( h0)-(h1 ) be satisfied ( but @xmath459 is allowed ) .
fix @xmath134 , and choose @xmath679 ( with @xmath680 defined as in lemma [ lm : sigma ] )
. then if @xmath196 is small enough ( depending on @xmath81 and @xmath114 ) the limit @xmath681 exists for an arbitrary vector @xmath109 in the dense subspace of @xmath682 spanned by finite linear combinations of vectors of the form @xmath683 , where @xmath684 , @xmath685 , and with @xmath686 .
if @xmath687 belongs to this space then @xmath688 furthermore @xmath689 , and @xmath154 has therefore a unique extension , also denoted by @xmath154 , to @xmath690 . on @xmath691 ,
the operator @xmath154 is isometric . for all @xmath692 , @xmath693 for the proof of this theorem
we refer to the proof of theorem 15 in @xcite , which is almost identical . from eq .
( [ eq : omega_+2 ] ) we see that vectors in the range of @xmath154 are limits of linear combination of vectors describing wave packet of dressed atom states and configurations of finitely many asymptotically freely moving photons .
physically , it is expected that the asymptotic evolution of every state with an energy below the ionization threshold of the atom ( that is @xmath694 ) and so small that the atom does not propagate with a velocity larger than one ( i.e. , @xmath695 ) can be approximated by linear combinations of vectors describing a dressed atom state and a configuration of finitely many freely propagating photons .
more precisely , one expects that @xmath696 this statement is called _
asymptotic completeness of rayleigh scattering_. due to technical difficulties , we can only prove asymptotic completeness for states with energy less than a threshold energy @xmath697 and assuming that the coupling constant @xmath39 is small enough .
the following theorem is our main result .
[ thm : acph ] assume that hypotheses ( h0)-(h2 ) are satisfied ( see eqs .
( [ eq : hyph0 ] ) , ( [ eq : gx ] ) , and ( [ eq : h2 ] ) ) . fix @xmath133 , and
choose @xmath698 ( with @xmath336 and @xmath56 as in lemma [ lm : sigma ] ) .
then , for @xmath699 sufficiently small , @xmath700 _ remark .
the allowed range of values of @xmath39 depends on the value of @xmath701 ; ( we need that @xmath702 ) , on the choice of @xmath114 ( @xmath39 must be small enough in order for eq .
( [ eq : sigma1 ] ) to hold true ) , on the value of @xmath703 ( @xmath704 ) , and on @xmath705 , with @xmath185 such that @xmath706 ( @xmath707 ) . as remarked in section [ sec : model ] , the assumption that @xmath39 is positive is not necessary , it only simplifies the notation ( but @xmath708 is not allowed , because in this case the fiber hamiltonian @xmath71 has embedded eigenvalues ) .
theorem [ thm : acph ] will be seen to follow from lemma [ lm : ac ] , where we show that it suffices to prove an analogous statement for a modified hamiltonian @xmath160 ( introduced in the next section ) and from theorem [ thm : ac ] in section [ sec : pos ] , where asymptotic completeness for @xmath160 is proven .
the fact that the bosons are massless leads to some technical difficulties connected with the unboundedness of the operator @xmath709 with respect to the hamiltonian .
however , as long as the infrared cutoff is strictly positive , the number of bosons with energy below @xmath212 is conserved .
this allows us to introduce a modified hamiltonian , where the dispersion law of the soft , non - interacting , photons is changed .
we define @xmath710 and we assume that the dispersion law @xmath369 has the following properties . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * hypothesis ( h3 ) .
* @xmath711 , with @xmath712 , @xmath713 , for @xmath714 , @xmath715 , for all @xmath416 , @xmath716 , and @xmath717 unless @xmath93 .
furthermore , @xmath718 for all @xmath719 .
( here @xmath210 is the infrared cutoff defined in hypothesis ( h1 ) . )
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the two hamiltonians , @xmath51 and @xmath160 , agree on states of the system without soft bosons . recall that @xmath720 is the characteristic function of the set @xmath721 and that the operator @xmath239 is the orthogonal projection onto the subspace of vectors describing states without soft bosons .
it is straightforward to check that @xmath51 and @xmath160 leave the range of the projection @xmath239 invariant and that @xmath722 the same conclusion can be reached using the unitary operator @xmath723 introduced in section [ sec : model ] . on the factorized hilbert space @xmath724 , the hamiltonians @xmath51 and @xmath160 are given by @xmath725 and we see explicitly that the two hamiltonians agree on states without soft bosons .
the modified hamiltonian @xmath160 , just like the physical hamiltonian @xmath51 , commutes with spatial translations , i.e. , @xmath726 = 0 $ ] , where @xmath727 is the total momentum of the system . in the representation of the system on the hilbert space @xmath728 , the modified hamiltonian @xmath160 is given by @xmath729 where @xmath730 has been defined in section [ sec : model ] .
the fiber hamiltonians @xmath71 and @xmath347 commute with the projection @xmath239 and agree on its range , @xmath731 in the proof of proposition [ prop : gs ] we have shown that if @xmath134 and @xmath732 then , for @xmath39 small enough , @xmath733 where @xmath734 is a simple eigenvalue of @xmath71 and of @xmath347 , as long as @xmath348 .
moreover , the corresponding dressed atom states coincide .
since the subspace @xmath99 is defined in terms of the dressed atom states @xmath98 , it follows that vectors in @xmath99 also describe dressed atom wave packets for the dynamics generated by the modified hamiltonian @xmath160 .
we remark that @xmath735 for all @xmath304 with @xmath338 , and for @xmath39 sufficiently small ; see eq . and corollary [ cor : spectrum ] .
next , we discuss the scattering theory for the modified hamiltonian @xmath160 . as in theorem [ thm : omega_+ ]
we fix @xmath113 and we choose @xmath736 .
then , by the assumption that @xmath737 for @xmath738 , and since @xmath739 commutes with @xmath51 and @xmath160 we have that @xmath740 for all @xmath104 .
it follows that the limit @xmath741 exists and that @xmath742 , for all @xmath743 and for all @xmath659 .
this and the fact that vectors in @xmath99 describe dressed atom states for @xmath51 and for @xmath160 show that the asymptotic states constructed with the help of the hamiltonians @xmath51 and @xmath160 coincide . on the extended hilbert space @xmath744 , we define the extended modified hamiltonian @xmath745 in terms of @xmath160 and @xmath746 we also define a modified version , @xmath747 , of the wave operator @xmath154 introduced in section [ sec : waveop ] .
[ lm : tilde - omega - mod ] let hypotheses ( h0 ) , ( h1 ) and ( h3 ) be satisfied ( @xmath459 in hypothesis ( h1 ) is allowed , and then @xmath748 ) . fix @xmath442 and @xmath749 .
then if @xmath196 is sufficiently small , depending on @xmath81 and @xmath114 , the limit @xmath750 exists for all @xmath751 .
moreover , the modified wave operator @xmath752 , defined by @xmath753 , agrees with @xmath147 on @xmath754 .
that is , @xmath755 for all @xmath756 .
we now extend the domain of @xmath147 to include arbitrarily many soft , non - interacting bosons . as a byproduct
we obtain a proof of . to start with , we recall the isomorphism @xmath757 introduced in section [ sec : model ] and define a unitary isomorphism @xmath758 separating interacting from soft bosons in the extended hilbert space @xmath759 . with respect to this factorization
the extended hamiltonian @xmath760 becomes @xmath761 , where @xmath762 . as an operator from @xmath763 to @xmath252 ,
the wave operator @xmath154 acts as @xmath764 where @xmath765 is given by @xmath766 while @xmath767 is given by @xmath768 where @xmath256 is the orthogonal projection onto the vacuum vector @xmath769 . in view of and
, the domain of @xmath147 can obviously be extended to @xmath770 . for the modified wave operator @xmath771
, we have @xmath772 , and from @xmath773 it follows that @xmath774 .
consequently , also @xmath752 is well defined on @xmath775 and @xmath776 .
we summarize the main conclusions in a lemma .
[ lm : ac ] let the assumptions of lemma [ lm : tilde - omega - mod ] be satisfied , and let @xmath147 be defined on @xmath777 , as explained above
. then @xmath778 in the factorization @xmath779 .
in particular , the following statements are equivalent : 1 .
2 . @xmath781 .
3 . @xmath782 .
4 . @xmath783 .
fix @xmath784 and choose @xmath785 ( recall from lemma [ lm : sigma ] that @xmath786 ) .
we choose numbers @xmath787 and @xmath788 such that @xmath789 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * definition .
* we pick a function @xmath790)$ ] such that @xmath791 on @xmath792 and @xmath793 on @xmath794 $ ] . our _ asymptotic observable _
@xmath159 is defined by @xmath795 where the energy cutoff @xmath26 is smooth and supported in @xmath796 . for the existence of this limit ,
see proposition [ prop : w ] below .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the physical meaning of the asymptotic observable is easy to understand : @xmath159 measures the number of photons that are propagating with an asymptotic velocity larger than @xmath788
. we will prove in section [ sec : pos ] that @xmath159 is _ positive _ when restricted to the subspace of vectors orthogonal to the space @xmath99 of wave packets of dressed atom states . instead of inverting the wave operator @xmath154 directly , we can then invert it with respect to the asymptotic observable @xmath159 .
more precisely , we define an operator @xmath797 , called the _ inverse wave operator _ , such that @xmath798 .
then , using the positivity of @xmath159 , we can construct an inverse of @xmath799 . in order to define @xmath800
, we need to split each boson state into two parts , the second part being mapped to the second fock - space of prospective asymptotically freely moving bosons . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * definition . *
we define @xmath801 as follows : let @xmath802 , where @xmath803 , @xmath804)$ ] , @xmath805 , @xmath806 on @xmath807 $ ] , @xmath808 $ ] while @xmath809 on @xmath810 and @xmath811
. then the inverse wave operator @xmath800 is given by @xmath812 where @xmath26 is a smooth energy cutoff supported in @xmath796 .
see appendix [ sec : factfock ] for the definition of the operator @xmath813 . for the existence of this limit , see proposition [ prop : w+ ] .
_ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ note that , since by definition @xmath814 , the photons which propagate with velocity larger than @xmath788 are asymptotically free . to prove this fact , notice first that lemma [ lm : sigma ] continues to hold with @xmath51 replaced by @xmath160 . hence , the assumption that @xmath815 ( where @xmath26 is the energy cutoff appearing in the definition of @xmath159 and @xmath800 ) with @xmath335 guarantees , for sufficiently small @xmath39 , that both the nucleus and the electron remain inside a ball of radius @xmath816 around the origin . in fact , using the assumption @xmath131 ( and @xmath39 small enough ) , we can prove , analogously to proposition 12 in @xcite , that @xmath817 for any @xmath818 with @xmath819 , @xmath820 and for any @xmath821 with @xmath815 . recall that @xmath44 is the coordinate of the center of mass of the atom . moreover , the assumption that @xmath822 implies that the electron and the nucleus remain exponentially bound for all times ; therefore , both the electron and the nucleus are localized inside the ball of radius @xmath816 . as a consequence , the interaction strength between the nucleus ( or the electron ) and those bosons counted by @xmath823 decays in @xmath104 at an integrable rate . to establish this fact rigorously we need the following lemma , similar to lemma 9 in @xcite .
assume that hypothesis ( h1 ) is satisfied and that @xmath824
. then , for every @xmath825 , there exists a constant @xmath826 such that @xmath827 moreover , if @xmath828 , we have @xmath829 _ remark . _ in the proof of the existence of the operators @xmath159 and @xmath800 , where we use this lemma , typically @xmath830 and @xmath831 . hence the r.h.s . of ( [ eq : dec2 ] )
gives a decay in time which is integrable if we choose @xmath832 large enough . to prove ( [ eq : decay ] ) , we first choose @xmath833 , with @xmath834 ( recall that @xmath835 and @xmath232 ) and we observe that @xmath836 using that @xmath837 , it follows that @xmath838 ) can be bounded by @xmath839 ( because @xmath840 is bounded )
. moreover the square of the first term on the r.h.s . of ( [ eq : decay1 ] ) can be estimated by @xmath841 for all @xmath44 and @xmath45 .
here we used that , from @xmath842 , @xmath843 , and since , by definition of @xmath441 , @xmath844 and @xmath845 , we have that @xmath846 . since @xmath847 ,
their fourier transform decay faster than any power , and hence ( [ eq : declast ] ) implies ( [ eq : decay ] ) .
to prove ( [ eq : dec2 ] ) , we use that @xmath848 because @xmath849 is finite ( because @xmath828 and by a simple commutation ) .
( [ eq : dec2 ] ) then follows from ( [ eq : decay ] ) .
the decay of the interaction determined in the last lemma is one of the two key ingredients for proving the existence of @xmath159 and @xmath800 .
the other one is a propagation estimate for the photons , analogous to proposition 24 in @xcite , but with the cutoff for @xmath850 ( in @xcite , @xmath45 is the position of the electron ) replaced by a cutoff for the asymptotic velocity @xmath851 of the center of mass of the nucleus - electron compound ( the reason why we can introduce here a cutoff in @xmath851 is that , because of ( [ eq : pex ] ) , we know it can not exceed @xmath81 ) . for the details of the proof of the next two proposition we refer to theorems 26 and 28 in @xcite .
[ prop : w ] assume that hypotheses ( h0 ) , ( h1 ) and ( h3 ) are satisfied .
fix @xmath784 , and choose @xmath443 .
suppose that @xmath852 with @xmath853 .
let @xmath788 , and @xmath854 be as defined above , and let @xmath855 be the operator of multiplication with @xmath856 .
then , for @xmath196 small enough ( in order for ( [ eq : sigma1 ] ) to hold true ) , @xmath857 exists , @xmath858 and @xmath159 commutes with @xmath160 . here
@xmath859 .
[ prop : w+ ] assume hypotheses ( h0 ) , ( h1 ) and ( h3 ) are satisfied . fix @xmath860 and choose @xmath698 .
suppose that @xmath852 with @xmath861 , and that @xmath854 and @xmath862 are defined as described above . if @xmath196 is so small that ( [ eq : sigma1 ] ) holds , then * the limit @xmath863 exists , and @xmath864 , for all @xmath865 ; * @xmath866 ; * @xmath867 . in this section
we prove the positivity of the asymptotic observable @xmath159 , restricted to the subspace of states orthogonal to wave packets of dressed atom states and not containing any soft bosons .
we need the following lemma .
[ lm : poscomm3 ] assume hypotheses ( h0 ) , ( h1 ) , ( h3 ) .
fix @xmath784 and choose @xmath868 .
suppose , moreover , that @xmath852 and @xmath853 .
put @xmath869 , where @xmath44 is the position of the center of mass of the atom . then , if @xmath39 is so small that ( [ eq : sigma1 ] ) holds true , we have that @xmath870f({h_{\text{mod } } } ) \geq ( 1-\beta ) f({h_{\text{mod } } } ) n f({h_{\text{mod } } } ) - c g f({h_{\text{mod}}})^2\ ] ] on the range of the projection @xmath239 .
this lemma follows from a straightforward estimate of the commutator @xmath871 $ ] , from ( [ eq : sigma1 ] ) , ( [ eq : sigma2 ] ) , and from lemma [ lm : phibounds ] .
[ thm : wpos ] assume that hypotheses ( h0)-(h3 ) are satisfied .
fix @xmath132 and choose @xmath872 ( with @xmath336 and @xmath56 as in lemma [ lm : sigma ] ) .
let the operator @xmath159 be defined as in proposition [ prop : w ] with @xmath873 .
then if @xmath603 is sufficiently small we can choose @xmath874 in the definition of @xmath159 such that @xmath875 here @xmath59 is a positive constant depending on @xmath39 , @xmath81 , @xmath114 , but independent of @xmath109 .
in particular , if @xmath565 and then @xmath876 on @xmath179 , then @xmath877 let @xmath878 . since @xmath879 is dense in @xmath880 , it is enough to prove that @xmath881 for every @xmath882 .
as before , we use the notation @xmath883 .
the first step consists in proving that there exists a constant @xmath59 , depending only on @xmath114 , such that , for every @xmath884 and for every @xmath169 , @xmath885 ^ 2 \\ & - c g \|f{\varphi}\|^2 + o(1),\qquad \text{as } \quad t \to\infty\ , .
\end{split}\ ] ] this inequality can be established as in the proof of theorem 27 in @xcite .
next , we observe that @xmath886 where @xmath256 denotes the orthogonal projection onto the fock vacuum @xmath144 .
the second term on the r.h.s . of the last equation can be written as an integral over fibers with fixed total momentum . making use of the fact that @xmath887 and of fubini s theorem to interchange the integration over @xmath888 and over @xmath67 , we obtain @xmath889 where @xmath890 is the orthogonal projection onto the dressed atom state @xmath98 , and @xmath891 is its orthogonal complement . for every fixed @xmath67 ,
the operator @xmath892 is a compact operator on @xmath893 , because @xmath894 ; ( since @xmath828 , this follows from lemma [ lm : sigma ] ) . by the continuity of the spectrum of @xmath347 on @xmath895 ( see corollary [ cor : spectrum ] ) , and the rage theorem ( see , for example , @xcite ) , it follows that @xmath896 as @xmath673 , pointwise in @xmath67 . using lebesgue s
dominated convergence theorem , we conclude that @xmath897 as @xmath673 . from ( [ eq : fnf ] )
we obtain @xmath898 for @xmath104 large enough , where we can assume @xmath899 without loss of generality .
( [ eq : fpf ] ) and ( [ eq : fnf2 ] ) allow us to apply lemma [ lm : teilfolge ] , with @xmath900 and @xmath901 ( it is easy to check that @xmath902 and @xmath903 are bounded and continuous ) .
we conclude that there exists a sequence @xmath904 with @xmath905 , as @xmath906 , such that @xmath907 from ( [ eq : step1 + 2 ] ) we infer that @xmath908 as @xmath906 . choosing @xmath909 and @xmath169 sufficiently small , we conclude that @xmath910 as @xmath906 . hence ,
by ( [ eq : teilfolge ] ) , there are constants @xmath911 and @xmath912 , depending only on @xmath114 , such that @xmath913 as @xmath906 . if @xmath132 and @xmath39 is small enough , we arrive at ( [ eq : claim ] ) by taking the limit @xmath906 .
( since we already know that the limit defining @xmath159 exists , it is enough to prove its positivity on some arbitrary subsequence ! ) [ lm : teilfolge ] suppose @xmath902 and @xmath903 are positive , continuous , bounded functions on @xmath914 , such that @xmath915 for all @xmath916 large enough , and @xmath917 then , for every @xmath321 , there exists a sequence @xmath918 , with @xmath919 , as @xmath906 , such that @xmath920 and @xmath921 define the sets @xmath922 : m_1 ( t ) < \frac{1}{1 + \delta } h_1 ( t ) \ } \ , , \quad \quad \text{for some } \delta>0.\ ] ] by the continuity of @xmath923 and @xmath924 the sets @xmath925 are measurable ( with respect to lebesgue measure on @xmath914 ) , for all @xmath926 .
denote by @xmath927 the lebesgue measure of a measurable set @xmath928 .
we show that @xmath929 in fact , if ( [ eq : wrong ] ) were false , then ( since @xmath930 for all @xmath931 ) @xmath932 and hence , for arbitrary @xmath169 , we could find a @xmath933 such that @xmath934 for all @xmath935 .
this would imply that @xmath936 where @xmath937 denotes the supremum of the bounded function @xmath938 and @xmath939 \backslash s_t$ ] is the complement of @xmath925 inside @xmath940 $ ] .
hence , we find @xmath941 for every @xmath942 . put @xmath943
. then we have @xmath944 by the assumption that @xmath945 for all @xmath926 large enough , we have @xmath946 , and thus , choosing @xmath947 , we find @xmath948 for all @xmath926 large enough .
this contradicts the boundedness of @xmath949 ( which follows from the boundedness of @xmath950 ) .
this proves ( [ eq : wrong ] ) , and implies that there exist @xmath169 and a sequence @xmath951 converging to infinity such that @xmath952 for all @xmath953 .
hence @xmath954 , for all @xmath955 .
next , we show that there exists a sequence @xmath956 , with @xmath919 as @xmath906 , such that @xmath957 , for all @xmath958 , and @xmath959 as @xmath906 . since , for all @xmath958 , @xmath960 , for some @xmath961 , the sequence @xmath962 automatically satisfies ( [ eq : cond1 ] ) .
thus the lemma follows if we can prove ( [ eq : cond2 ] ) . to this end
we argue again by contradiction .
if there were no sequence @xmath963 satisfying ( [ eq : cond2 ] ) then there would exist @xmath964 and @xmath965 such that @xmath966 , for all @xmath967 .
but then , for an arbitrary @xmath961 with @xmath968 , @xmath969 } { { \rm d}}s \ , h_2 ( s ) \geq \alpha \frac{\mu ( s_{t_m}^c \cap [ \tau , t_m])}{t_m } \\ & \geq \alpha \frac{\mu ( s_{t_m}^c)}{t_m } - \frac{\alpha \tau}{t_m } \geq \alpha { \varepsilon}- \frac{\alpha \tau}{t_m } \end{split}\ ] ] for all @xmath961 with @xmath968 .
taking @xmath970 , this contradicts the assumption ( [ eq : ass2 ] ) . using the positivity of the asymptotic observable @xmath159
, we can complete the proof of asymptotic completeness for the hamiltonian @xmath160 .
our proof is based on induction in the energy .
the following simple lemma is useful .
[ lm : induction ] assume that hypotheses ( h0)-(h3 ) are satisfied .
fix @xmath134 and choose @xmath868 .
the wave operators @xmath971 and @xmath154 are defined as in lemma [ lm : tilde - omega - mod ] and in theorem [ thm : omega_+ ] , respectively .
suppose that @xmath972 , for some @xmath973 .
then , for every @xmath974 , there exists @xmath975 such that @xmath976 if @xmath977 and @xmath978 then @xmath979 . the interpretation of this lemma is simple : if we know that asymptotic completeness holds for vectors with energy lower than @xmath980 , then it continues to be true if we add asymptotically free photons to these vectors ( no matter what the total energy of the new state is ) . for the proof of this lemma
we refer to lemma 20 of @xcite . using this lemma
we can prove asymptotic completeness for @xmath160 ; the proof is similar to the proof of theorem [ thm : ac ] in @xcite .
we repeat it here , because it is very short , and because it explains the ideas behind all the tools introduced in section [ sec : scatt ] .
[ thm : ac ] assume that hypotheses ( h0)-(h3 ) are satisfied .
fix @xmath981 and choose @xmath982 ; ( with @xmath336 and @xmath56 defined as in lemma [ lm : sigma ] ) .
if @xmath699 is sufficiently small , then @xmath983 the proof is by induction in energy steps of size @xmath984 .
we show that @xmath985 holds for @xmath93 , by proving this claim for all @xmath986 . since @xmath160 is bounded below ,
is obviously correct for @xmath6 large enough . assuming that holds for @xmath987
, we now prove it for @xmath988 .
since @xmath989 is closed ( by theorem [ thm : omega_+ ] ) and since @xmath990 , it suffices to prove that @xmath991 for @xmath992 .
here we use lemma [ lm : ac ] . fix @xmath993 with @xmath994 and
choose @xmath852 real - valued , with @xmath995 on @xmath179 and @xmath996 .
we define the asymptotic observable @xmath159 in terms of @xmath26 , as in proposition [ prop : w ] . by theorem [ thm : wpos ] , the operator @xmath997 is strictly positive on @xmath998 , and hence onto , if @xmath39 is small enough . given @xmath351 in this space , we can therefore find a vector @xmath999 such that @xmath1000 by proposition [ prop : w+ ] , @xmath1001 and @xmath1002 .
furthermore , by part ( ii ) of proposition [ prop : w+ ] , @xmath1003 has at least one boson in the outer fock space , and thus an energy of at most @xmath1004 in the inner one .
that is , @xmath1005w_{+}{\varphi}.\ ] ] hence we can use the induction hypothesis @xmath1006 . by lemma [ lm : induction ] , it follows that @xmath1007 for some @xmath1008 .
we conclude that @xmath1009 where @xmath1010 is the projection onto the orthogonal complement of the vacuum .
this proves the theorem .
let @xmath1011 be a complex hilbert space , and let @xmath1012 denote the @xmath185-fold symmetric tensor product of @xmath1011 .
then the bosonic fock space over @xmath1011 , @xmath1013 is the space of sequences @xmath1014 , with @xmath1015 , @xmath1016 , and with the scalar product given by @xmath1017 where @xmath1018 denotes the inner product in @xmath1019 .
the vector @xmath1020 is called the vacuum . by @xmath1021
we denote the dense subspace of vectors @xmath109 for which @xmath1022 , for all but finitely many @xmath185 .
the number operator @xmath468 is defined by @xmath1023 .
the creation operator @xmath657 , @xmath1024 , is defined on @xmath1025 by @xmath1026 and extended by linearity to @xmath1027 . here
@xmath128 denotes the orthogonal projection onto the symmetric subspace @xmath1028 .
the annihilation operator @xmath658 is the adjoint of @xmath657 .
creation- and annihilation operators satisfy the canonical commutation relations ( ccr ) @xmath1029 = ( g , h),\hspace{3em } [ a^{\#}(g),a^{\#}(h ) ] = 0.\ ] ] in particular , @xmath1030 = \|h\|^2 $ ] , which implies that the graph norms associated with the closable operators @xmath658 and @xmath1031 are equivalent .
it follows that the closures of @xmath658 and @xmath1031 have the same domain . on this common domain
we define the self - adjoint operator @xmath1032 the creation- and annihilation operators , and thus @xmath1033 , are bounded relative to the square root of the number operator : @xmath1034 more generally , for any @xmath1035 and any integer @xmath185 , @xmath1036 let @xmath1038 and @xmath1039 be two hilbert spaces and let @xmath1040 .
we define @xmath1041 by @xmath1042 in general @xmath1043 is unbounded ; but if @xmath1044 then @xmath1045 . from the definition of @xmath1031
it easily follows that @xmath1046 if @xmath1047 on @xmath1038 then these equations imply that @xmath1048 let @xmath1050 be an operator on @xmath1011 .
then @xmath1051 is defined by @xmath1052 for example @xmath709 . from the definition of @xmath1031
we infer that @xmath1053 = a^*(bh ) \quad [ { \mathrm{d}\gamma}(b),a(h ) ] = -a(b^*h),\ ] ] and , if @xmath1054 , @xmath1055 = \phi(ibh).\ ] ] note that @xmath1056 .
let @xmath1038 and @xmath1039 be two hilbert spaces .
we define a linear operator @xmath1057 by @xmath1058u\hspace{3em}\text{for } h=(h_{(0)},h_{(\infty)})\in{\mathfrak{h}}_1\oplus{\mathfrak{h}}_2 .
\end{split}\ ] ] this defines @xmath1059 on finite linear combinations of vectors of the form @xmath1060 . from the ccrs it follows that @xmath1059 is isometric .
its closure is isometric and onto , hence unitary .
suppose @xmath1061 and @xmath1062 are linear operators on @xmath1011 and @xmath1063 is defined by @xmath1064 .
then @xmath1065 and consequently @xmath1066 .
we define @xmath1067 where @xmath1068 is as defined in sect .
[ sec : a2 ] .
it follows that @xmath1069 which is the identity if @xmath1070 . in this case
@xmath1071{\breve{\gamma}}(j)\\ \label{eq : ugamma - phi}{\breve{\gamma}}(j)\phi(h ) & = [ \phi(j_0 h)\otimes 1 + 1\otimes \phi(j_{\infty}h)]{\breve{\gamma}}(j).\end{aligned}\ ] ] we define the scattering identification @xmath1072 by @xmath1073 and extend it by linearity to @xmath1074 .
( note that this definition is symmetric with respect to the two factors in the tensor product .
) there is a second characterization of @xmath678 which can be useful .
let @xmath1075 be defined by @xmath1076 .
then @xmath1077 , with @xmath1059 as above .
since @xmath1078 , the operator @xmath678 is unbounded , but it can be proved that @xmath1079 is bounded , for any @xmath1080 .
in this section we review standard estimates that are used throughout this paper to bound the interaction .
[ lm : phibounds ] assume hypothesis ( h0)-(h1 ) .
let @xmath1086 with @xmath1087 and choose @xmath1088 .
then there exists @xmath1089 such that @xmath1090 with @xmath1091 as in eq .
( [ eq : fx ] ) . for @xmath1092 , @xmath1093 where @xmath1094 ; ( see eqs .
( [ eq : gxx ] ) , ( [ eq : fx ] ) ) .
note that @xmath1095 eq .
( [ eq : phi1 ] ) follows from lemma [ lm : estim ] , because @xmath1096 is bounded ( see lemma [ lm : sigma ] ) and from hypothesis ( h1 ) .
( [ eq : phi2 ] ) follows from ( [ eq : phi1 ] ) because @xmath1097 | scattering of photons at an atom with a _
dynamical _ nucleus is studied on the subspace of states of the system with a total energy below the threshold for ionization of the atom ( _ rayleigh scattering _ ) .
the kinematics of the electron and the nucleus is chosen to be non - relativistic , and their spins are neglected . in a simplified model of a hydrogen atom or a one - electron ion interacting with the quantized radiation field in which the helicity of photons is neglected and the interactions between photons and the electron and nucleus are turned off at very high photon energies and at photon energies below an _ arbitrarily small _ , but _ fixed _ energy ( infrared cutoff ) , _ asymptotic completeness of rayleigh scattering _ is established rigorously . on the way towards proving this result , it is shown
that , after coupling the electron and the nucleus to the photons , the atom still has a stable ground state , provided its center of mass velocity is smaller than the velocity of light ; but its excited states are turned into resonances .
the proof of asymptotic completeness then follows from extensions of a positive commutator method and of propagation estimates for the atom and the photons developed in previous papers .
the methods developed in this paper can be extended to more realistic models .
it is , however , not known , at present , how to remove the infrared cutoff . |
the _ infrared astronomical satellite ( iras ) _ detected a significant population of galaxies that emit large fractions of their bolometric energy in the mid- and far - infrared ( ir ) bands .
the majority of luminous ir galaxies ( lirgs : @xmath5 @xmath6 @xmath7 @xmath8 ) are powered by massive stars that have recently formed in their central kiloparsec @xcite .
the far - ir luminosities of these galaxies therefore measure the amount of star formation activity @xcite .
local starburst galaxies with far - ir luminosities of @xmath9 @xmath8and normal spiral galaxies with @xmath5 @xmath10 @xmath8also show evidence for active star formation in their central ( @xmath11 kpc ) regions , even though their far - ir luminosities are much lower than those of lirgs @xcite .
the formation of o and b stars affects the physical and chemical state of the interstellar medium through stellar uv radiation and through the effects of stellar winds and supernovae @xcite .
strong near - ir h@xmath0 emission lines have been observed from the central regions of galaxies , especially from starburst galaxies and lirgs .
since h@xmath0 vibration - rotation emission can trace the interaction of the neutral ism with the uv radiation from hot young stars and its interaction with sources of dynamical energy such as supernovae , the h@xmath0 luminosity may also be a good tracer of star formation activity .
the main motivation of this paper is to explore how well we can use this additional , readily observable tracer of the interaction of massive stars with the dense interstellar medium to study massive star formation in the inner parts of galactic disks .
the observational part of this paper emphasizes new , large - beam observations of emission in the vibration - rotation transitions of h@xmath0 in the near - ir from nearby late - type galaxies and low luminosity starbursts . for normal galaxies , observations of near - ir h@xmath0 lines
have been restricted to local bright regions because the average surface brightness of the h@xmath0 emission is low . in a previous study
, we used an instrument optimized for observations of low surface brightness line emission in the near - ir to detect enormously extended ( @xmath12 or @xmath13 pc ) h@xmath0 emission along the plane of the galaxy near the galactic center @xcite .
the integrated and dereddened h@xmath0 @xmath1 luminosity from the extended region ( @xmath14 @xmath8 ) dominates over that from the previously known 5 pc circumnuclear gas ring ( @xmath15 @xmath8 , @xcite ) .
our results for the inner 400 pc of the milky way motivated us to begin a study of h@xmath0 emission from the central kiloparsec of other nearby galaxies where it may be possible to detect extended , low surface brightness h@xmath0 emission using the same large - aperture fabry - perot spectrometer used in the galactic work .
the large - beam observations we present here make it possible to compare for the first time the h@xmath0 emission to the far - ir continuum and far - ir fine - structure line emission at similar angular resolution and over similar angular scales .
the observations also give us measurements of h@xmath0 line strengths from nearby objects with linear scales comparable to those obtained with slit spectrometers for the more distant luminous ir galaxies . with this broad range of spatially compatible observations ,
we investigate whether it is possible to formulate a generalized star formation law which can be applied to galaxies over a wide range of levels in star formation activity . in order to analyze the h@xmath0 line data and to compare them with other star formation tracers , we need to identify the excitation mechanism responsible for the h@xmath0 emission . in the central regions of galaxies , however , the excitation mechanism for the h@xmath0 emission has long been the subject of intense controversy ( see the reviews of @xcite ; @xcite ) .
@xcite observed nearby spiral galaxies using a circular variable filter ( @xmath16 ) and a beam full - width at half maximum ( fwhm ) of 196 .
they detected , , and lines and concluded that the h@xmath0 line populations are non - thermal .
however , @xcite presented high spectral resolution data ( @xmath17 ) and pointed out that the flux measurement of @xcite could be affected by unresolved stellar absorption lines .
they argued that the h@xmath0 emission is mostly from shocks .
@xcite suggested that high spectral resolution observations are essential to discriminate between the excitation mechanisms .
the ultimate energy source responsible for the excitation of the h@xmath0 vibration - rotation emission can lie either in the absorption of uv photons from the ambient interstellar radiation field or from local ob stars , or in shock waves excited by supernovae , protostellar flows , or other energetic events @xcite .
hydrogen molecules that absorb far - uv radiation make a transition to an excited electronic state . at low densities ,
this transition is followed by fluorescence to the ground electronic state and a cascade through that state s vibration - rotation levels @xcite .
this interaction between h@xmath0 molecules and uv radiation takes place at the edges of molecular clouds where hydrogen makes the transition from predominantly atomic to predominantly molecular gas , in the photodissociation region ( pdr : @xcite ; @xcite ) .
if @xmath18 @xmath19 @xmath20 @xmath21 , the relative intensities of vibration - rotation transitions in the near - ir arising in uv - excited h@xmath0 are insensitive to density or to uv field strength @xcite and depend only on the branching ratios in the downward cascade . in dense (
@xmath18 @xmath6 @xmath22 @xmath21 ) gas heated by intense far - uv radiation , @xcite showed that inelastic h@xmath23h@xmath0 and h@xmath0@xmath23h@xmath0collisions can heat the gas at the cloud edge to temperatures approaching 1000 k. a combination of the collisional de - excitation and collisional excitation in the high temperature layer alters the h@xmath0 level populations in this uv - excited gas .
the expected changes in relative line strength with density have been verified by observations of molecular cloud edges @xcite . in shocked regions , the h@xmath0 vibration - rotation energy level populations in the ground electronic state
are thermalized by collisions and the level populations are usually in lte @xcite .
because the level populations are thermalized and reflect a similar range of temperatures , shock - heated gas and hot , dense gas heated by uv radiation are difficult to distinguish .
most previous observations of the central regions of galaxies showed that the h@xmath0 line ratios of to lie between the ratios one would expect for fluorescently excited h@xmath0 and what one would see in observations of uv - heated but collisionally excited gas or of gas collisionally heated in shocks @xcite .
these intermediate ratios suggest that both pure uv - fluorescence and thermalized uv - excited ( or shock excited ) gas are responsible for the h@xmath0 emission @xcite .
detection of h@xmath0 lines arising from upper vibrational states with @xmath24 is a strong indication of the presence of uv - fluorescent excitation since the high energies of the upper states make it very hard to excite these transitions thermally .
there are marginal detections of such transitions from a few galaxies , e.g. , a giant region ngc 5451 in m101 @xcite , arp 299 @xcite , and some ultraluminous ir galaxies @xcite .
another method for discriminating between the h@xmath0 excitation mechanisms is to measure the h@xmath0 nuclear spin degeneracy ratio ( ortho- to para - ratio ) , which may deviate from its formation value if the gas is thermally processed , but remains constant in the case of non - thermal excitation .
observations of and showed that the h@xmath0 emission in the galactic centers arises mostly from uv - fluorescence @xcite .
more recently , @xcite have observed vibration - rotation transition lines of h@xmath0 from the nuclei of seven ulirgs and concluded that the line emission arises from dense pdrs irradiated by intense far - uv radiation fields . in galactic sources , we have used observations of h@xmath0 lines from high vibrational states @xcite , as well as comparisons of extended h@xmath0 emission and tracers of uv excited pdrs such as far - ir continuum and ] emission to show that much of the h@xmath0 emission arises from uv - excited gas @xcite .
this comparison with far - ir lines and continuum has been more difficult in other galaxies because of the significant differences between the ] and far - ir beam sizes and the typical slit sizes used for near - ir spectroscopy of h@xmath0 . in this work
, we will use both h@xmath0 line ratios and a study of various tracers of star forming clouds in the context of pdr models to learn more about the relative significance of various types of h@xmath0 excitation . in section [ sec :
observations ] , we present a description of our spectroscopic observations of the starburst galaxies m82 and ngc 253 , and the normal spiral galaxies ngc 6946 and ic 342 . in section [ sec : results ]
we first show the observed h@xmath0 spectra and their extinction - corrected line fluxes .
we then show the spatial distribution of the h@xmath0 line intensity and compare it to the distribution of other pdr tracers , e.g. , co line emission and sub - mm continuum . in section [ sec : discussion ]
we analyze the nature of the h@xmath0 excitation mechanism by comparing the measured h@xmath0 line ratios with thermal and non - thermal models .
we also calculate the luminosity ratios of h@xmath0 to far - ir and ] to far - ir , and discuss the implications of the observed ratios by comparison to predictions by pdr models .
section [ sec : conclusions ] summarizes our conclusions .
we have observed emission lines resulting from vibration - rotation transitions of h@xmath0 from the central @xmath25 kpc regions of m82 , ngc 253 , ngc 6946 , and ic 342 .
table [ tbl-1 ] lists the basic parameters of the four galaxies in our sample .
we made the observations at the 1.5 m telescope of the cerro tololo inter - american observatory ( ctio ) and the 5 m hale telescope of the palomar observatory with the university of texas near - ir fabry - perot spectrometer ( utfps ) .
the spectrometer was designed to observe extended and low surface brightness line emission by maximizing the area - solid angle product @xcite . in order to subtract the telluric oh line emission , we chopped the secondary mirror from the source position to @xmath26 ( at ctio ) or @xmath27 ( at palomar ) at 0.5 hz . the fabry - perot etalon was scanned to cover the emission profiles , and each spectral step consisted of one chopping cycle : object sky sky object ( see table [ tbl-2 ] ) .
the detailed procedure for the utfps observations is described in @xcite .
the instrument , as used for the observations described in this paper , differed in one respect from the instrument described in @xcite . for the 1996 september run
, the fabry - perot etalon was housed in an enclosure cooled thermoelectrically to @xmath28c to reduce the background signal and to stabilize the instrument against drifts in the plate spacing .
the beam shape on all runs was approximately cylindrical @xcite .
we indicate the beam size at each telescope using the equivalent disk size , @xmath29 , the diameter of a cylindrical beam whose solid angle is the same as the integrated solid angle of the actual beam pattern .
since the one - dimensional beam pattern was close to that of a box function , the fwhm of the beam pattern is almost the same as @xmath29 . after the 1995 december run at the palomar observatory
, we changed the optical configuration inside the instrument .
the equivalent disk size on the 1996 september run ( @xmath30 ) was smaller than that on the 1995 december run ( @xmath31 ) .
the telescope pointing uncertainty was , in all cases , less than 10 percent of the beam size .
as the area - solid angle product of the instrument changes , the resolving power of the spectrometer also changes by a small amount , i.e. , r=@xmath32 varies between 2300 and 3000 .
table [ tbl-2 ] therefore also lists the effective spectral resolution for each wavelength and observing run .
the flux calibration was done by measuring standard stars at the center of the beam profile .
the flux calibration uncertainty is less than 20 percent .
table [ tbl-2 ] summarizes the observation logs , and lists the instrument parameters and the flux calibration stars . on the 1995 december run and the 1996 september run
, we used the hale telescope , where the utfps beam has a linear size of @xmath33 pc in our target galaxies , to map the central regions of these galaxies ( except ngc 6946 , where only a single point was observed ) .
we observed h@xmath0 line emission along the major axis or the molecular bar known previously from co observations ( see table [ tbl-1 ] ) stepping either 10 or 15between the individual measurements . at some positions
, we also observed line emission . on the 1995 october ctio run
, we observed the , , and lines in ngc 253 with a beam size of 81 ( @xmath341 kpc in this galaxy ) .
this large beam covers the entire region scanned during the 11 beam observations in ngc 253 .
we observed h@xmath0 emission lines from m82 with the ir cryogenic spectrometer ( crsp ; @xcite ) at the 2.1 m telescope of the kitt peak national observatory ( kpno ) in 1996 december .
the crsp is a long - slit spectrometer with a slit length of 824 ( 135 pixels @xmath35 061 ) .
we used a 300 l / mm grating and a slit width of 2.7 pixels ( or 165 ) and the resulting spectral resolution was @xmath36 @xmath37 @xmath38 km s@xmath39 at @xmath40 , corresponding to r@xmath411060 .
the grating was tilted to take spectra in the @xmath42 band which includes the h@xmath0 2.12183 , 2.22329 , and 2.24771 emission lines .
the crsp slit used toward m82 was rotated to match the major axis ( p.a .
= 73 ) of the co molecular bar @xcite . in order to cover the extended emission from the central region
, we also took spectra with the slit displaced by @xmath43 perpendicular to its length : ( @xmath44 , @xmath45 ) @xmath37 ( @xmath46 , @xmath47 ) , ( 0 , 0 ) , and ( @xmath48 , @xmath49 ) .
we co - added 9 pixels along the spatial direction ( or slit length direction ) and co - added the data taken at all three slit positions to make high signal - to - noise spectra reflecting the near - ir emission from a rectangular beam of @xmath50 . in the central region of m82 ,
the h@xmath0emission may extend up to 10 in the minor axis direction ( see section [ sec : distr_m82 ] ) .
even though the co - added rectangular beam did not include all of the flux from m82 , the h@xmath0 line ratios from these data are more reliable than those derived from fabry - perot observations made at different times because the ratios are less dependent on the sky conditions ( see section [ sec : h2ratio ] ) .
the crsp measurements can therefore contribute to our analysis of the h@xmath0 excitation mechanism .
figures @xmath51 show the h@xmath0spectra taken with the utfps toward m82 , ngc 253 , ngc 6946 , and ic 342 . on each h@xmath0 spectrum , we have overlaid a @xmath52co @xmath53spectrum taken with a similar beam size and smoothed in the spectral direction with a gaussian smoothing function to the resolution of the utfps to allow us to compare the radial velocities and the line shapes more carefully . in general ,
those h@xmath0 line profiles with high signal - to - noise are similar to the co profiles from the same positions .
the similarity argues for a similar distribution of line emissivity with galactocentric radius , implying that the h@xmath0 emission and co emission arise from the same molecular clouds . at the m82 positions ( e27,n13 ) and
( w27 , s13 ) , however , the co and h@xmath0 profiles differ .
at least part of the change in shape and velocity profile may result from the beam pattern difference between the co and h@xmath0 observations .
the spatial 16 beam of the nobeyama radio telescope is gaussian while that of the utfps 16 beam is cylindrical .
this shape difference could work in concert with the large gradient in rotational velocity at these positions to produce the profile differences .
figure [ fig-08 ] shows selected spectra taken by crsp toward m82 .
the h@xmath0 lines are prominent and detected at all positions .
the and lines are clearly detected at the ( e105 , n32 ) , ( w105 , n32 ) , and ( w210 , s64 ) positions , but are confused with stellar absorption features at the ( 0 , 0 ) position where the continuum level is highest .
the relative strength of and in this work agrees with that of @xcite ( see section [ sec : h2ratio ] ) .
there is a possible detection of the line at ( w210 , s64 ) , but this line was not detected at other positions . in figure [ fig-02 ]
, we overlaid the h@xmath0 and lines from m82 , as observed with the utfps .
the h@xmath0 emission at ( 0 , 0 ) is very similar to the line profile of the transition while the match in line shapes is less clear at ( e14e , n6 ) where the signal to noise ratio of the observation is small .
fluxes and upper limits for the h@xmath0 and lines toward ngc 253 and ngc 6946 are listed in table [ tbl-3 ] .
table [ tbl-4 ] lists line ratios , principally for the to ratio , along the major axis of m82 .
s / n for the ratio determination ranges from 2 to 6 . however , temporal variations in telluric absorption may increase the uncertainty for the line ratios derived from the utfps data .
we determine the h@xmath0 line flux by integrating across the velocity interval containing the line .
we define the line emission bandwidth , @xmath54 , based on the @xmath52co @xmath53 or @xmath55line spectra and list the values in table [ tbl-3 ] . for each profile , we integrate over this velocity interval , subtracting a continuum value determined from the average of the channels at the two ends of the spectrum .
we label the flux in the portion of the spectrum where we are integrating over the line as ( @xmath56 ) and that over the spectral region containing only continuum as ( @xmath57 ) where the subscript @xmath58 is an index for the individual samples along the spectrum . the integrated line flux ( @xmath59 ) is calculated as : @xmath60 where @xmath61 is the average of @xmath57 and @xmath62 is the interval between the channels in the scanning mode ( see table [ tbl-2 ] ) .
the error ( or noise ) of the integrated line flux is derived from the standard deviation , @xmath63 , of @xmath57 : @xmath64 where @xmath65 and @xmath66 are the numbers of individual spectral samples in the emission and continuum bands , respectively .
table [ tbl-3 ] lists the observed positions and the measured fluxes of the h@xmath0 , , and emission lines .
note that the flux values , at the same wavelength and the same position , vary with the beam size because the emission regions are extended .
the fwhm ( @xmath67 km s@xmath39 ) of the instrumental profile of crsp is wider than the typical fwhm of m82 emission lines . assuming the line profile is gaussian
, we derive the h@xmath0line flux by using the single profile fitting task in the _ iraf _ package , splot .
the error in the flux is derived using equation ( [ eq : fluxsigma ] ) with @xmath68 and @xmath69 @xmath37 @xmath70 . the molecular hydrogen flux that we observe likely arises from the surfaces of giant molecular clouds exposed to far - uv radiation @xcite .
therefore , as for the large - scale co line emission from galaxies , the area filling factor for the observed h@xmath0 line emission is likely to be less than unity , so that much of the emergent h@xmath0 will suffer only modest extinction and the foreground screen model for extinction is appropriate .
if the h@xmath0 emission is widespread and arises from the surfaces of clouds , then there would only be a loss of half the h@xmath0 flux ( from the back side of each cloud ) , and a further loss from the foreground extinction .
the screen model is highly effective in modeling the extended h@xmath0 line emission in the central kiloparsec of the milky way . here
the h@xmath0 emission would not be detectable at all if the emitting sources were uniformly mixed with molecular gas clouds , since co observations indicate these clouds typically have @xmath71 @xmath72 @xcite . the far - uv radiation that gives rise to the observed h@xmath0 line emission arises from a slightly older population of stars than those that excite regions .
these older stars are not likely to be still embedded in their natal clouds and can therefore produce the observed widespread h@xmath0 emission .
in contrast , the recombination lines likely emerge from regions excited by very early type stars that are still embedded in their natal molecular clouds .
therefore , one might expect the extinction of the recombination lines to be significantly higher , and the mixed extinction scenario may be more likely to apply to these lines .
however , the application of a screen model even to the observed br@xmath73 flux from both ultraluminous ir galaxies ( ulirgs ) @xcite and normal and nearby starburst galaxies @xcite results in a reasonably tight br@xmath73 to far - ir correlation that suggests no large extinction corrections are required .
the foreground screen model is applicable for all of the galaxies in our sample .
the most controversial galaxy in our sample is the nearby , highly inclined ( @xmath74 ) galaxy , m82 @xcite .
extensive modeling of the observed h recombination line emission from m82 suggests @xmath71 @xmath37 3 to 9 magnitudes when a uniform foreground screen model is applied @xcite .
however , a model where the dust and regions are homogeneously distributed also fits the data and suggests that @xmath71 @xmath37 25 to 52 magnitudes @xcite .
we argue that while a mixed model _ may _
apply for the recombination lines , it is not appropriate for the h@xmath0 lines due to the argument outlined above . to compare the extinction corrected results , in table [ tbl-5 ] , we applied both foreground model and mixed model to the h@xmath0 data of m82 . in converting from visual extinction , @xmath71 , in table
[ tbl-5 ] to ir extinction , @xmath75 , at different wavelengths , we used the ir interstellar reddening law ( section 7.8 . in @xcite and references therein ) : @xmath76 where @xmath77 is in units of . in section
[ sec : h2ratio ] and table [ tbl-4 ] , where we compare the h@xmath0 line ratios , we corrected for differential extinction for lines at different wavelengths .
our new h@xmath0 measurements of galaxies are particularly useful for comparison with other observations of line and continuum radiation from neutral atomic and molecular gas because the size of our utfps beam is more comparable to the sizes of beams used to measure the other pdr tracers than are the arcsec - scale beam sizes of typical near - ir slit spectrometers . here we compare our h@xmath0results to far - ir and submillimeter continuum , and ] 158 and @xmath52co line data assembled from the literature .
figure [ fig-09 ] shows the integrated intensities of h@xmath0 , @xmath52co @xmath53 @xcite , and @xmath52co @xmath78lines @xcite and the specific intensity of submillimeter continuum at 450 @xcite along the major axis of m82 ( p.a .
@xmath37 @xmath79 ; see tables [ tbl-1 ] and [ tbl-3 ] ) .
the 450 data were obtained with a 9 beam .
we smoothed these data to match the spatial resolution of our 16 beam observations .
the beam sizes of the co @xmath53@xcite and the co @xmath78
@xcite data are slightly larger ( 16 and 18 , respectively ) than the h@xmath0beam . in general , the distribution of co , submillimeter continuum , and h@xmath0 emission agree rather well
. the differences are consistent with a common emission region with excitation differences accounting for the small differences in source size .
co @xmath53observations show emission peaks on either side of the nucleus ( separation @xmath80 ) , indicating a rotating molecular ring @xcite .
@xcite observed a smaller separation ( @xmath81 or 270 pc ) of the corresponding peaks in co @xmath78 .
thus , higher gas temperature and higher density clouds may exist at the inner edges of the rotating molecular ring .
while the convolved scan map at 450 that we show in figure [ fig-09 ] has a plateau at the nucleus , the original diffraction - limited map shows weak double peaks separated by @xmath82 @xcite .
this small separation supports the idea that the rotating molecular ring has a temperature gradient , since the dust continuum emission scales at least linearly with temperature @xcite .
the similarities in the co , submillimeter continuum , and h@xmath0 maps are consistent with a picture in which the h@xmath0 emission from m82 arises primarily from the molecular clouds in the circumnuclear ring .
figure [ fig-10 ] shows the integrated intensities of h@xmath0 , @xmath52co @xmath53 @xcite , and @xmath52co @xmath83@xcite as well as the specific intensity of submillimeter continuum at 850 @xcite along the molecular bar in ngc 253 which has an angular extent of @xmath84 at a position angle of 64 .
the molecular bar is tilted from the major axis ( p.a .
= 51 , @xcite ) and apparently rotates as a rigid body @xcite .
the co @xmath53 data were obtained at the owens valley radio observatory with a @xmath85 beam .
we smoothed them to match our utfps spatial resolution .
the beam sizes of the co @xmath83 line and the 850 continuum observations are 14 and 15 , respectively .
the h@xmath0 emission distribution in figure [ fig-10 ] has a sharp peak at the center and extends with relatively low intensities out to @xmath86 kpc from the nucleus .
the slight enhancement of h@xmath0intensity at the southwestern bar ( at the positions between @xmath87 pc and @xmath88 pc shown on the horizontal axis in figure [ fig-10 ] ) may be related to the southwest peak that is also present in the overlaid co @xmath53 distribution .
the concentration of the h@xmath0 emission in this galaxy can be estimated by comparing the flux measurements with different aperture sizes .
@xcite derived an h@xmath0 flux from their @xmath89 slit of @xmath90 w m@xmath91 . from our scanned observations along the bar with the 11 beam ( @xmath92 ) ,
the summed h@xmath0 flux is @xmath93 w m@xmath91 .
from the 81 beam observations at the ctio the flux is @xmath94 w m@xmath91 ( see tables [ tbl-3 ] and [ tbl-5 ] ) .
we conclude that the h@xmath0 emission from the central @xmath2 kpc region is dominated by the 39 long molecular bar .
@xmath52co @xmath95 observations of ngc 6946 show that the molecular gas is concentrated in the central 10 region ( or 0.3 kpc at the distance of 5.5 mpc , @xcite ) .
using near - ir long - slit spectroscopy , @xcite also found that the h@xmath0 emission extends up to 12 along the major axis .
the flux in our 11 beam is @xmath96 w m@xmath91 .
we expect that our beam covers most of the h@xmath0emission in the central region of ngc 6946 .
an aperture synthesis map of @xmath52co @xmath53 in the central starburst region of ic 342 @xcite shows a 7(or 60 pc at the distance of 1.8 mpc ) diameter ring - like structure within the extended ( @xmath97 ) molecular bar .
our beam size , 16 covers the whole molecular ring structure .
we also observed regions offset by 16 along the molecular bar ( see table [ tbl-3 ] ) .
figure [ fig-11 ] shows scan maps of the integrated intensities of the h@xmath0 , @xmath52co @xmath53@xcite , and @xmath52co @xmath83 @xcite lines . since the spatial resolution of the co @xmath53 observations is 4and that of the co @xmath83 observations is 7
, we smoothed the co maps to match to our utfps spatial resolution of 16 . in this galaxy , the h@xmath0 emission region is unresolved and therefore significantly more compact than the co emission region .
this difference between ic 342 and m82 and ngc 253 may , however , be due to a more extreme difference in emissivities rather than to a separation of emission regions . @xcite
observed the central region of ic 342 and found an h@xmath0 emission region corresponding to the 7 co ring .
if h@xmath0 line emission traces molecular gas illuminated by far - uv photons , its distribution and strength should be related to those of other pdr probes .
we have compiled far - ir continuum , ] line , and ] line emission results from the literature .
far - ir continuum , ] , and ] emission have been taken with low spatial resolution ( @xmath98 ) .
the individual beams therefore cover @xmath2 kpc regions in the centers of nearby galaxies ( @xmath99 mpc ) .
our big beam fabry - perot data cover comparable size scales .
the co interferometer map of m82 @xcite shows that the minor - axis width of the rotating ring - like bar is @xmath100 .
our scan map along the major axis with a 16beam therefore most likely covers most of the h@xmath0 emission region in the center . the total h@xmath0 flux , @xmath101 , from the @xmath102 region within our map is @xmath103 w m@xmath91 ( see table [ tbl-5 ] ) .
ngc 253 is also an edge - on galaxy with the inclination of 78@xcite .
the minor - axis width of the molecular bar , from the co observations , is @xmath104 @xcite , which is about the same as our 11 beam . for this galaxy
, we have the 81 ctio measurement to provide the total h@xmath0 flux .
the active star forming regions in ngc 6946 and ic 342 are relatively small compared to those in the starburst galaxies , m82 and ngc 253 .
one utfps beam covers the h@xmath0 emission region .
we converted the total h@xmath0 flux values into extinction corrected flux values assuming that the foreground model is applicable ( see section [ sec : extinction ] and table [ tbl-5 ] ) .
when making use of iras 60 and 100 observations , one can define the total far - ir continuum flux , @xmath105 , as the flux measured from the source within a bandpass of 80 centered at 82.5 @xcite .
in addition to the iras observations , the far - ir fluxes of the galaxies in our h@xmath0 sample were measured using the university of texas far - ir photometer ( see the references in @xcite ) on the kuiper airborne observatory ( kao ) and using the long wavelength spectrometer ( lws ; @xcite ) on the infrared space observatory ( iso ; @xcite ) .
the kao observations spatially resolved the emission regions with a beam size of @xmath106 .
the spatial resolution of the iras data , which were processed using a maximum correlation method algorithm @xcite , is @xmath107 at 100 , and that of the iso / lws is @xmath108 @xcite . in section [ sec : distr ] , we found that most of the submillimeter dust continuum emission arises from regions @xmath109 in diameter . even though the measured luminosities from the kao instrument are slightly less than those from the iras and iso , the different aperture sizes do not significantly affect the measured luminosity . compiling the far - ir luminosities from the iras , the kao , and the iso / lws
, we list in table [ tbl-6 ] the average @xmath105 of the published values . the ] 63 flux , @xmath110 , and the ] 158 flux , @xmath111 were measured with the u. c. berkeley cryogenic tandem fabry - perot spectrometer on the kao ( see the references in @xcite ) and by the iso / lws .
the fwhm beam sizes are 55 and 69 for the kao and iso respectively . since the emission regions are concentrated in the central @xmath112 of the galaxies ( sections [ sec : distr ] and [ sec : l_fir ] ) ,
the flux values measured by both kao and iso can be compared with our h@xmath0 measurement of the galaxies .
table [ tbl-6 ] lists the ] and ] line fluxes .
a resolution of the long - running debate about the excitation mechanism for the vibration - rotation h@xmath0 lines in galaxies and about the physical nature of the line formation region is critical if we are to reap any benefit from the diagnostic power of these lines .
we need to examine the observational evidence in the context of models for the formation and excitation of h@xmath0 to decide whether uv excitation followed by a radiative cascade or collisional excitation account for the population of the upper states of the near - ir transitions . we need to determine if any collisionally excited gas we do see was heated by uv light incident on the neutral cloud or by mechanical energy in the form of shocks . like giant molecular clouds in our own galaxy , the ism of other spirals and of starburst galaxies
is bathed in uv radiation from main - sequence ob stars that have already freed themselves from the dense cores in which they formed .
orion serves as a good example for the galactic molecular clouds .
the h@xmath0 emission attributable to the pdrs produced by uv light has a luminosity of 34 @xmath8 over a 15 pc diameter region centered on @xmath113 ori .
the luminosity of the same line attributable to the well - known and very prominent orion / kl shock is only 2.5 @xmath8 @xcite .
observations of several h@xmath0 vibration - rotation transitions with the utfps , including detection of the line , and comparisons to other pdr tracers led @xcite to conclude that the majority of the h@xmath0 emission from several giant molecular clouds in the large magellanic cloud is a result of uv excitation .
the ubiquity of uv emission incident on the surfaces of molecular clouds in the inner regions of galaxies means that these galaxies will always produce some uv - excited h@xmath0 emission .
the real question then is whether this emission or emission from shocks produced by some combination of protostellar jets , supernovae , and large - scale gas flows will dominate the observed flux . in this section
, we examine whether uv excitation is sufficient to account for all of the h@xmath0 emission observed toward the galaxies in our sample .
the ratios of h@xmath0 lines arising in states with different upper state energies and the h@xmath0 ortho- to para - ratios give us a way to estimate the fraction of the emission produced by each excitation process within the observed beam . in pdrs where @xmath114 @xmath19 @xmath22 @xmath21 ( sources where a radiative cascade determines the level populations ) , the ratios of to and to are @xmath115 and @xmath116 , respectively @xcite . in shocked regions ,
e.g. , bipolar outflows @xcite and supernova remnants @xcite , the observed ratio of to line flux is @xmath117 , which can be explained by a combination of j- and c - type bow shocks @xcite .
if the level populations of the h@xmath0 molecules were in lte , this ratio would be appropriate for gas with @xmath118 k @xcite . at this temperature ,
the ratio of to is only @xmath119 ( see table [ tbl-4 ] ) . in table
[ tbl-4 ] , the observed h@xmath0 ratios of to from m82 , ngc 253 , and ngc 6949 are @xmath120 , which lies between the ratios in the thermal and non - thermal excitation cases .
if we assume that h@xmath0 emission arises from a mixture of thermal excitation at one lte temperature and non - thermal excitation following the rules for a uv - excited cascade , we can estimate the fraction of the total emission in h@xmath0 lines contributed by uv - excited gas . in the discussion that follows
, we use @xmath121 to represent the fraction of the h@xmath0 emission ( summed over all h@xmath0transitions ) contributed by uv fluorescence .
appendix [ sec : re ] gives a detailed explanation of how we derive @xmath121 from the observed line ratios .
figures [ fig-12 ] , [ fig-13 ] , and [ fig-14 ] show the @xmath121 values derived from our utfps and crsp data from m82 and ngc 253 .
we also plot the values of @xmath121 derived from other observations of h@xmath0 emission in those galaxies @xcite .
the results show that , at almost all positions in both galaxies , the majority of the total h@xmath0 emission comes from regions excited radiatively by uv photons .
figure [ fig-14 ] shows the values of @xmath121 derived from a variety of h@xmath0 vibration - rotation transitions for ngc 253 .
most lines give values of @xmath121 between 0.6 and 1 , again indicating the predominance of uv - excited emission .
if some of the remaining ( thermal ) emission arises from dense gas in pdrs , the non - thermal fraction may be even higher than the results from the analysis used in the figures indicates .
@xcite modeled the h@xmath0 emission from warm pdrs , where the h@xmath0vibrational level populations are thermalized by collisional de - excitation . in these regions , the h@xmath0 line has excitation temperature of @xmath122 k. if we use the relative line intensities derived assuming an excitation temperature , @xmath123 k for the collisionally excited gas , the derived @xmath121 values are closer to unity than those derived assuming @xmath124 k. our analysis shows that h@xmath0 emission from the central regions of galaxies is mostly dominated by non - thermal excitation .
uv radiation may also be responsible for some of the remaining thermal h@xmath0 emission if that emission arises in dense pdrs .
the analysis presented in appendix [ sec : h2fraction ] and illustrated in figure [ fig-15 ] points out that the picture one gets from the line alone is highly biased toward collisionally excited emission , because this line s upper state energy is lower than those of or lines
. a uv - excited radiative cascade populates many rotational states in higher vibrational levels . in purely radiatively excited
h@xmath0 , the line accounts for only 1.6% of the total h@xmath0 emission , while the same line from 2000 k gas in lte accounts for 8.5% of the ro - vibrational intensity @xcite . as a result , even when only a small fraction of the emission arises in uv - excited regions , _ most _ of the overall h@xmath0 emission is a product of uv excitation . in such cases ,
the thermal contribution to the emission could arise in dense pdr gas where uv radiation is the ultimate source of the emission , or it could come from shocks associated with supernova remnants or large - scale flows .
a comparison of h@xmath0 luminosity with fine structure line emission from pdrs in galaxies will help distinguish between these possibilities . in starburst nuclei ,
the far - ir luminosity arises mostly from pdrs illuminated by recently formed ob stars .
far - ir emission has been used to estimate the star formation rates ( sfrs ) of such galaxies @xcite .
the @xmath125 c@xmath126 emission at 158 from galactic nuclei also arises mostly from the warm photodissociated surfaces of far - uv exposed molecular clouds , with lesser contributions from low density ionized gas regions , and from the cold neutral medium @xcite .
the average ] to far - ir luminosity ratio for the four galaxies in our sample is @xmath127 ( see also figure [ fig-16 ] ) , where the 0.27 represents the standard deviation about the mean ratio .
the fairly constant ratio is consistent with a pdr origin for both @xcite . in figure
[ fig-16 ] , we also plot the ratio of h@xmath0 to far - ir luminosity versus the far - ir luminosity from our observations .
the @xmath3 ratio is almost constant , with a mean value of @xmath128 .
@xcite found a similar proportionality between the extinction corrected h@xmath0 and far - ir luminosities from lirgs ( @xmath129 @xmath19 @xmath5 @xmath19 @xmath130 ) .
their data give a value for @xmath3 from the lirgs of @xmath131 , which is consistent with our value for starbursts and normal spiral galaxies . in calculating the mean ratio for nearby galaxies
, we assumed that the foreground model was applicable when making the extinction correction to the m82 h@xmath0 data ( see section [ sec : extinction ] ) .
if we applied the mixed model result , 52 magnitudes of extinction , to the observed h@xmath0 line flux , the corresponding ratio of h@xmath0 line to far - ir continuum of m82 would be very large , @xmath3 @xmath41 @xmath132 ( see tables [ tbl-5 ] and [ tbl-6 ] ) .
this ratio is two orders of magnitude larger than the extinction corrected line to continuum ratios typical for ir bright and lirg galaxies @xcite and six times larger than ratio found for the most extreme h@xmath0 luminous system , ngc 6240 @xcite .
the constant ratio of @xmath3 and @xmath133 suggests that the far - ir , ] , and h@xmath0 emission regions and their emission mechanisms are directly related . in examining this possibility , we assume that the h@xmath0 luminosity can represent the total h@xmath0 transition luminosity .
if the h@xmath0emission is mostly from pdrs and its energy level populations are partly thermalized , the relationship between the h@xmath0 and the total h@xmath0 transition luminosity is given by equations ( [ eq : re_t ] ) and ( [ eq : i_v ] ) in sections [ sec : re ] and [ sec : h2fraction ] .
when @xmath134 , 0.8 , and 1 , the ratio of total h@xmath0 luminosity to luminosity in the i(1,0)s(1 ) line is 23 , 34 , and 63 , respectively .
since the fraction of the non - thermal excitation , @xmath121 , observed in the galaxies in our sample is @xmath135 , the error introduced by using the line as an indicator of the total h@xmath0 transition intensity is a factor of @xmath136 ( or @xmath137 ) .
this uncertainty is comparable to the standard deviation of the observed @xmath3 ratio from normal and starburst galaxies ( @xmath138 ) , and lirgs ( @xmath139 ) . using the line as a proxy for the total h@xmath0luminosity
, we now compare the observed h@xmath0 emission to the intensity that independently derived pdr models would predict for the same objects . in this section ,
we examine the possibility that the observed h@xmath0emission arises in pdrs by comparing the average observed h@xmath0 intensity to the intensity predicted from pdr models with parameters set by observed far - ir continuum and ] and ] line strengths .
the emission from pdrs can be parameterized by the cloud density , @xmath140 , and the strength of the far - uv radiation field illuminating the cloud , @xmath141 @xcite .
the fine structure line ratio @xmath142 and the fine structure line to far - ir continuum ratio @xmath143 ( see the values in table [ tbl-6 ] ) are relatively free from beam filling factor effects and can be used to derive the density and local uv field in the pdr .
we compared the observed flux of ] , ] , and far - ir ( table [ tbl-6 ] ) to the pdr models of @xcite , and list the best - fit values of @xmath18 and @xmath141 in table [ tbl-7 ] .
these values are different from the previous results in @xcite .
one reason for the difference is that the pdr model of @xcite used in this work includes additional gas heating due to polycyclic aromatic hydrocarbons ( pahs ) which results in a lower predicted gas density @xcite .
the h@xmath0 line intensity , @xmath144 , is also a function of the strength of the external far - uv radiation field , @xmath141 , impinging on the pdr , and the gas density , @xmath18 , in the pdr . in order to compare this predicted h@xmath0 intensity for a derived @xmath18 , @xmath141 pair with the observed values , e.g. , @xmath3 in figure [ fig-16 ] , it is necessary to derive the intensity ratio of h@xmath0 to far - ir , @xmath145 from the models .
we can convert the derived far - uv radiation field strength , @xmath141 in table [ tbl-7 ] , into far - ir intensity , @xmath146 , assuming that most of the far - uv photons are absorbed by grains and reradiated in the far - ir : @xmath147 where the factor of 2 is for the photons with @xmath148 .
figure [ fig-17 ] shows the intensity ratio , @xmath145 , as a function of @xmath141 and @xmath18 . to calculate the h@xmath0 intensity
, we used a second pdr model that calculates abundances and temperatures in a plane - parallel slab for a given @xmath18 and @xmath141 @xcite and uses a monte carlo method for the radiative transfer @xcite .
we assume solar metallicity in the models because the h@xmath0 intensity is not sensitive to the metallicity @xcite . in figure [ fig-17 ] , we plot the locations of the galaxies in our sample in ( @xmath141 , @xmath18 ) space , as derived from the ] and ] observations using the pdr models .
the figure also shows the model predictions for the i@xmath149 ratio .
table [ tbl-7 ] lists the values , as well as the predicted strengths of the h@xmath0 line .
since the derived @xmath18 values of ngc 6946 and ngc 253 are lower than the model domain , we extrapolated the grids .
the resulting values of @xmath145 are listed in table [ tbl-7 ] .
the average of the h@xmath0 to far - ir ratios predicted by the pdr models from the fine - structure line observations is @xmath150 which is in very close agreement with the value of @xmath128 measured directly ( see section [ sec : h2ciifir ] ) .
this agreement argues strongly for uv excitation in pdrs as the primary source of h@xmath0 and , by implication , as the dominant cause of h@xmath0 emission overall in the central regions of galaxies .
our results are in good agreement with those of @xcite who conclude that the observed h@xmath0 line emission from a sample of seven ulirgs arises from dense pdrs exposed to strong ( @xmath141 @xmath41 @xmath151 ) radiation fields . in our study of the inner 400 pc of the milky way
, we found that most h@xmath0 emission arises as a result of uv excitation and that the ratio of h@xmath0 luminosity to far - ir luminosity is @xmath152 @xcite .
as shown in sections [ sec : h2ratio ] , [ sec : h2ciifir ] , and [ sec : model ] , the results for other galaxies are similar ; uv excitation dominates and the @xmath3 ratios are @xmath128 for spiral galaxies and local starburst galaxies and @xmath131 for lirgs .
the relationship between far - ir intensity and h@xmath0intensity as conditions in a pdr vary can be explained in terms of the competition between h@xmath0 absorption and dust absorption of far - uv photons .
@xcite showed analytically that the penetration of far - uv photons into a cloud is determined by h@xmath0self - shielding rather than by dust extinction when : @xmath153 in figure [ fig-17 ] , the dotted line divides the ( @xmath141 , @xmath18 ) plane into the h@xmath0 self - shielding dominant region and the region where shielding by dust is the major protector of h@xmath0 . in the h@xmath0 self - shielding case , the uv - excited h@xmath0 intensity , @xmath144 , is proportional to the far - uv field strength , @xmath141 because any additional uv photons are absorbed by h@xmath0 and most of these absorptions result in a fluorescent cascade . in this case , the @xmath145 ratio is always @xmath154 . in the case where dust shielding dominates ,
the h@xmath0 intensity is independent of the far - uv field but scales with the gas density , @xmath18 and the @xmath145 ratio is @xmath155 .
the distribution of our sample galaxies in the ( @xmath141 , @xmath18 ) plane ( figure [ fig-17 ] ) shows that the ratio of @xmath156 , derived from the observations of ] and ] lines and far - ir continuum ( see section [ sec : model ] ) , is lower than 80 in all cases , consistent with the observed @xmath145 ratio .
dust absorption of far - uv photons should therefore dominate over h@xmath0self - shielding in the central regions of these galaxies .
the observed constant ratio of h@xmath0 intensity to far - ir intensity is surprising in view of the results of the pdr modelling . in the dust - shielding regime
, this constant ratio can only come about if the gas density increases as the far - uv field strength increases . for a uniform medium , with an observed @xmath157 @xmath37 @xmath158 ,
the implied relationship is @xmath159 this inferred relation joins a long string of similar correlations between mass or density and surface density indicators and proxies for the star formation rate .
these include : the schmidt law @xcite , as determined from h@xmath160 and co j=@xmath161 observations @xcite , the scaling of @xmath18 and g@xmath162 derived from ] , ] and far - ir intensities @xcite , the strong correlation of the bolometric luminosity and virial mass of dense galactic cores derived from cs and far - ir observations @xcite and the close correlation of bolometric luminosity with the amount of dense gas in galaxies @xcite .
one possibility , then , is that the relationship between density and incident uv field inferred from the constant @xmath145 ratio is telling us that the surfaces of molecular clouds in galaxies , after stars have begun to form , have higher densities when the local uv field is higher .
this conclusion is not unreasonable , given that less dense clumps will be ionized and photo - evaporated more rapidly and that the average densities of surviving clumps should be higher when the incident uv field is stronger .
we have analyzed near - ir h@xmath0 emission from the central @xmath163 kpc regions in the nearby starburst galaxies , m82 and ngc 253 , and the vigorously star - forming galaxies , ngc 6946 and ic 342 .
the line ratios of various h@xmath0 vibration - rotation lines , e.g. , , and to , show that , while both thermal and non - thermal excitation are responsible for the emission , non - thermal excitation predominates .
in addition to the h@xmath0 line ratios , the constant ratio between h@xmath0 , ] , and far - ir luminosity suggests that the emission in the two lines and the far - ir continuum originate together in photodissociation regions .
we also compared scan maps of h@xmath0 emission along the major axes ( or molecular bars ) of the galaxies with maps of co and sub - mm emission .
the distribution of h@xmath0 line emission correlates with the sub - mm continuum emission , and also agrees well with the distribution of co emission .
the h@xmath0 emission traces active star formation regions where the far - uv fields are intense and the gas is dense and warm .
the correlation of h@xmath0 and sub - mm emission suggests that clouds have similar structure or that the star formation and cloud distributions are similar in the central regions of these galaxies .
we compare the luminosities of h@xmath0 , ] 158 , and far - ir continuum .
the @xmath3 ratios of normal spiral and starburst galaxies are the same . comparing the observed ratio to pdr models ,
we conclude that the column of warm h@xmath0 line emitting gas is limited by the extinction of far - uv photons by dust , rather than by h@xmath0 self - shielding in the pdrs . in this case
, the h@xmath0intensity is only sensitive to the gas density . the constant @xmath3 ratio over the large far - ir luminosity ranges of the galaxies
, then requires that the gas density in the pdrs increase as the far - uv ( or far - ir ) intensity increases .
we thank mark r. swain for help with observations at the palomar observatory , ewine van dishoeck for providing pdr model results , and michael luhman and thomas benedict for their work on the utfps .
this work was supported by nsf grant 95 - 30695 . s.p . appreciates the warm hospitality of the staff of the institute of space and astronautical science in japan .
assuming that h@xmath0 emission arises from a mixture of thermal and non - thermal excitation , the total h@xmath0 intensity summing over all vibration - rotation transitions , @xmath164 , and the intensity of a single transition , @xmath165 , can be expressed as @xmath166 where the superscripts @xmath167 and @xmath168 refer to the thermal and non - thermal excitation , respectively .
we can define the fraction of non - thermal h@xmath0 intensity in the observed h@xmath0intensity as @xmath169 the fractional intensity of thermal or non - thermal emission in a single transition relative to the total thermal or non - thermal emission is : @xmath170 table [ tbl-8 ] lists the adopted values of @xmath171 and @xmath172 for h@xmath0 , , , and from model calculations @xcite .
we assumed that the excitation temperature is 2000 k for thermalized gas . from the above equations we can derive the h@xmath0 line ratio as @xmath173 where the subscript ` 1 ' denotes the transition line . from equation ( [ eq : h2ratio ] )
, we can rewrite , @xmath121 , as a function of the observed h@xmath0 ratio of another transition to the line : @xmath174 the h@xmath0 at 2.1218 line is the most commonly observed near - ir h@xmath0 line because the wavelength resides in a clear part of the atmospheric window and the line is usually brighter than other h@xmath0 lines . when the level populations of excited h@xmath0 molecules are partly thermalized , the emission is enhanced relative to most other h@xmath0 lines because of its lower excitation requirements .
we need to check whether the h@xmath0 emission can represent the total h@xmath0 intensity .
we can generalize the fraction of the non - thermal excitation in a single transition , @xmath175 , in terms of @xmath121 as : @xmath176^{-1 } \ .
\label{eq : re_v}\ ] ] figure [ fig-15 ] shows @xmath175 versus @xmath121 for various transitions . for example , when @xmath177 , about 20 percent of the h@xmath0 emission arises from non - thermal excitation . on the other hand , we can extrapolate to the total h@xmath0vibration - rotation transition intensity , @xmath164 , from a single h@xmath0transition line emission , @xmath165 , as @xmath178 alton , p. b. , davies , j. i. , & bianchi , s. 1999 , , 343 , 51 aumann , h. h. , fowler , j. w. , & melnyk , m. 1990 , , 99 , 1674 becklin , e. e. , fomalont , e. b. , & neugebauer , g. 1973 , , 181 , l27 becklin , e. e. , gatley , i. , matthews , k. , neugebauer , g. , sellgren , k. , werner , m. w. , & wynn - williams , c. g. 1980 , , 236 , 441 black , j. h. , & van dishoeck , e. f. 1987 , , 322 , 412 bker , t. , frster - schreiber , n. m. , & genzel , r. 1997 , , 114 , 1883 bker , t. , van der marel , r. p. , & vacca , w. d. 1999 , , 118 , 831 burston , a. j. , ward , m. j. , & davies , r. i. 2001 , , 326 , 403 burton , m. g. , brand , p. w. j. l. , geballe , t. r. , & webster , a. s. 1989 , , 236 , 409 burton , m. g. , hollenbach , d. j. , & tielens , a. g. g. m. 1990 , , 365 , 620 burton , m. g. 1992 , aust . j. physics , 45 , 463 canzian , b. , mundy , l. g. , & scoville , n. z. 1988 , , 333 , 157 carral , p. , hollenbach , d. j. , lord , s. d. , colgan , s. w. j. , haas , m. r. , rubin , r. h. , & erickson , e. f. 1994 , , 423 , 223 choi , m. , evans , n. j. , ii , gregersen , e. m. , & wang , y. 1995 , , 448 , 742 clegg , p. e. , et al .
1996 , , 315 , l38 cox , a. n. 1999 , allen s astrophysical quantaties , ( 4th ed . ;
new york : springer - verlag ) crawford , m. k. , genzel , r. , townes , c. h. , & watson , d. m. 1985 , , 291 , 755 davies , r. i. , sternberg , a. , lehnert , m. , & tacconi - garman , l. e. 2003 , , 597 , 907 davis , c. j. , smith , m. d. , eislffel , j. , & davies , j. k. 1999 , , 308 , 539 doyon , r. , wright , g. s. , & joseph , r. d. 1994 , , 421 , 115 eckart , a. , downes , d. , genzel , r. , harris , a. i. , jaffe , d. t. , & wild , w. 1990 , , 348 , 434 engelbracht , c. w. , rieke , m. j. , rieke , g. h. , & latter , w. b. , 1996 , , 467 , 227 engelbracht , c. w. , rieke , m. j. , rieke , g. h. , kelly , d. m. , & achtermann , j. m. 1998 , , 505 , 639 frster - schreiber , n. m. , genzel , r. , lutz , d. , kunze , d. , & sternberg , a. 2001 , , 552 , 544 freedman , w. , et al .
1994 , , 427 , 628 gatley , i. , beattie , d. h. , lee , t. j. , jones , t. j. , & hyland , a. r. 1984 , , 210 , 565 genzel , r. et al .
1998 , , 498 , 579 goldader , j. d. , joseph , r. d. , doyon , r. , & sanders , d. b. 1995 , , 444 , 97 goldader , j. d. , joseph , r. d. , doyon , r. , & sanders , d. b. 1997a , , 474 , 104 goldader , j. d. , joseph , r. d. , doyon , r. , & sanders , d. b. 1997b , , 108 , 449 gry , c. et al .
2001 , the iso handbook , vol .
the long wavelength spectrometer , sai-99 - 077/dc , ver .
1.2 , esa harrison , a. , puxley , p. , russell , a. , & brand , p. 1998
, , 297 , 624 helou , g. , khan , i. r. , malek , l. , & boehmer , l. 1988 , , 68 , 151 hollenbach , d. j. , & tielens , a. g. g. m. 1999 , reviews of modern physics , 71 , 173 hughes , d. h. , gear , w. k. , & robson , e. i. 1994 , , 270 , 641 israel , f. p. , white , g. j. , & baas , f. 1995 , , 302 , 343 ishizuki , s. , kawabe , r. , ishiguro , m. , okumura , s. k. , morita , k .- i . , chikada , y. , kasuga , t. , & doi , m. 1990a , , 355 , 436 ishizuki , s. , kawabe , r. , ishiguro , m. , okumura , s. k. , morita , k. , chikada , y. , & kasuga , t. 1990b , , 344 , 224 jansen , d. j. , van dishoeck , e. f. , black , j. h. , spaans , m. , & sosin , c. 1995 , , 302 , 223 joyce , r. r. , fowler , a. m. , & heim , g. b. 1994 , in proc .
spie , vol .
2198 , instrumentation in astronomy viii , ed .
d. l. crawford & e. r. craine , 725 kaufman , m. j. , wolfire , m. g. , hollenbach , d. j. , & luhman , m. l. 1999 , , 527 , 795 kawara , k. , nishida , m. , & gregory , b. 1990 , , 352 , 433 kennicutt , r. c. 1998 , , 498 , 541 kessler , m. f. , et al .
1996 , , 315 , l27 klumpe , e. w. 1999 , phd thesis , univ . of texas at austin koornneef , j. , & israel f. p. 1996
, new astronomy , 1 , 271 lester , d. f. , carr , j. s. , joy , m. , & gaffney , n. 1990 , , 352 , 544 lo , k. y. , et al .
1984 , , 282 , l59 lord , s. d. , hollenbach , d. j. , haas , m. r. , rubin , r. h. , colgan , s. w. j. , & erickson , e. f. 1996 , , 465 , 703 luhman , m. l. , jaffe , d. t. , keller , l. d. , & pak , s. 1994 , , 436 , l185 luhman , m. l. , jaffe , d. t. , keller , l. d. , & pak , s. 1995 , , 107 , 184 luhman , m. l. , & jaffe , d. t. 1996 , , 463 , 191 luhman , m. l. , jaffe , d. t. , sternberg , a. , herrmann , f. , & poglitsch , a. 1997 , , 482 , 298 lumsden , s. l. , puxley , p. j. , & doherty , r. m. 1994 , , 268 , 821 lynds , c. r. , & sandage , a. r. 1963 , , 137 , 1005 madden , s. c. , geis , n. , genzel , r. , herrmann , f. , jackson , j. , poglitsch , a. , stacey , g. j. , & townes , c. h. 1993 , , 407 , 579 malhotra , s. et al .
2001 , , 561 , 766 mao , r. q. , henkel , c. , schulz , a. , zielinsky , m. , mauersberger , r. , strzer , h. , wilson , t. l. , & gensheimer , p. 2000 , , 358 , 433 mauersberger , r. , henkel , c. , wielebinski , r. , wiklind , t. , & reuter , h. -p .
1996 , , 305 , 421 moorwood , a. f. m. , & oliva , e. 1990 , , 239 , 78 mouri , h. & taniguchi , y. 1992 , , 386 , 68 mouri , h. 1994 , , 427 , 777 nakai , n. , hayashi , m. , handa , t. , sofue , y. , & hasegawa , t. 1987 , , 39 , 685 negishi , t. , onaka , t. , chan , k. -w .
, & roellig , t. l. 2001 , , 375 , 566 nilson , p. 1973 , uppsala astron .
, 6 paglione , t. a. , et al .
2001 , , 135 , 183 pak , s. , jaffe , d. t. , & keller , l. d. 1996a , , 457 , l43 pak , s. , jaffe , d. t. , & keller , l. d. 1996b , in asp conf .
102 , the galactic center , ed .
r. gredel ( san francisco : ast ) , 28 pak , s. , jaffe , d. t. , van dishoeck , e. f. , johansson , l. e. b. , & booth , r. s. 1998 , , 498 , 735 pak , s. , yu , y. , lee , d .- h . , min , k .- w . , & van dishoeck , e. f. 2003 , journal of the korean physical society , 42 , s88 pence , w. 1980 , , 239 , 54 puxley , p. j. , & hawarden , t. g. , & mountain , c. m. 1988 , , 234 , 29 puxley , p. j. , & hawarden , t. g. , & mountain , c. m. 1990 , , 364 , 77 puxley , p. j. 1991 , , 249 , 11p puxley , p. j. , ramsay howat , s. k. , & mountain , c. m. 2000 , , 529 , 224 richter , m. j. , graham , j. r. , & wright , g. s. 1995 , , 454 , 277 rieke , g. h. , lebofsky , m. j. , thompson , r. i. , low , f. j. , & tokunaga , a. t. 1980 , , 238 , 24 rogstad , d. h. , & shostak , g. s. 1972 , , 176 , 315 satyapal , s. , watson , d. m. , pipher , j. l. , forrest , w. j. , greenhouse , m. a. , smith , h. a. , fischer , j. , & woodward , c. e. 1997 , , 483 , 148 schmidt , m. 1959 , , 129 , 243 shen , j. , & lo , k. y. 1995 , , 445 , l99 shirley , y.l . , evans , n.j .
ii , young , k.e . , knez , k. , & jaffe , d.t .
2003 , , in press shull , j. m. , & beckwith , s. 1982 , , 20 , 163 smith , b. j. , & harvey , p. m. 1996 , , 468 , 139 smith , m. d. 1991 , , 252 , 378 smith , m. d. 1994 , , 266 , 238 sofue , y. , doi , m. , & ishizuki , s. 1988 , , 40 , 511 sohn , j. , ann , h. b. , pak , s. , & lee , h. m. 2001 , journal of the korean astronomical society , 34 , 17 solomon , p.m. , downes , d. , and radford , s.j.e .
1992 , , 387 , l55 stacey , g. j. , townes , c. h. , poglitsch , a. , madden , s. c. , jackson , j. m. , herrmann , f. , genzel , r. , & geis , n. 1991 , , 382 , l37 steppe , h. , mauersberger , r. , schulz , a. , & baars , j. w. m. 1990 , , 233 , 410 sternberg , a. 1989 , , 347 , 863 sternberg , a. , & dalgarno , a. 1989 , , 338 , 1997 sugai , h. , davies , r. i. , malkan , m. a. , mclean , i. s. , usuda , t. , & ward , m. j. 1999 , , 527 , 778 telesco , c. m. 1988 , , 26 , 343 thronson , h. a. , walker , c. k. , walker , c. e. , & maloney , p. 1987 , , 318 , 645 tielens , a. g. g. m. , & hollenbach , d. j. 1985 , , 291 , 722 turner , j. l. , & ho , p. t. p. 1983
, , 268 , l79 wright , e. l. 1991 , , 381 , 200 young , j. s. , xie , s. , kenney , j. p. d. , & rice , w. l. 1989 , , 70 , 699 xie , s. , young , j. , & schloerb , f. p. 1994
, , 421 , 434 lccccccc m82 & 09 55 52.6 & @xmath179 40 47 & 1 & 65 & 2 & 3.6 & 3 + ngc 253 & 00 47 33.1 & @xmath180 17 16 & 4 & 64 & 5 & 2.5 & 6 + ngc 6946 & 20 34 52.2 & @xmath181 09 14 & 7 & 62 & 8 & 5.5 & 9 + ic 342 & 03 46 48.1 & @xmath182 05 46 & 10 & 14 & 11 & 1.8 & 12 + lcccccccl m82 & 2.12183 & 1995 dec & palomar & 16 & 100 & 156 & 43.8 & hr 8505 + & & 1996 sep & palomar & 11 & 105 & 164 & 35.0 & hr 7924 + & 2.24771 & 1996 sep & palomar & 11 & 111 & 173 & 35.0 & hr 8781 + ngc 253 & 2.12183 & 1995 oct & ctio & 81 & 125 & 192 & 43.8 & hr 6084 , 7525 , 8728 + & & 1996 sep & palomar & 11 & 105 & 164 & 35.0 & hr 7924 + & 2.24771 & 1995 oct & ctio & 81 & 132 & 203 & 35.0 & hr 6084 + & & 1996 sep & palomar & 11 & 111 & 173 & 35.0 & hr 8781 + & 2.20139 & 1995 oct & ctio & 81 & 129 & 198 & 43.8 & hr 6084 , 8728 + ngc 6946 & 2.12183 & 1996 sep & palomar & 11 & 105 & 164 & 35.0 & hr 7924 + & 2.24771 & 1996 sep & palomar & 11 & 111 & 173 & 35.0 & hr 8781 + ic 342 & 2.12183 & 1995 dec & palomar & 16 & 100 & 156 & 25.0 & hr 8505 + lrrrrrrrr m82 & 11.1 & 194 & 14 & 6 & @xmath183 & @xmath184 & @xmath185 & + & & & 0 & 0 & @xmath183 & @xmath186 & @xmath187 & + & 15.7 & 274 & 41 & 19 & @xmath188 & @xmath189 & & + & & & 27 & 13 & @xmath188 & @xmath190 & & + & & & 14 & 6 & @xmath183 & @xmath191 & & + & & & 0 & 0 & @xmath183 & @xmath192 & & + & & & -14 & -6 & @xmath193 & @xmath194 & & + & & & -27 & -13 & @xmath193 & @xmath195 & & + & & & -41 & -19 & @xmath193 & @xmath196 & & + + ngc 253 & 11.1 & 135 & 45 & 22 & @xmath197 & @xmath198 & & + & & & 36 & 18 & @xmath197 & @xmath199 & & + & & & 27 & 13 & @xmath197 & @xmath200 & & + & & & 18 & 9 & @xmath197 & @xmath201 & & + & & & 9 & 4 & @xmath197 & @xmath202 & & + & & & 0 & 0 & @xmath203 & @xmath204 & @xmath205 & + & & & -9 & -4 & @xmath206 & @xmath207 & & + & & & -18 & -9 & @xmath206 & @xmath208 & & + & & & -27 & -13 & @xmath206 & @xmath209 & & + & & & -36 & -18 & @xmath206 & @xmath210 & & + & & & -45 & -22 & @xmath206 & @xmath211 & & + & 81 & 982 & 0 & 0 & @xmath203 & @xmath212 & @xmath213 & @xmath214 + + ngc 6946 & 11.1 & 296 & 0 & 0 & @xmath215 & @xmath216 & @xmath217 & + + ic 342 & 15.7 & 137 & 5 & 15 & @xmath218 & @xmath219 & & + & & & 0 & 0 & @xmath218 & @xmath220 & & + & & & -5 & -15 & @xmath218 & @xmath221 & & + lcccccc model & & & & 0.46 & 0.56 & 0.18 + _ thermal _ model & & & & 0.21 & 0.082 & 0.0057 + + m82 & @xmath222 & 31.6 & 9.6 & @xmath223 & @xmath224 & + & & 26.3 & 8.0 & @xmath225 & @xmath226 & + & & 21.0 & 6.4 & @xmath227 & @xmath228 & + & & 15.8 & 4.8 & @xmath229 & @xmath230 & + & & 10.5 & 3.2 & @xmath231 & @xmath232 & + & & 5.3 & 1.6 & @xmath233 & @xmath234 & + & & 0.0 & 0.0 & @xmath235 & @xmath236 & + & & -5.3 & -1.6 & @xmath237 & @xmath238 & + & & -10.5 & -3.2 & @xmath239 & @xmath240 & + & & -15.8 & -4.8 & @xmath241 & @xmath242 & + & & -21.0 & -6.4 & @xmath243 & @xmath244 & + & & -26.3 & -8.0 & @xmath245 & @xmath246 & + & & -31.6 & -9.6 & @xmath247 & @xmath248 & + & 11.1 & 14 & 6 & & @xmath249 & + & & 0 & 0 & & @xmath250 & + + ngc 253 & 11.1 & 0 & 0 & & @xmath251 & + & 81 & 0 & 0 & & @xmath252 & @xmath253 + + ngc 6946 & 11.1 & 0 & 0 & & @xmath254 & + lclclclcl m82 & 86.9 & 1 & 82.1 & 2,3,4,5,6,7 & 128 & 4 & 169 & 4 + ngc 253 & 69.9 & 1 & 64.7 & 4,5,6,7 & 51.9 & 4 & 37.6 & 4 + ngc 6946 & 3.74 & 1 & 3.68 & 3,4,5,6,7 & 10.3 & 4 & 5.90 & 4 + ic 342 & 3.68 & 1 & 12.2 & 2,3,5,7 & 38.7 & 8 & 32.6 & 8 + | the central regions of many interacting and early - type spiral galaxies are actively forming stars .
this process affects the physical and chemical properties of the local interstellar medium as well as the evolution of the galaxies .
we observed near - infrared h@xmath0 emission lines : @xmath1 , , , and from the central @xmath2 kpc regions of the archetypical starburst galaxies , m82 and ngc 253 , and the less dramatic but still vigorously star - forming galaxies , ngc 6946 and ic 342 . like the far - infrared continuum luminosity
, the near - infrared h@xmath0 emission luminosity can directly trace the amount of star formation activity because the h@xmath0 emission lines arise from the interaction between hot and young stars and nearby neutral clouds .
the observed h@xmath0 line ratios show that both thermal and non - thermal excittion are responsible for the emission lines , but that the great majority of the near - infrared h@xmath0 line emission in these galaxies arises from energy states excited by ultraviolet fluorescence .
the derived physical conditions , e.g. , far - ultraviolet radiation field and gas density , from ] and ] lines and far - infrared continuum observations when used as inputs to photodissociation models , also explain the luminosity of the observed h@xmath0 line .
the ratio of the h@xmath0 line to far - ir continuum luminosity is remarkably constant over a broad range of galaxy luminosities ; @xmath3 @xmath4 , in normal late - type galaxies ( including the galactic center ) , in nearby starburst galaxies , and in luminous ir galaxies ( lirgs : @xmath5 @xmath6 @xmath7 @xmath8 ) . examining this constant ratio in the context of photodissociation region models
, we conclude that it implies that the strength of the incident uv field on typical molecular clouds follows the gas density at the cloud surface . |
the distribution function ( df ) provides the most general and complete way of statistical description of dark matter ( dm ) haloes .
it carries maximum information on the spatial and velocity distributions of particles in such objects .
our knowledge on the df is still being improved , mostly due to numerical experiments . in the last few years cosmological simulations
have revealed increasingly detailed features of phase - space structure of dm haloes .
these numerical results provide useful constraints on theoretical models of the df .
one property of interest in this field is the anisotropy of the velocity dispersion tensor .
it has been demonstrated that the outer parts of the haloes exhibit more radially anisotropic trajectories than the halo centre ( see e.g. coln , klypin & kravtsov 2000 ; fukushige & makino 2001 ; wojtak et al .
2005 ; mamon & okas 2005 ; cuesta et al .
this feature , besides the well - studied density profile , has been considered as the main point of reference in the attempts at construction of a reliable model of the df .
so far , a few approaches to this problem have been proposed .
cuddeford ( 1991 ) generalized the osipkov - merritt model ( osipkov 1979 ; merritt 1985 ) to the df which generates an arbitrary anisotropy in the halo centre and becomes fully radial at infinity .
although an analytical inversion for these models exists , the anisotropy profile can not be reconciled with the numerical results : the rise from central to outer anisotropy is too sharp and the outer orbits are too radial ( see mamon & okas 2005 ) .
an & evans ( 2006a ) noticed that a non - trivial profile of the anisotropy can be obtained from a sum of dfs with a constant anisotropy for which an analytical inversion is known ( cuddeford 1991 ; kochanek 1996 ; wilkinson & evans 1999 ) .
however , the resulting anisotropy profiles are decreasing functions of radius and do not agree with those measured in cosmological simulations . recently a very elegant method has been presented by baes & van hese ( 2007 ) .
the authors introduced a general ansatz for the anisotropy profile and then , for a given potential - density pair , derived the df as a series of some special functions .
this approach works well under the condition that the potential can be expressed as an elementary function of the corresponding density .
this requirement , however , is not satisfied by many models , including the nfw density profile ( navarro , frenk & white 1997 ) which is commonly used as a good approximation of the universal density profile of dm haloes .
the df inferred from the simulation gives a possibility to test directly the analytical models .
according to the jeans theorem any spherically symmetric system in the state of equilibrium should possess a df which is a function only of energy and the absolute value of angular momentum .
this theoretical postulate was taken into account in the computation carried out by voglis ( 1994 ) and natarajan , hjorth & van kampen ( 1997 ) . in the first case
the df was obtained for a single relaxed halo which formed from cosmologically consistent initial conditions .
it was shown that there were two main contributions to the df , the halo population and the core population of particles .
both were effectively described by two independent phenomenological fits .
natarajan et al .
( 1997 ) determined the df for a sample of cluster - size haloes formed in cosmological simulations .
their selection of objects included those with substructures and departing from equilibrium .
they also discussed and took into account in their calculation the effect of boundary conditions defined by the virial sphere .
however , the final results were not used to test quantitatively any model of the df .
it seems that two main approaches to study the df , namely theoretical modelling and feedback from the simulations , evolved rather separately barely crossing each other .
the rare exceptions include the work of okas & mamon ( 2001 ) who used the eddington formula to derive numerically the df following from the nfw profile in the isotropic case and that of widrow ( 2000 ) who considered more general cuspy profiles and osipkov - merritt anisotropy . this paper is devoted to combining both approaches and providing a coherent analysis of the df from the viewpoints of the simulations as well as the model construction .
our main aim is to propose a phenomenological model of the df that recovers the results from the simulations as accurately as possible .
our effort is mainly motivated by the future applications of the derived df to the dynamical modelling of galaxy clusters .
although subhaloes in general have different density and velocity distributions than dm particles ( diemand , moore & stadel 2004 ) , massive subhaloes ( those likely to host galaxies ) are distributed like dm particles ( faltenbacher & diemand 2006 ) . although the correspondence between the massive subhaloes and galaxies in real clusters remains to be proven , our results should at least in principle
be applicable to kinematic data sets for galaxy clusters .
the traditional approach to do such modelling was to reproduce the velocity dispersion profile of the galaxies by solving the jeans equation ( see e.g. katgert , biviano & mazure 2004 ) .
it is well known however that from the dispersion alone one can not constrain all the interesting parameters ( such as the virial mass , the concentration of the nfw profile and anisotropy ) because of the density - anisotropy degeneracy .
one can break this degeneracy by using the fourth order velocity moment , the kurtosis , and solving an additional higher - order jeans equation ( okas 2002 ; okas & mamon 2003 ; okas et al . 2006 ; wojtak & okas 2007 ) .
although this approach has many advantages ( e.g. it does not require the knowledge of the full df ) , it has been applied till now only for constant - anisotropy models and the calculation of velocity moments requires the binning of the data in which some information is lost . since the number of galaxies with measured redshifts per cluster is still rather low ( of the order of a few hundred for the best - studied , nearby clusters ) it is essential that all the available information is used .
this can be obtained by fitting the projected df to the data directly .
a few approaches along these lines have been attempted already .
for example , mahdavi & geller ( 2004 ) used a simple df of the form @xmath11 ( which yields constant anisotropy ) to constrain the mass profile and orbital structure using combined kinematic data sets for nearby galaxy groups and clusters , while van der marel et al .
( 2000 ) in their study of cnoc clusters did not assume an explicit form for the energy - dependent part of the df , but still used the constant anisotropy .
wojtak et al .
( 2007 ) used a simplified form of the projected isotropic df constructed from the projected density combined with a gaussian distribution for the line - of - sight velocities to study the properties of members versus interlopers in simulated kinematic data sets .
none of the dfs used so far , however , reflects accurately the true properties of cluster - size dm haloes found in @xmath12-body simulations .
the paper is organized as follows .
section 2 provides the theoretical framework and defines all the basic quantities used later on in the paper . in the next section
we discuss the details of the computation of the df of dm particles in the haloes formed from cosmological simulations and provide examples of the results .
section 4 is devoted to the derivation of a phenomenological model of the df ; we discuss the separability of the df in energy and angular momentum and present an explicit formula for the @xmath4-dependent part of df .
an extensive comparison of the model with the simulations is presented in section 5 , where we also provide an analytical approximation for the energy - dependent part of the df obtained for an average halo .
finally , the discussion follows in section 6 .
this section summarizes the theoretical background of the paper .
first , we introduce scaling properties consistent with the nfw density profile .
we will use this profile in the paper , but our approach is not restricted to this particular density distribution and can be easily generalized to any profile consistent with simulations ( see below ) .
second , we briefly describe the relation between the differential df and the df itself . finally , we discuss the consequences of the finite volume of the virialized area of a halo .
in particular , the relation between the df and its differential form is properly modified to account for this effect .
it is a well known fact that the density profiles of dm haloes formed in cosmological simulations exhibit striking similarity .
nfw showed that most of them are well fitted within the virial sphere by the universal two - parameter profile which can be expressed in the following way @xmath13 where @xmath14 .
the two free parameters are the scale radius @xmath15 and the mass enclosed within the sphere of this radius @xmath16 .
the ( positive ) gravitational potential inferred from the poisson equation reads ( cole & lacey 1996 ; see also okas & mamon 2001 ) @xmath17 where the velocity unit @xmath18 is related to the circular velocity @xmath19 at the scale radius via @xmath20 .
let us note that @xmath15 , @xmath18 and @xmath16 define a set of natural units of the nfw model . by scaling any quantity by a proper combination of them we remove the explicit dependence on the free parameters of the nfw model .
this is an essential property if we want to study the dynamics of a whole class of haloes with nfw - like density profiles .
hereafter , we will keep this scaling in all equations in the text . in many places
we will also use a unit of the angular momentum @xmath21 as a substitute for @xmath22 .
the df is a fundamental concept in statistical mechanics of @xmath12-body systems .
it describes the phase - space density of particles of such a system without any detailed knowledge of the time evolution of @xmath12 trajectories . following the jeans theorem ,
a steady - state df , which is of interest for us here , depends on the phase - space coordinates only through the integrals of motion .
although the shape of dm haloes is in general better approximated by a three - axial ellipsoid rather than a sphere ( see e.g. gottlber & yepes 2007 ) , it is still very effective to assume spherical symmetry in dynamical approach . given that the streaming motions and internal rotation within the virial sphere are negligible compared to higher velocity moments , spherical symmetry implies that the df can be expressed as @xmath23 where @xmath3 is the positively defined binding energy and @xmath4 the absolute value of the angular momentum per unit mass @xmath24 the gravitational potential in equation ( [ e_def ] ) is related to the df through the poisson equation @xmath25 the most natural and straightforward probe of @xmath26 in numerical experiments is the so - called differential df defined in the following way @xmath27 one may intuitively interpret this function as mass density in energy - angular momentum space .
the df itself can be simply derived dividing @xmath28 by the volume @xmath29 of the hypersurface of constant energy and angular momentum embedded in the phase space @xmath30 it is easy to show that the volume of this hypersurface reads ( see appendix a ) @xmath31 where @xmath32 is the radial period of an orbit given by the following integral over radius from the pericentre to the apocentre @xmath33}}.\ ] ] the upper panel of fig .
[ g_vol ] shows a contour map of @xmath29 ( dotted lines ) calculated for the nfw gravitational potential ( [ psi_nfw ] ) .
the @xmath34 line is the profile of maximum angular momentum which consists of points corresponding to circular orbits .
this curve divides the energy - angular momentum plane into an area describing the physical orbits of a system ( below @xmath35 ) and the zone not permitted by mechanics ( above @xmath35 ) .
note that we are using the scaling relations introduced in the previous subsection so the results do not depend explicitly on the halo mass @xmath16 and the scale radius @xmath15 . in some places later on we will refer to the inverse function for @xmath34 by @xmath36 .
voglis ( 1994 ) and natarajan et al . (
1997 ) showed that the dependence of @xmath32 on the angular momentum is very weak and could be neglected without loss of precision .
this is understandable if we note that the nfw - like potentials are still not so far away from the isochrone potential @xmath37 which leads to purely energy - dependent @xmath38 proportional to @xmath39 ( binney & tremaine 1987 ) .
following natarajan et al .
( 1997 ) we will use this feature to simplify expression ( [ g(e , l ) ] ) . to do this
we first note that the volume of the hypersurface of constant energy @xmath40 is given by @xmath41 taking advantage of the weak dependence of @xmath42 on @xmath4 , equation ( [ t_r ] ) , we get @xmath43 on the other hand , one can show that @xmath44 reads ( see appendix a ) @xmath45 where @xmath46 is the apocentre radius of the radial orbit . inserting ( [ g_e_gen ] ) into ( [ g_app ] ) one immediately gets a very simple approximation for @xmath29 involving only a one - dimensional integral without singularities , in contrast with @xmath47 derived by expression ( [ t_r ] ) .
we find that this approximation reproduces the exact formula ( [ g(e , l ) ] ) with enough accuracy .
taking advantage of its numerical simplicity we use it in majority of our calculations .
so far we have discussed the relation between the df and its differential form for an infinite system . in practice , however , we restrict our numerical analysis to the interior of the virial sphere which separates the equilibrium part of a halo from the infall region .
we define the virial radius @xmath48 of this sphere by @xmath49 where @xmath50 is the virial mass , @xmath51 is the present critical density and @xmath52 is the virial overdensity .
another parameter commonly used to describe the size of the virial sphere in terms of @xmath15 is the concentration @xmath53 .
the existence of the boundary of the virialized part of the halo implies that @xmath38 given by ( [ t_r ] ) must be replaced by @xmath54}},\ ] ] where the upper limit of the integral is a minimum of the virial radius and the radius at the apocentre ( see appendix a for details ) . combining ( [ t_r_trun ] ) with ( [ g(e , l ) ] ) one gets a general formula for the volume @xmath29 in the presence of a spherical boundary of a halo .
contrary to the conclusion of natarajan et al .
( 1997 ) , we find that the approximation ( [ g_app ] ) is no longer justified for orbits extending beyond the virial sphere ( @xmath55 ) .
this follows from the fact that angular momentum dependence of ( [ t_r_trun ] ) becomes non - negligible and the integral ( [ g_e ] ) can not be simplified to the form of ( [ g_app ] ) . using ( [ g(e , l ) ] ) and
( [ t_r_trun ] ) with the nfw potential , we calculated @xmath29 for a halo limited by the virial sphere of radius @xmath56 ( see the lower panel of fig .
[ g_vol ] ) .
as expected , the result differs from an infinite system by the orbits with @xmath55 and remains unchanged for trajectories wholly included within the virial sphere .
for our @xmath12-body simulation we have assumed the wmap3 cosmology ( spergel et al .
2007 ) with matter density @xmath57 , the cosmological constant @xmath58 , the dimensionless hubble parameter @xmath59 , the spectral index of primordial density perturbations @xmath60 and the normalization of power spectrum @xmath61 .
we have used a box of size @xmath62 and @xmath63 particles . thus the particle mass was @xmath64 .
starting at redshift @xmath65 we followed the evolution using the mpi version of the adaptive refinement tree ( art ) code ( kravtsov , klypin & khokhlov 1997 ) .
we identified clusters with the hierarchical friends - of - friends ( fof ) algorithm ( klypin et al .
1999 ) with a linking length of 0.17 times the mean inter - particle distance which roughly corresponds to an overdensity of 330 .
we have selected 36 clusters at redshift @xmath66 in the range of virial mass ( 0.15 - 2 ) @xmath67 , where the virial overdensity parameter appropriate for our cosmological model was assumed @xmath68 ( okas & hoffman 2001 ) .
our sample did not include clusters with two substructures of approximately the same mass and a poor fit of the nfw profile suggestive of a recent major merger .
starting from the fof position of the cluster we have determined the highest density peak as the final centre of the clusters .
this centre coincides with the position of the most massive substructure found at the linking length 8 times shorter and also with the position of the halo found by the bdm halo finder ( klypin et al .
1999 ) . in the first step of the computation
we calculate the binding energy ( [ e_def ] ) and the angular momentum ( [ l_def ] ) per unit mass for each particle within the virial sphere of each halo .
spherical symmetry implies that we have to apply in ( [ e_def ] ) the radial profile of the gravitational potential @xmath69 where @xmath70 .
however , the mass profile of the equilibrium part of a halo reaches no further than the virial radius . on the other hand , all analytical models of the df
involve the density profile extending to infinity .
we found that the only coherent way to reconcile both facts is to split the integral ( [ pot_num_0 ] ) in two parts @xmath71 the first term is evaluated numerically by integration of a discrete mass profile .
the second term is an analytical extension with the nfw density profile which is an assumption of the df model introduced in the following section .
its contribution to the potential is a constant equal to @xmath72 .
[ halo_1_sc ] shows the resulting energies and angular momenta of particles inside the virial sphere of one of the simulated haloes .
the profile of the maximum angular momentum ( solid line ) and the profile of vanishing radial velocity at the virial sphere ( dashed line ) were calculated for the exact gravitational potential given by ( [ pot_num_1 ] ) .
all particles occupy the area permitted by mechanics or lie very close to the boundary line . interestingly ,
quite a large fraction of them have orbits extending beyond the virial sphere .
as noted in the previous section , we keep @xmath73 and @xmath21 as units of energy and angular momentum respectively .
the parameters of the nfw model were obtained for each halo by fitting the nfw formula to the density profile measured in logarithmic radial bins . in the next step
we determine for each halo the differential df given by ( [ n(e , l ) ] ) . in this calculation
we used our own version of the fiestas ( field estimator for arbitrary spaces ) algorithm designed to infer the density field from a scatter diagram embedded in a space of any number of dimensions ( see ascasibar & binney 2005 for more details ) . as a result of this computation
we get an estimate of @xmath28 at all points of the energy - angular momentum plane corresponding to the particles inside the virial sphere .
once @xmath28 is calculated the df can be easily obtained via ( [ n(e , l ) ] ) .
as discussed in section 2 , we used approximation ( [ g_app ] ) for the orbits contained inside the virial sphere and the exact formula ( [ g(e , l ) ] ) with ( [ t_r_trun ] ) for trajectories extending beyond @xmath48 .
we found that the additional advantage of expression ( [ g_app ] ) is that it could be evaluated at any point of the energy - angular momentum plane .
this helps us to keep the estimates of the df obtained for points with angular momentum lying slightly above @xmath34 . in order to derive a contour map or a profile of the df
we introduce a regular dense mesh on the energy - angular momentum plane and find the median value of the df in each cell .
such a set of median points is considered as the final numerical approximation of the df and is used in preparation of all plots in this paper .
[ halo_1_df ] shows two examples of the resulting contour maps obtained for two different haloes .
the unit of the df in this and following figures is @xmath74 .
the interval between the iso - df lines is fixed at value @xmath75 of the logarithmic scale .
the lack of the df estimation in the lower part of each diagram arises from the fact that this zone is occupied by very few particles ( see e.g. fig . [ halo_1_sc ] ) so that no information on the distribution can be retrieved .
let us note that this is an effect of the finite mass resolution .
a general form of the df for spherical systems in the state of equilibrium is a function of energy and the absolute value of angular momentum @xmath26 . in our approach
we assume that the df is separable in energy and angular momentum @xmath76 this is the first assumption that considerably narrows the family of possible solutions .
therefore , it is necessary to check how robust it is .
we address this problem in the next section , where we present an extensive comparison of the analytical model with the simulations .
the angular momentum part of the df in equation ( [ df_mod_gen ] ) specifies the anisotropy of velocity dispersion tensor .
this quantity is commonly described with the so - called anisotropy parameter @xmath77 where @xmath78 and @xmath79 are the radial and the tangential velocity dispersions respectively and we assume there are no streaming motions .
the values of this parameter range from @xmath80 for circular orbits to @xmath81 for purely radial trajectories . fig .
[ beta_sing ] shows the average anisotropy profile of the simulated haloes used for the measurement of the df .
the light gray rectangle in the background of the plot indicates the position of the virial radius .
it is clearly seen that the anisotropy is typically a growing function of radius , with values @xmath82 in the halo centre and @xmath83 at the virial sphere ( see e.g. mamon & okas 2005 and cuesta et al .
2007 for comparison ) . on the other hand
, the considerable width of the interquartile range of the measured @xmath84 ( dark gray region ) signifies that the profiles of single haloes differ among each other . occasionally flat or decreasing profiles
are measured .
it seems that a simple and general enough analytical model of the anisotropy should possess at least three free parameters which determine asymptotic values of @xmath84 for small and large radii and a scale of transition between them .
we proceed with the construction of such a model by introducing a proper ansatz for @xmath85 .
louis ( 1993 ) showed that the following asymptotes of the angular momentum part of the df @xmath86 where @xmath8 is an angular momentum constant , lead to constant anisotropy @xmath7 at infinity ( @xmath87 ) and @xmath88 in the halo centre .
this result can be easily generalized to the case of a non - isotropic velocity distribution in both limits of radius .
first , let us note that the central part of the halo is dominated by the particles with small angular momenta , namely @xmath89 . then , remembering that the df of constant anisotropy takes the form ( hnon 1973 ; binney & tremaine 1987 ; okas 2002 ) @xmath90 it is easy to notice that the formula ( [ f_l_louis ] ) can be rewritten in the following way @xmath91 where @xmath6 is the central anisotropy of a system . as shown by an & evans ( 2006b ) ,
the upper limit for @xmath6 is equal to @xmath92 , where @xmath93 is the density profile near the halo centre .
this means that for the nfw density model we have @xmath94 .
the simplest function obeying the asymptotic conditions formulated above is a double power - law function @xmath95 as shown in the following section , this ansatz leads to a very realistic anisotropy profile that fits well the @xmath84 profiles of simulated haloes .
furthermore , the simplicity of formula ( [ f_l ] ) guarantees that the energy part of the df can be quite easily calculated via the inversion of the integral equation @xmath96 the key idea of this procedure lies in an analytical simplification of the right - hand side of ( [ rho_df ] ) to a one - dimensional integral .
the resulting equation is then solved numerically for @xmath9 .
the technical details of this calculation are summarized in appendix b. once the full form of the df is determined one can also calculate the velocity moments .
all formulae are reduced to one - dimensional integrals which can be easily evaluated numerically ( see appendix c ) .
= 17.5 cm the top row of fig .
[ df_th ] shows the anisotropy , dispersion @xmath78 and kurtosis @xmath97 of the radial velocity inferred from the model of the df .
the calculations were carried out assuming the nfw density profile and four sets of model parameters chosen to illustrate the flexibility of the model : @xmath98 and @xmath99 ( solid and dashed lines respectively ) ; @xmath100 ( dotted line ) ; @xmath101 and @xmath102 ( dashed - dotted line ) . in all cases the transition value of @xmath103 was used .
the dispersion profiles for the two models with increasing @xmath84 , as expected , differ only for large radii which is the effect of different values of @xmath7 .
interestingly , the corresponding kurtosis profiles clearly signify flat - topped velocity distribution in the outer part of the halo ( @xmath104 ) , highly peaked distribution in the centre ( @xmath105 ) and roughly gaussian for radii around @xmath15 ( @xmath106 ) . on the other hand ,
non - increasing @xmath84 profiles lead to less peaked velocity distributions in the centre .
it seems therefore that the typical anisotropy of dm haloes , as shown in fig .
[ beta_sing ] , is expected to coincide with the kurtosis rapidly growing towards the halo centre ( see also fig . [ 4haloes ] below ) . as we will see in the following section , this is one of the most characteristic features of the phase - space structure of massive dm haloes . in the bottom panels of fig .
[ df_th ] we plotted the dfs corresponding to four sets of model parameters , as described above .
the three panels from the left to the right show the energy part of the df , contour maps and the profiles for three fixed values of angular momentum .
the plots reveal some interesting signatures of the specific shape of @xmath84 profile . for example
, the inclination of the iso - df lines with respect to the energy axis decreases with increasing @xmath6 : more isotropic @xmath107 at the centre corresponds to more vertical iso - df lines ; also the shape of the lines is somewhat different .
these features are also to some extent visible in the contour maps of the df for two simulated haloes in fig .
[ halo_1_df ] .
the upper map represents a halo with an increasing @xmath84 , whereas the second one depicts the case of a decreasing @xmath84 profile .
both haloes are analyzed in terms of velocity moments and the df in the following section .
recently mamon & okas ( 2005 ) found that the simulation data are well reproduced by the anisotropy profile of the form @xmath108 where @xmath109 is the radius where @xmath110 . assuming @xmath111 and @xmath112 in our df model , we made a comparison of the resulting anisotropy with the functional form ( [ ml_beta ] ) .
[ ml_beta_f ] shows both anisotropies for three values of @xmath109 .
note that both @xmath84 profiles have similar shapes , although our anisotropy profile has a somewhat sharper rise at small radii .
we also find that the radius @xmath113 characteristic of the df model , for which @xmath107 is the mean of the limiting values @xmath114 depends weakly on @xmath6 and @xmath7 . for parameter ranges leading to @xmath84 profiles covering the interquartile area of anisotropy from the simulation ( @xmath115 , @xmath116 and @xmath117 ) ,
this radius is well ( within 5 percent accuracy ) approximated by @xmath118
the df proposed in the previous section is a phenomenological model in the sense that it possesses free parameters whose values should be adjusted to the simulation data .
all three parameters were introduced to determine a family of anisotropy profiles so that it is @xmath84 that is most sensitive to the variations of @xmath6 , @xmath7 and @xmath8 .
consequently , we decided to constrain the parameters of the model by fitting the @xmath84 profile inferred from the df model to the median profile measured in simulated dm haloes .
the best - fitting parameters are : @xmath119 , @xmath120 and @xmath121 .
the corresponding best - fitting profile of the anisotropy is plotted as a dashed line in the lower left panel of fig .
[ moments ] .
once the model parameters are adjusted the df can be compared with its counterpart measured from the simulation .
[ df_comp ] shows this comparison in terms of a contour map and the profiles for constant angular momentum or energy .
dark gray regions in all panels indicate the interquartile range of the df values within the halo sample .
the lighter gray area in the background of the upper diagram marks the points of vanishing radial velocity at the virial radius @xmath48 .
its boundaries are fixed by the first and third quartile of virial radii in the halo sample , @xmath122 and @xmath123 respectively .
= 7.8 cm although some deviations of the model ( dashed lines ) from the results of the simulations are visible , in general the theoretical profiles are included within the interquartile range or lie very close to its boundaries .
as expected , the strongest discrepancy between the model and the simulation is present in the part of the energy - angular momentum plane populated by the particles with orbits extending beyond the virial sphere ( the area to the left of the @xmath124 line ) .
however , given that this is the only part of the energy - angular momentum plane affected by the infalling material , we think that the observed differences are acceptable . a critical point of the derivation of the df presented in the previous section was the factorization introduced by equation ( [ df_mod_gen ] ) . in order to inspect the robustness of this assumption we propose a simple test .
we calculate the ratio of the df from the simulation to the energy part of the df model with parameters adjusted to the anisotropy profile from the simulation . under the assumption that the real df is factorizable in energy and angular momentum
, we can expect that the resulting ratio should be a weak function of energy equal to @xmath85 given by ( [ f_l ] ) .
[ separability ] shows that the variations of this ratio with respect to @xmath85 are of the same order as the width of the interquartile range which means that separability is acceptable from the statistical point of view .
a small systematic deviation can be seen for @xmath125 . however , this is certainly a local feature since this trend is not repeated in other profiles .
let us emphasize that this test of separability depends strongly on the reliability of @xmath85 .
one can imagine that an incorrect form of @xmath85 would likely lead to a negative result of the test , whether @xmath26 is separable or not .
on the contrary , a positive result of such a test in our case means that not only is the assumption of factorization valid but the approximation for @xmath85 is reasonable as well .
further comparison between the simulation and the df model can be done in terms of velocity moments .
this is depicted in fig .
[ moments ] where the dispersion and kurtosis of the radial and tangential velocity are plotted . in the bottom part of this figure
we show the profiles of the anisotropy @xmath84 and @xmath126 parameter which measures the anisotropy of a tensor of the fourth velocity moment . by analogy with the parameter @xmath84 we defined @xmath127 in the following way @xmath128 the dashed lines in each panel of fig .
[ moments ] are the model predictions , except for the @xmath84 profile ( lower left panel ) which is a fit of the model providing constraints on parameter values given in the previous subsection .
theoretical dispersion profiles coincide very well with the profiles from the simulation .
we notice quite a good agreement also for the @xmath127 parameter . on the other hand , theoretical profiles of the kurtosis
are systematically biased towards higher values , but typically by less than 10 percent .
nevertheless , their shapes clearly recover the shapes of the median profiles from the simulation
. moreover , for both the radial and tangential velocity a characteristic growth of @xmath129 from value @xmath130 around the virial radius up to @xmath131 in the halo centre , is seen .
although the kurtosis bias is enclosed within acceptable limits ( kurtosis is known to be sensitive to any noise ) , it would be desirable to find out the reason for this behaviour .
since our statistical samples consist of @xmath132 particles per radial bin , we ruled out a possibility of a bias of the kurtosis estimator ( see okas & mamon 2003 ) .
we also excluded the possibility that this is caused by some specific assumptions of the model .
for example , changing the nfw density distribution to the 3d sersic profile , which fits the simulation data even better ( navarro et al .
2004 ; merritt et al . 2005 ; prada et al .
2006 ) , we still encounter the same bias . in addition , perturbing the model parameters of @xmath85 does not explain the situation either .
we therefore conclude that the slight discrepancy in the predictions of our model concerning the kurtosis must signify reaching the limitations of the theoretical approach based on using the global , smooth gravitational potential of a system .
we suppose that the problem is caused by the presence of substructures which perturb locally the trajectories of particles with respect to the orbits determined by the global potential of a halo .
what one gets from the simulation is really a convolution of the velocity distribution expected from the model involving a global potential with the distribution of velocity perturbations occurring due to density fluctuations .
the estimation of the importance of this effect is a complicated task since the perturbation of the particle orbit depends on many variables , such as the distribution of substructures , softening of the potential and particle velocity .
however , some qualitative conclusions can be drawn .
first , note that low - velocity particles are affected by the density perturbations more strongly .
consequently , the peak of the resulting velocity distribution is suppressed and the tails are preserved which may effectively decrease the kurtosis ( see fig .
[ moments ] ) .
second , the effect of the perturbation on the velocity dispersion is a higher order correction compared to the dispersion obtained for a system with a global potential .
this means that the resulting dispersion profiles are barely changed and they are still expected to coincide well with theoretical predictions .
it seems an intriguing issue that the profile of tangential kurtosis signifies gaussianity of the velocity distribution at radii around @xmath15 where the logarithmic slope of the density profile is equal to @xmath133 .
one could suppose that some signatures of the so - called isothermal sphere are locally present .
interestingly , this statement is also supported by the shape of the df for @xmath134 that is the energy range of particles at @xmath135 . referring to the middle panel of fig .
[ df_comp ] it is easy to notice that the df grows exponentially with energy , as expected for systems not very different from the isothermal sphere .
the distribution of the radial velocity , on the other hand , takes the gaussian form for radii around @xmath136 .
this difference could be a consequence of the non - vanishing anisotropy parameter , which is not accounted for in the classical formulation of the isothermal model ( binney 1982 ; binney & tremaine 1987 ) .
so far we tested the df model for a typical halo associated with the median properties of our halo sample . in order to check
the applicability of our model more extensively we repeat such comparison for single haloes .
the df in this case is expected to differ from one halo to another due to the observed variety of anisotropy profiles .
results of this analysis are summarized in fig . [
4haloes ] . to save space we included only five haloes with representative , rather different anisotropy profiles ( upper panels ) , from the most strongly increasing profile in the left panel to a decreasing one on the right - hand side .
the second and fifth panels correspond to the haloes for which contour maps of the df are shown in fig .
[ halo_1_df ] ( the top and bottom panel respectively ) .
we restricted the number of profiles to those most essential : we plot the dispersion and kurtosis of the radial velocity and the anisotropies @xmath84 and @xmath127 .
we also show the profiles of the df for three values of angular momentum or energy . in all panels
the solid lines represent simulation results , whereas dashed lines are the predictions of the model . as in fig .
[ moments ] , dashed lines in the case of parameter @xmath84 indicate best fitting profiles of the model .
gray regions in the panels of two bottom rows mark the interquartile ranges of the df which describe the scatter of points resulting from the fiestas algorithm . from the analysis of fig .
[ 4haloes ] we conclude that all profiles , regardless of the anisotropy , are very well reproduced by our model of the df . in general
, the theoretical df does not exceed the limits of the interquartile range or lies very close to its boundaries ( see two bottom rows of panels in fig . [ 4haloes ] ) .
surprisingly , we find that the agreement is usually almost equally good when the model is applied to the haloes with massive substructures which were rejected from our sample .
this is certainly good news for the future applications of our df to the dynamical modelling of galaxy clusters which very often display signatures of recent major mergers .
.values of the parameters used in the approximation of the energy part of the df ( [ f_e_appr ] ) .
the first column lists the parameters .
the second column gives the parameter values for the model fitted to the anisotropy for the average halo from the simulation ( @xmath119 and @xmath120 ) and the third one gives the values which reproduce the df for the anisotropy profile ( [ ml_beta ] ) ( @xmath111 and @xmath112 ) . in both cases
@xmath121 was assumed .
[ cols="^ , < , < " , ] the df discussed in the first subsection is typical in a sense that it describes the statistical macrostate of dm particles in a typical massive halo . with future applications in mind we decided to provide an analytical approximation for the energy part @xmath9 of this df which could be used as a substitute for a rather complicated procedure described in appendix b. we found that the following expression reproduces the numerical df with good accuracy @xmath137}\nonumber \\ & & \times(1-e)^{-\alpha_{2}/\{1+\exp[(1-e)/e_{2})]\ } } \label{f_e_appr},\end{aligned}\ ] ] where the values of the parameters are listed in the middle column of table [ parameters ] . for completeness
we recall that the angular momentum part of the df is given by ( [ f_l ] ) with @xmath119 , @xmath120 and @xmath121 .
we have verified that the errors of the dispersion , kurtosis and both anisotropies , when using this approximate formula in the integrals for velocity moments , do not exceed 5 percent within the radial range @xmath138 ( see fig .
[ errors ] ) .
note that the general form of expression ( [ f_e_appr ] ) can be effectively used to approximate the df model also for other sets of parameters .
as a second example we include in the third column of the table also the parameters of a model with @xmath111 , @xmath112 and @xmath121 which mimics the anisotropy profile ( [ ml_beta ] ) with @xmath139 .
we have studied the df of dm particles inside the virial spheres of the haloes of mass @xmath0@xmath1 formed in the standard @xmath2cdm cosmological @xmath12-body simulation . in the first part of the paper we presented results of the calculation of the df from the simulation in the form most suitable for comparison with theoretical models .
then we proposed a phenomenological model of the df .
the model in its part dependent on angular momentum involves three free parameters which specify the anisotropy profile , namely its asymptotic values and the scale of transition between them .
we demonstrated that this parametrization is sufficient to reproduce accurately the simulation results in terms of velocity moments as well as the df itself .
the only discrepant point we encountered was a small but statistically significant bias of the theoretical kurtosis with respect to its profiles measured from the simulation .
this is probably caused by the presence of substructures perturbing the trajectories of low - velocity particles . in section 5
we showed that the velocity distribution of a typical halo changes from a flat - topped distribution ( @xmath140 ) in the outer part to a peaked one ( @xmath141 ) near the centre .
this behaviour was noticed and discussed by others before ( e.g. kazantzidis , magorrian & moore 2004 ; wojtak et al .
the analysis of the df presented here suggests that this property of the velocity distribution is correlated with the profile of the anisotropy increasing with @xmath142 : @xmath84 profiles growing faster with @xmath142 imply more rapid growth of the kurtosis towards the centre .
as demonstrated by taylor & navarro ( 2001 ) , the profile of the phase - space density @xmath143 in dm haloes is well fitted by a power - law function .
it seems that the status of this relation is as well established as the nfw fit of the density profile .
we checked that @xmath144 profiles inferred from the df model with parameters adjusted to the median anisotropy ( see section 5 ) coincide well with the corresponding power - law functions ( see fig .
[ q_profile ] ) . in this comparison
we assumed the logarithmic slopes of @xmath145 ( in the case of the dispersion of the radial velocity ) and @xmath146 ( in the case of the total dispersion ) , the values obtained from the simulations by dehnen & mclaughlin ( 2005 ) .
the relative residuals of both @xmath144 profiles are of the same order as the scatter of points from the simulations in fig . 1 of dehnen & mclaughlin ( 2005 ) .
note that this happens when the df model is tuned to the mean trend of the @xmath84 parameter .
therefore one could suspect that both relations , the mean profile of the anisotropy and @xmath147 , are two aspects of some deeper relation .
a more general parametrization of the df might provide some insights towards a more fundamental understanding of this phenomenon .
although the whole analysis presented in this paper was done in the framework of the nfw density profile , the equations for the numerical inversion ( [ inv_final ] ) were derived for an arbitrary density distribution . using this general form one can immediately obtain a family of dfs with our general anisotropy profiles for any potential - density pair .
for the commonly used density profiles it is easy to introduce the phase space units analogous to @xmath15 , @xmath18 and @xmath16 in our case .
this reduces the role of the parameters of the density profile to scaling properties so that the final df model would not explicitly depend on them .
given our very general parametrization of the @xmath84 profile , our df model is expected to provide some impact on the solution of the classical problem of mass - anisotropy degeneracy for spherical systems . in order to obtain more reliable estimates of mass profiles
, one could assume the anisotropy profile from the simulation and keep the density profile as the only degree of freedom of the df .
one could then apply the maximum likelihood approach of projected df , as described e.g. by mahdavi & geller ( 2004 ) .
a more advanced and simulation - independent approach would be to treat the anisotropy profile as an unknown quantity , described by the three parameters introduced in our formulation . as a result one would obtain an estimate of the mass profile as well as anisotropy .
both methods require an additional study of the df in projection and extensive tests on mock data sets
. this will be the subject of our follow - up papers .
the simulations have been performed at the altix of the lrz garching .
rw and e are grateful for the hospitality of astrophysikalisches institut potsdam and institut dastrophysique de paris where part of this work was done .
rw thanks a. knebe for helpful advice and g. bou for fruitful discussions .
this work was partially supported by the polish ministry of science and higher education under grant nn203025333 as well as by the polish - german exchange program of deutsche forschungsgemeinschaft and the polish - french collaboration program of lea astro - pf . an j. h. , evans n. w. , 2006a , aj , 131 , 782 an j. h. , evans n. w. , 2006b , apj , 642 , 752 ascasibar y. , binney j. , 2005 , mnras , 356 , 872 baes m. , van hese m. , 2007 , a&a , 471 , 419 binney j. , 1982 , mnras , 200 , 951 binney j. , tremaine s. , 1987 , galactic dynamics .
princeton univ . press , princeton cole s. , lacey c. , 1996 , mnras , 281 , 716 coln p. , klypin a. a. , kravtsov a. v. , 2000 , apj , 539 , 561 cuddeford p. , 1991 ,
mnras , 253 , 414 cuddeford p. , louis p. , 1995 ,
mnras , 275 , 1017 cuesta a. j. , prada f. , klypin a. , moles m. , 2007 , submitted to mnras , arxiv:0710.5520 dehnen w. , mclaughlin d. e. , 2005 , mnras , 363 , 1057 diemand j. , moore b. , stadel j. , 2004 , mnras , 352 , 535 faltenbacher a. , diemand j. , 2006 , mnras , 369 , 1698 fukushige t. , makino j. , 2001 , apj , 557 , 533 gottlber s. , yepes g. , 2007 , apj , 664 , 117 hnon m. , 1973 , a&a , 24 , 229 katgert p. , biviano a. , mazure a. , 2004 , apj , 600 , 657 kazantzidis s. , magorrian j. , moore b. , 2004 , apj , 601 , 37 klypin a. , gottlber s. , kravtsov a. v. , khokhlov a. m. , 1999 , apj , 516 , 530 kochanek c. s. , 1996 , apj , 457 , 228 kravtsov a. v. , klypin a. , khokhlov a. m. , 1997 , apjs , 111 , 73 louis p. d. , 1993 , mnras , 261 , 283 okas e. l. , 2002 , mnras , 333 , 697 okas e. l. , hoffman y. , 2001 , in spooner n. j. c. , kudryavtsev v. , eds , proc .
3rd international workshop , the identification of dark matter . world scientific , singapore , p. 121 okas e. l. , mamon g. a. , 2001 , mnras , 321 , 155 okas e. l. , mamon g. a. , 2003 , mnras , 343 , 401 okas e. l. , wojtak r. , gottlber s. , mamon g. a. , prada f. , 2006 , mnras , 367 , 1463 mahdavi a. , geller m. j. , 2004 , apj , 607 , 202 mamon g. a. , okas e. l. , 2005 , mnras , 363 , 705 merritt d. , 1985 , aj , 90 , 1027 merritt d. , navarro j. f. , ludlow a. , jenkins a. , 2005 , apj , 624l , 85 natarajan p. , hjorth j. , van kampen e. , 1997 , mnras , 286 , 329 navarro j. f. , frenk c. s. , white s. d. m. , 1997 , apj , 490 , 493 navarro j. f. et al . , 2004 , mnras , 349 , 1039 osipkov l. p. , 1979
, soviet astron .
lett . , 5 , 42 prada f. , klypin a. a. , simonneau e. , betancort - rijo j. , patiri s. , gottlber s. , sanchez - conde m. a. , 2006 , apj , 645 , 1001 saha p. , 1992 ,
mnras , 254 , 132 spergel d. n. et al . , 2007 ,
apjs , 170 , 377 taylor j. e. , navarro j. f. , 2001 , apj , 563 , 483 van der marel r. p. , magorrian j. , carlberg r. g. , yee h. k. c. , ellingson e. , 2000 , aj , 119 , 2038 voglis n. , 1994 , mnras , 267 , 379 widrow l. m. , 2000 , apjs , 131 , 39 wilkinson m. i. , evans n. w. , 1999 , mnras , 310 , 645 wojtak r. , okas e. l. , 2007 , mnras , 377 , 843 wojtak r. , okas e. l. , gottlber s. , mamon g. a. , 2005 , mnras , 361 , l1 wojtak r. , okas e. l. , mamon g. a. , gottlber s. , prada f. , moles m. , 2007 , a&a , 466 , 437
the volume @xmath29 of the hypersurface @xmath148 of constant energy and angular momentum is defined by the integral @xmath149 introducing spherical coordinates and changing variables into @xmath3 , @xmath4 and radius @xmath142 one gets @xmath150 where @xmath151 is the radial velocity and the integral is equal to the radial period of the orbit . using the radii of the pericentre @xmath152 and the apocentre @xmath153
, one can rewrite the final formula for @xmath29 in the following way @xmath154 integrating @xmath29 over the angular momentum we get the volume @xmath44 of the hypersurface of constant energy @xmath155 changing the order of the integrals and performing the integral over the angular momentum one obtains @xmath156 where @xmath46 is the apocentre radius of the radial orbit ( @xmath157 ) . considering a system of finite size @xmath158 in phase space
, one has to recalculate @xmath29 with the realistic hypersurface of constant @xmath3 and @xmath4 given by the cartesian product @xmath159 . in particular , for a spherical system with a boundary in the form of a sphere of radius @xmath48 the upper limit of the integral in ( [ g_der2 ] ) and ( [ ge_der2 ] )
must be replaced by @xmath160 and @xmath161 respectively .
the energy part of the df @xmath9 introduced in section 4 is related to the density profile by @xmath162 although the main part of this paper concerns the df consistent with the nfw profile , we keep within the appendix a general density @xmath163 so that the final formulae of inversion could be applied to any potential - density pair . changing variables in the integral ( [ inv_1 ] ) into the energy and angular momentum one gets @xmath164 where @xmath165 and @xmath166 .
the integral over the @xmath167 variable is evaluated analytically so that ( [ inv_2 ] ) can be rewritten in the form @xmath168 with a kernel of the integral given by @xmath169 where @xmath170 stands for the hypergeometric function .
equation ( [ inv_3 ] ) is a volterra integral of the first kind . in the general case of models with varying anisotropy , when @xmath171 and @xmath172 , it has no analytical solution for @xmath9 due to the complexity of expression ( [ kernel ] ) .
however , as shown by cuddeford & louis ( 1995 ) , this kind of integral can be quite easily inverted numerically .
below we adapt their method to our problem . for @xmath173 and @xmath174 the integral kernel can encounter a singularity , i.e. @xmath175 . in order to avoid this behaviour , we define a smooth integral kernel @xmath176 which is free of such a feature @xmath177 by analogy we introduce a smooth energy part of the df which is a regular function for energy approaching @xmath178 @xmath179 the value of @xmath180 is determined from the limit of @xmath163 at small potential ( @xmath181 and @xmath182 ) in equation ( [ inv_3 ] ) , for which @xmath183 which immediately gives @xmath180 in terms of @xmath6 , @xmath7 and the asymptotic slopes of the potential and density . in particular , for the nfw density profile one gets @xmath184 .
using formulae ( [ kernel_sm ] ) and ( [ f_e_sm ] ) we can rewrite equation ( [ inv_3 ] ) in the following form @xmath185 where @xmath186 and @xmath187 stands for all coefficients in front of the integral ( [ inv_3 ] ) . following cuddeford & louis ( 1995 )
we introduce discrete vectors of the potential @xmath188 , radius @xmath189 and density @xmath190 , where @xmath191 is an integer number and @xmath192 . for any @xmath193 we can split the integral ( [ inv_4 ] ) into a sum @xmath194 in order to apply any numerical algorithm to invert equation ( [ inv_5 ] ) with respect to @xmath195 ,
one has to assume that @xmath196 is sufficiently small so that the variations of @xmath195 and @xmath197 within subsequent integration ranges are negligible .
then one can approximate both functions by their values at @xmath198 , i.e. the middle points of the energy range .
this approach was used by cuddeford & louis ( 1995 ) and favoured over other methods involving higher order interpolation ( see e.g. saha 1992 ) . applying this approximation to equation ( [ inv_5 ] ) we get @xmath199 with @xmath200 and the matrix @xmath201 defined in the following way @xmath202,\end{aligned}\ ] ] where @xmath203 is the incomplete beta function .
as shown by cuddeford & louis ( 1995 ) , the solution of ( [ inv_6 ] ) for @xmath204 can be obtained by evaluating iteratively the following expression @xmath205 with the initial value @xmath206 given by @xmath207
all non - vanishing moments of the radial or tangential velocity at a given radius @xmath142 are defined by the integral @xmath208 where @xmath209 is a subscript indicating the velocity component in spherical coordinates and @xmath210 is an integer number . by analogy with steps ( [ inv_2 ] ) and ( [ inv_3 ] ) from appendix b
, one can evaluate the angular momentum part of the integral and simplify the expression to the following form @xmath211 where the kernel functions for the radial ( @xmath212 ) and tangential ( @xmath213 ) component of the velocity field are given respectively by @xmath214 once @xmath9 is calculated in terms of the @xmath215 vector it is easy to evaluate numerically the integral ( [ vmom ] ) and obtain the profile of any velocity moment .
an interesting property of the model is the ratio of any non - vanishing moment of the tangential velocity to the corresponding moment of the radial velocity in the limit of small and large radii . introducing spherical coordinates in ( [ vel_mom_1 ] ) and performing the integral with two asymptotes of @xmath85 given by ( [ f_l_asympt ] )
, one can show that this ratio is the following function of @xmath6 or @xmath7 @xmath216 | we study the distribution function ( df ) of dark matter particles in haloes of mass range @xmath0@xmath1 . in the numerical part of this work we measure the df for a sample of relaxed haloes formed in the simulation of a standard @xmath2cdm model .
the df is expressed as a function of energy @xmath3 and the absolute value of the angular momentum @xmath4 , a form suitable for comparison with theoretical models . by proper
scaling we obtain the results that do not depend on the virial mass of the haloes .
we demonstrate that the df can be separated into energy and angular momentum components and propose a phenomenological model of the df in the form @xmath5^{-\beta_{\infty}+\beta_{0}}l^{-2\beta_{0}}$ ] .
this formulation involves three parameters describing the anisotropy profile in terms of its asymptotic values ( @xmath6 and @xmath7 ) and the scale of transition between them ( @xmath8 ) .
the energy part @xmath9 is obtained via inversion of the integral for spatial density .
we provide a straightforward numerical scheme for this procedure as well as a simple analytical approximation for a typical halo formed in the simulation .
the df model is extensively compared with the simulations : using the model parameters obtained from fitting the anisotropy profile , we recover the df from the simulation as well as the profiles of the dispersion and kurtosis of radial and tangential velocities . finally , we show that our df model reproduces the power - law behaviour of phase space density @xmath10 .
galaxies : clusters : general galaxies : kinematics and dynamics cosmology : dark matter |
inequalities for polynomials and rational functions have rich history and numerous applications in different branches of mathematics , in particular in approximation theory ( see , for example , @xcite , @xcite and references therein ) . one of the best known inequalities is bernstein inequality for the derivative of polynomial on the interval @xmath0 $ ] .
this inequality was generalized and improved in several directions , for instance , instead of one interval it was considered on compact subsets of the real line ( see @xcite , @xcite ) and on circles ( @xcite,@xcite ) , on lemniscates @xcite , on systems of jordan curves @xcite , and more recently , on analytic jordan arcs @xcite
. for rational functions , bernstein type inequalities were shown for a circle @xcite , on compact subsets of the real line and circle ( see @xcite , @xcite , @xcite ) . for improvements of some of such inequalities , see @xcite and references therein .
the aim of this paper is to extend the mentioned above borwein - erdlyi inequality for rational functions from a circle to an arbitrary analytic jordan curve when poles are from a given compact set away from the curve .
the obtained inequality is asymptotically sharp . as a consequence we get a similar estimate on an analytic arc as well .
all the results are formulated in terms of normal derivatives of green functions of corresponding domains . for the necessary background on potential theory
, we refer to @xcite and @xcite .
the basic theme of this paper is similar to a recently developed method in @xcite , but most of the details are rather different .
this approach is based on using the borwein - erdlyi inequality as a model case , estimates for green s functions , fast decreasing polynomials , gonchar - grigorjan estimate of the norm of holomorphic part of meromorphic functions , and an appropriate interpolation ; for the case of an analytic arc we apply `` open - up '' mapping ( a rational function ) .
although bernstein inequality was developed to prove inverse theorems in approximation theory , we use direct approximation theorems to establish theorem [ thm : analrational ] .
the structure of the paper is the following : in section [ sec : statement ] we formulate the statements of new results , in section [ sec : proofone ] auxiliary facts are collected and the main theorem is proved , in section 4 we consider the case of one analytic arc , the last section is devoted to sharpness .
we denote the unit disk by @xmath1 , @xmath2 is called exterior of the unit disk and @xmath3 denotes the extended complex plane .
we frequently use @xmath4 for green s function of domain @xmath5 with pole at @xmath6 .
[ thm : analrational ] let @xmath7 be an analytic jordan curve , @xmath8 .
let @xmath9 be the interior of @xmath10 , @xmath11 be the exterior domain @xmath12 and let @xmath13 be a closed set .
denote the two normals to @xmath10 at @xmath14 by @xmath15 and @xmath16 , @xmath17 , where @xmath18 and @xmath19 are pointing inward and outward respectively .
then , for any rational function @xmath20 with poles in @xmath21 only , we have @xmath22 where the sum with @xmath23 ( or @xmath24 ) is taken over all poles of @xmath25 in @xmath26 ( or in @xmath11 , respectively ) , counting multiplicities , and @xmath27 denotes an error term that depends on @xmath10 , @xmath14 and @xmath21 , tends to @xmath28 as the total degree of @xmath25 tends to infinity and is independent of @xmath25 itself . applying an appropriate open - up mapping
, we obtain [ thm : analrationalarc ] let @xmath29 be an analytic jordan arc , @xmath30 not endpoint .
denote the two normals to @xmath31 at @xmath32 by @xmath33 and @xmath34 , @xmath35 .
let @xmath36 be the complementing domain and let @xmath37 be a closed set .
then , for any rational function @xmath38 with poles in @xmath21 only , we have @xmath39 where the sum with @xmath24 is taken over all poles of @xmath25 in @xmath40 counting multiplicities , and @xmath27 denotes an error term that depends on @xmath31 , @xmath32 and @xmath21 , tends to @xmath28 as the total degree of @xmath25 tends to infinity and is independent of @xmath25 itself .
theorem [ thm : analrational ] is asymptotically sharp as the following theorem shows .
[ thm : analratsharp ] we use the notations of theorem [ thm : analrational ] .
then there exists a sequence of rational functions @xmath41 with @xmath42 with poles in @xmath21 such that @xmath43 where @xmath44 tends to @xmath28 as @xmath45 and it depends on @xmath10 , @xmath46 and @xmath21 too and the sum with @xmath23 ( or @xmath24 ) is taken over all poles of @xmath25 in @xmath26 ( or in @xmath11 , respectively ) , counting multiplicities .
a particular sequence of rational functions showing sharpness in theorem [ thm : analrationalarc ] can be obtained from theorem 2 from @xcite in standard way if we take a point from @xmath21 and apply a fractional - linear mapping which maps this point to infinity .
the error term @xmath44 in theorems [ thm : analrational ] and [ thm : analrationalarc ] can not be dropped in general , even for polynomials , see @xcite .
we need the following `` rough '' bernstein type inequality on jordan curves . [
prop : rough_berns ] let @xmath10 be a @xmath47 smooth jordan curve and @xmath48 be a closed set .
then , there exists @xmath49 such that for any @xmath8 and any rational function @xmath25 with poles in @xmath21 only , we have @xmath50 this quickly follows from an idea attributed to szeg ( cauchy integral formula around the point , see e.g. @xcite , p. 133 ) .
fix @xmath8 and and put @xmath51 , @xmath52 and let @xmath53 .
now fix @xmath54 arbitrarily , @xmath55 .
let @xmath56 be the total order of poles of @xmath25 in @xmath57 , @xmath58 be the total order of poles of @xmath25 in @xmath59 .
obviously , @xmath60 .
since the poles are from a compact set , it is standard ( see , e.g. @xcite , lemma 1 ) that there exist @xmath61 and @xmath62 such that if @xmath63 , @xmath64 and @xmath65 , @xmath55 , then @xmath66 let @xmath67 ( assume @xmath68 ) and we use bernstein - walsh estimate for @xmath25 on @xmath57 and @xmath59 ( see , e.g. @xcite , theorem 5.5.7 , p. 156 ) : @xmath69 where the sum is taken over all poles @xmath23 of @xmath25 in @xmath70 , each of them appearing as many times as the order of the pole of @xmath25 at @xmath23 and @xmath64 .
it is again standard ( see , e.g. , lemma 1 from @xcite ) that there exist @xmath71 and @xmath72 such that if @xmath73 , then @xmath74 .
this way we can estimate the sum in the previous displayed formula as follows @xmath75 we apply cauchy integral formula @xmath76 the proposition is proved with @xmath77 which is independent of @xmath25 , @xmath78 and @xmath14 .
recall , for a given curve @xmath10 , @xmath26 denotes the interior of @xmath10 and @xmath11 denotes the unbounded component of @xmath79 . as earlier , @xmath80 and @xmath81 . with these notations , @xmath82 . using kellogg - warschawski theorem
( see e.g. @xcite p. 49
, theorem 3.6 ) , if the boundary is @xmath83 smooth , then the riemann mappings of @xmath84 onto @xmath85 respectively and their derivatives can be extended continuously to the boundary . under analyticity assumption , we can compare the riemann mappings as follows .
[ prop : green_deriv_g_one_g_two]let @xmath86 be fixed .
then there exist two riemann mappings @xmath87 , @xmath88 such that @xmath89 and @xmath90 , @xmath55 .
if @xmath82 is analytic , then there exist @xmath91 such that @xmath92 extends to @xmath93 , @xmath94 and @xmath95 is a conformal bijection , and similarly , @xmath96 extends to @xmath97 , @xmath98 and @xmath99 is a conformal bijection .
this proposition is proposition 7 in @xcite . from now on ,
we fix such two conformal mappings .
these mappings and domains are depicted on figure [ fig : analext ] .
, @xmath100 , the domain @xmath101 and the possible location of poles ] we may assume that @xmath101 and @xmath102 are of positive distance from one another ( by slightly decreasing @xmath103 , if necessary ) .
denote the normal vector ( of unit length ) to @xmath10 at @xmath46 or @xmath104 at @xmath105 pointing inward by @xmath106 or @xmath107 respectively .
similarly , the outward normal vectors are denoted by @xmath108 and @xmath109 .
[ prop : green_deriv_unitdisk ] the following hold for arbitrary @xmath110 , @xmath111 with @xmath112 , @xmath113 if @xmath114 @xmath115 and if @xmath116 , then @xmath117 this proposition is a slight generalization of proposition 8 from @xcite with the same proof .
the next result is based on a general construction of fast decreasing polynomials by totik , see @xcite , corollary 4.2 and theorem 4.1 too .
[ thm : fast_decr]let @xmath118 be a compact set , @xmath119 be a boundary point .
assume that @xmath120 satisfies the touching - outer - disk condition , that is , there exists a closed disk ( with positive radius ) such that its intersection with @xmath120 is @xmath121 .
let @xmath122 be a finite set and @xmath123 .
then there exist @xmath124 with the following properties .
for any given multiplicity function @xmath125 introduce @xmath126 where the sum is taken for @xmath127 and there exists a rational function @xmath128 such that @xmath129 , @xmath130 , @xmath128 has poles at the points of @xmath131 only and the order of the pole of @xmath128 at @xmath127 is at most @xmath132 and for @xmath133 we have @xmath134 roughly speaking , we transform for each @xmath127 the fast decreasing polynomials to move their poles from @xmath135 to @xmath23 and multiply them together .
let @xmath127 be fixed and let @xmath136 .
obviously , @xmath137 is a compact set and also satisfies the touching - outer - disk condition at @xmath138 .
using theorem 4.1 on page 2065 from @xcite , we know that there are @xmath139 depending on @xmath23 , @xmath120 and @xmath140 only and we get a polynomial @xmath141 such that @xmath142 , @xmath143 , @xmath144 and for @xmath133 @xmath145 let @xmath146 .
we immediately have @xmath129 , @xmath147 , and @xmath128 has poles at the points of @xmath131 only , the order of the pole of @xmath128 at @xmath127 is at most @xmath132 .
we multiply the estimates above together , so @xmath148 where @xmath149 and @xmath150 . in this subsection
we construct some auxiliary sets and the fast decreasing rational functions .
we also use @xmath151 in the following sense : if @xmath152 , then @xmath153 and @xmath154 .
let @xmath155 where @xmath156 , @xmath157 is fixed .
it is important that @xmath158 depends on @xmath57 only .
let @xmath159 , this disk touches the unit disk at @xmath105 . for every @xmath160 $ ] , @xmath161 consists of exactly two points , @xmath162 and @xmath163 .
it is easy to see that the length of the two arcs of @xmath164 lying between @xmath165 and @xmath166 are different , therefore , by reindexing them , we can assume that the shorter arc is going from @xmath165 to @xmath166 counterclockwise .
elementary geometric considerations show that for all @xmath167 , @xmath168 with @xmath169 , we have ( since @xmath170 ) @xmath171 let @xmath172 obviously , this @xmath173 is a compact set and satisfies the touching - outer - disk condition at @xmath105 of theorem [ thm : fast_decr ] .
consider @xmath174 \cup\phi_{1}\left[k_{v}^{*}\cap{\mathbb{d}}^{*}\right]\cup g_{1}.\ ] ] this is a compact set and also satisfies the touching - outer - disk condition at @xmath175 of theorem [ thm : fast_decr ] .
obviously , @xmath176 , @xmath177 , @xmath178 and if @xmath179 , then @xmath180 and @xmath181 too .
take any finite set @xmath182 with arbitrary multiplicity function such that the total multiplicity is at most @xmath183 . now applying theorem [ thm : fast_decr ]
, there exists a fast decreasing rational function for @xmath184 at @xmath14 and its degree is @xmath185 , where @xmath186 .
denote the poles of @xmath128 by @xmath187 counting multiplicities .
then @xmath188 , @xmath189 , @xmath190 on if @xmath191 , moreover @xmath192 note that @xmath193 and @xmath194 depend on @xmath195 and @xmath196 but they are independent of the multiplicity function and @xmath197 and @xmath198 . for simplicity
, we put @xmath199 .
it is easy to decompose @xmath25 into sum of rational functions , that is , @xmath200 where @xmath201 is a rational function with poles in @xmath202 , @xmath203 and @xmath204 is a rational function with poles in @xmath102 .
this decomposition is unique .
put @xmath205 , @xmath206 , then @xmath207 .
denote the poles of @xmath208 by @xmath209 , @xmath210 and the poles of @xmath211 by @xmath212 , @xmath213 ( with counting the orders of the poles ) .
now fix @xmath214 we use a gonchar - grigorjan type estimate for @xmath204 on @xmath57 ( see theorem 1 in @xcite ) so there exists @xmath215 such that we have @xmath216 obviously , we have @xmath217 consider @xmath218 we know that @xmath219 and @xmath220 .
we use the fast decreasing rational function @xmath128 from subsection [ sec : touchcircles ] .
we decompose `` the essential part of '' @xmath221 as follows @xmath222 where @xmath223 is a rational function , @xmath224 and @xmath225 is holomorphic in @xmath226 .
we use the decomposition @xmath227 we apply the gonchar - grigorjan type estimate ( see theorem 1 @xcite ) again for @xmath225 on @xmath226 , this way the following sup norm estimate holds @xmath228 where @xmath229 is a constant independent of @xmath221 .
furthermore , we can estimate @xmath230 on @xmath231 as follows @xmath232 we also need to estimate @xmath128 outside @xmath26 as follows . using bernstein - walsh estimate ,
we can write for @xmath231 @xmath233 where we use that the set @xmath234 is bounded , @xmath235 therefore , for all @xmath231 , @xmath236 this way we can continue ( [ est : phi_one_e_v ] ) and we use @xmath237 and that @xmath223 is a rational function with no poles outside @xmath226 and the maximum principle for @xmath223 @xmath238 and here we used that @xmath201 has no pole in @xmath59 and the maximum principle .
we can estimate these three sup norms with the help of ( [ eq : f_one_norm_est ] ) and ( [ eq : phi_one_norm ] ) , ( [ eq : f_one_norm_est ] ) and ( [ eq : phi_one_e_norm_est ] ) , ( [ eq : phi_one_norm ] ) , ( [ eq : f_one_norm_est ] ) .
hence we have for @xmath231 @xmath239 in this subsection we construct an approximation to @xmath225 which is holomorphic on a larger set containing @xmath226 .
later we will use properties and only .
the approximation is done by interpolating @xmath225 as follows .
@xmath225 is holomorphic in @xmath240 .
put @xmath241,\ ] ] where @xmath242 is @xmath243 and @xmath198 is @xmath244 . for simplicity , put @xmath245 where @xmath210 and @xmath246 .
introduce @xmath247 for the interpolation as follows @xmath248 where @xmath249 , @xmath250 are from @xmath251 , the first @xmath185 coincide with the @xmath252 s ( i.e. @xmath253 for @xmath254 ) and are arbitrary anyway .
modulus of @xmath247 can be written as follows ( when @xmath255 ) @xmath256 where @xmath257 .
note that if @xmath258 , then @xmath259 .
on @xmath260 , @xmath261 can be estimated from below as follows .
we know that @xmath101 and @xmath251 are disjoint and they are of positive distance from one another .
therefore there exists @xmath262 such that for all @xmath263 and @xmath264 , the modulus of the derivative of the blaschke factor at @xmath167 with pole at @xmath24 is at least @xmath265 : @xmath266 .
this implies that for all @xmath263 and @xmath264 @xmath267 moreover there is a similar upper estimate .
that is , there exists @xmath268 such that for all @xmath263 and @xmath264 @xmath269 multiplying for all blaschke factors ( appearing in ) , we obtain for @xmath270 , @xmath271 we also have a sharper lower estimate : @xmath272 we define the approximating rational function ( see e.g. @xcite chapter viii ) @xmath273 it is well known that @xmath274 does not depend on @xmath275 .
since @xmath105 is a double zero of @xmath247 , therefore @xmath274 and @xmath276 coincide there with @xmath225 and @xmath277 respectively .
the error of the approximating rational function @xmath274 can be written as @xmath278 where @xmath279 and recall that @xmath225 is holomorphic in @xmath280 and is continuous on @xmath281 .
for @xmath282 the error can be estimated as follows @xmath283 which tends to @xmath28 as @xmath284 , because @xmath285 and @xmath286 is fixed , that is @xmath287 so , @xmath274 is a rational function with poles in @xmath288 only and we know that @xmath289 where @xmath27 is independent of @xmath25 and depends only on @xmath10 and tends to @xmath28 as @xmath284 , furthermore @xmath290 now we interpolate and approximate @xmath291 .
we have the following description of the growth of green s function .
[ lem : green_est ] there exists @xmath292 depending on @xmath59 only and is independent of @xmath198 and @xmath25 such that for all @xmath293 and @xmath294 we have @xmath295 we can express green s function in the following way for @xmath296 , @xmath297 where @xmath298 is blaschke factor . hence @xmath299 here we use that there exists @xmath300 such that for all @xmath301 , @xmath302 and there exists @xmath303 such that for all @xmath304 , and @xmath305 , @xmath306 finally , the directional derivative of @xmath307 at @xmath167 from direction @xmath308 can be estimated as @xmath309 and integrating it along the radial ray @xmath310 $ ] , we obtain .
now we give the approximating rational function as follows @xmath311 where @xmath312 can be arbitrary as long as @xmath313 and @xmath314 , and @xmath247 is defined above .
we remark that we use the same interpolating points , but we need a different @xmath312 for the error estimate .
now we construct @xmath315 for the estimate and investigate the error .
we use @xmath316 , @xmath317 .
we give four jordan arcs that will make up @xmath315 .
let @xmath318 be the ( shorter , circular ) arc between @xmath319 and @xmath320 , @xmath321 be the longer circular arc between @xmath322 and @xmath323 , @xmath324 and similarly @xmath325 be the two segments connecting @xmath321 and @xmath318 .
finally let @xmath315 be the union of @xmath321 , @xmath326 , @xmath318 and @xmath327 .
the figure [ fig : gamma2 ] depicts these arcs and @xmath173 defined above .
, scaledwidth=60.0% ] we estimate the error of @xmath328 to @xmath329 on each integral separately : @xmath330 for the first term , we use bernstein - walsh estimate for the rational function @xmath204 on @xmath59 and the fast decreasing rational function @xmath128 as follows .
if @xmath331 , then with , @xmath332 ( uniformly in @xmath305 ) , therefore @xmath333 where we used ( [ eq : f_two_norm_est ] ) .
now we use the fast decreasing property of @xmath128 as follows .
we use @xmath334 hence , if @xmath335 , then by , @xmath336 , and then @xmath337 where @xmath338 and @xmath339 and therefore the fast decreasing rational function @xmath128 is small , see , in other words , @xmath340 therefore we can write @xmath341 where the coefficient tends to @xmath28 because , with the definition of @xmath342 , we can write @xmath343 and @xmath344 since @xmath345 . integrating along @xmath321 , we can write for @xmath282 @xmath346 here we used @xmath316 and @xmath347 .
we estimate the third term , the integral on @xmath318 , as follows for @xmath282 @xmath348 here , @xmath349 and @xmath350 .
roughly speaking , @xmath204 grows and this time @xmath128 grows too and only @xmath351 decreases .
we are going to estimate the growths with the help of bernstein - walsh estimate and blaschke factors . for simplicity , put @xmath352 , hence @xmath353 , @xmath354 .
this latter holds because @xmath355 .
moreover , @xmath356 , hence @xmath357 and together with @xmath358 , it implies that @xmath359 .
therefore @xmath360 .
moreover , @xmath361 ( if @xmath362 ) .
first , bernstein - walsh estimate for @xmath204 on @xmath59 yields for @xmath363 ( and in particular , if @xmath364 ) that @xmath365 as for @xmath247 , we use and , hence for @xmath366 @xmath367 as for @xmath128 , we use bernstein - walsh estimate for @xmath128 on @xmath368 and that @xmath369 .
therefore , @xmath370 and @xmath371 now we are going to multiply together these estimates .
consider the quotients @xmath372 for @xmath373 .
first , @xmath374 is continuous and holomorphic ( near @xmath105 , using the main branch of the logarithm since @xmath375 ) , @xmath376 because @xmath377 .
so there exists @xmath378 such that @xmath379 and this is uniformly true for all @xmath380 ( since @xmath251 is compact in @xmath381 ) .
now , taking the real part and integrating along radial rays ( see also the proof of lemma 9 in @xcite ) , we get @xmath382 applying this estimate for the product of , , and we can write @xmath383 where we used @xmath384 -2 $ ] and at the second step , the definition of @xmath342 and at the third step , again the the definition of @xmath342 and that @xmath198 is large ( hence @xmath385 ) at the fourth step . therefore , the integral over @xmath318 can be written as ( see ) @xmath386 where we used that @xmath387 ( if @xmath388 ) , and the length of @xmath318 is at most @xmath389 . for @xmath326 and @xmath327 , we apply the same estimate which we detail for @xmath326 only .
we again start with the integral for @xmath282 @xmath390 since @xmath391 , we can rewrite it in the form @xmath392 where @xmath393 ( with @xmath162 ) .
we use similar steps to estimate @xmath204 and @xmath247 and @xmath128 .
we use the estimate for @xmath394 , for @xmath395 and for @xmath128 .
we also use @xmath396 , so @xmath397 which we continue using @xmath398 ( see ) and @xmath399 and for the sum .
so @xmath400 which tends to @xmath28 if @xmath401 and we can continue this estimate using @xmath402 , so @xmath403 now continuing we write @xmath404 where @xmath402 and @xmath405 .
summarizing these estimates on @xmath321 , @xmath318 and @xmath326 ( and also on @xmath327 ) , and using with the exponential decay of @xmath406 , we have uniformly for @xmath407 , @xmath408 where @xmath27 tends to @xmath28 as @xmath284 but it is independent of @xmath25 and @xmath204 .
obviously , @xmath328 is a rational function with poles in @xmath288 only with @xmath409 and therefore we uniformly have for @xmath407 @xmath410 that is , @xmath411 where @xmath27 tends to @xmath28 as @xmath284 but it is independent of @xmath25 . since @xmath105 is double zero of @xmath247 , so we have @xmath412 consider the `` constructed '' rational function @xmath413 recall that the poles of @xmath223 are @xmath414 , @xmath210 , the poles of both @xmath274 and @xmath328 @xmath415 , @xmath213 and the poles of @xmath128 are @xmath252 , @xmath254 ( counting multiplicities ) .
hence , this function @xmath416 has poles at @xmath417 ( and with exactly the same multiplicities ) , and @xmath416 has poles at @xmath415-s and here the multiplicities may change a bit .
we use the identity @xmath418 to calculate the derivatives as follows @xmath419 where the second term is zero because of the fast decreasing rational function ( @xmath189 ) and for the first term we can apply the rough bernstein type inequality ( proposition [ prop : rough_berns ] ) in the following way ( @xmath420 ) : @xmath421 where @xmath422 . here
we use that @xmath423 because @xmath424 , the derivatives of blaschke factors are bounded away from @xmath28 in modulus ( @xmath425 is closed and disjoint from the unit circle ) and @xmath426 is bounded away from @xmath28 .
similarly for @xmath427 . there is a uniform upper estimate for the normal derivatives : @xmath428 therefore @xmath429 where we used that @xmath430 this way we need to consider @xmath431 only .
the derivatives at @xmath105 of the original @xmath25 and @xmath416 coincide , because of , and , so @xmath432 as for the sup norms , we use , , , so we write @xmath433 where @xmath27 tends to @xmath28 as @xmath434 but it is independent of @xmath25 , this follows from the considerations after and
. now we apply borwein - erdlyi inequality for @xmath416 as follows : @xmath435 where we also used the definition of @xmath249 s ( see ) .
now we apply proposition [ prop : green_deriv_unitdisk ] and with @xmath436 and the uniform upper estimate so we can continue the main estimate ( [ est : main_est_for_h ] ) @xmath437 where in the last step we used and the properties of @xmath128 . here
, @xmath44 tends to @xmath28 as @xmath434 but it is independent of @xmath25 , this follows from the consideration and @xmath436 . using that @xmath438 and summarizing these estimates
, we have @xmath439 which is the assertion of theorem [ thm : analrational ] .
here we use an analytic open - up tool to transform the arc setting ( @xmath440 plane ) to the curve setting ( @xmath441 plane ) . for a jordan curve @xmath10
, @xmath442 denotes the interior of @xmath10 and @xmath443 denotes the exterior of @xmath10 , @xmath444 .
[ prp : openup ] let @xmath29 be an analytic jordan arc . then there exist a rational function @xmath445 and an analytic jordan curve @xmath10 such that @xmath445 is a conformal bijection from @xmath446 and from @xmath443 onto @xmath447 .
this is a special case of @xcite , proposition 5 and actually can be established with the joukowskii mapping @xmath448 and using a suitable linear transformation .
this mapping is depicted on figure [ fig : openup ] .
denote the inverse of @xmath449 restricted to @xmath442 by @xmath450 and that restricted to @xmath443 by @xmath451 .
we need the mapping properties of @xmath445 as regards the normal vectors . for the full details ,
we refer to @xcite p. 879 .
briefly , there are exactly two @xmath452 , @xmath453 such that @xmath454 .
denote the normal vectors to @xmath10 pointing outward by @xmath455 and the normal vectors pointing inward by @xmath456 . by reindexing @xmath457 and @xmath458 and the exchanging @xmath459 and @xmath460
, we may assume that the normal vector @xmath461 is mapped by @xmath445 to the normal vector @xmath462 .
this immediately implies that @xmath463 , @xmath464 , @xmath465 are mapped by @xmath445 to @xmath466 , @xmath466 , @xmath462 respectively .
moreover , we need to relate the normal derivatives of green s functions as follows .
[ prp : greenopenup ] using the notations above , for the green s functions of @xmath467 and @xmath468 and for @xmath469 we have @xmath470 and , similarly for the other side , @xmath471 this proposition follows immediately from proposition 6 from @xcite and is a two - to - one mapping analogue of proposition [ prop : green_deriv_unitdisk ] .
use proposition [ prp : openup ] , and consider @xmath472 on the analytic jordan curve @xmath10 at @xmath473 ( where @xmath474 ) and put @xmath475 and @xmath476 , similarly as above .
obviously , @xmath477 .
we have then @xmath478 where @xmath24 runs through the poles of @xmath479 ( counting multiplicities ) .
now use proposition [ prp : greenopenup ] and @xmath480 , hence @xmath481 which we wanted to prove .
the idea is as follows . on the unit circle ( @xmath167 plane )
, we use some special rational functions for which borwein - erdlyi inequality is sharp
. then we transfer it back to @xmath10 ( @xmath441 plane ) and approximate it with rational functions . in other words , we do the `` reconstruction step '' in the `` opposite direction '' . recall that @xmath482 and @xmath257 is the blaschke factor with pole at @xmath483 .
first , we state the cases when we have equality in borwein - erdlyi inequality .
[ prp : besharpness ] suppose @xmath416 is a blaschke product with all poles are either inside or outside the unit circle , that is , @xmath484 where all @xmath485 , or @xmath486 where all @xmath487 .
then @xmath488 where the sum with @xmath23 ( or @xmath24 ) is taken over all poles of @xmath416 in @xmath226 ( or in @xmath381 , respectively ) , counting multiplicities .
this proposition is contained in borwein - erdlyi inequality as stated in @xcite pp .
324 - 326 .
the proof starts as follows .
take @xmath198 ( not necessarily different ) points from @xmath489 , denote them by @xmath490 and let @xmath491 obviously @xmath492 .
now we `` transfer '' @xmath493 to @xmath441 plane .
let @xmath494 be the sum of principal parts of @xmath495 . and approximate @xmath496 with rational functions with poles outside @xmath10 as follows .
it is immediate that @xmath497 is holomorphic in @xmath498 .
fix any @xmath499 arbitrarily .
take a rational function @xmath500 such that @xmath501 and @xmath502 .
hence @xmath503 is a conformal mapping from @xmath504 onto itself .
let @xmath505 and @xmath506 and @xmath507 .
fix a smooth jordan curve @xmath508 such that @xmath509 , @xmath510 and @xmath511 .
these are depicted on figure [ fig : sharpthree ] . , @xmath441 and @xmath167 and the mappings ] put @xmath512\ ] ] and denote a fekete polynomial ( monic polynomial with zeros at a fekete point set ) of @xmath513 with degree @xmath514 by @xmath515 .
we are going to use the sharpness of bernstein - walsh lemma ( see , e.g. @xcite , theorem 5.5.7 ( b ) , p. 156 ) therefore we put @xmath516 where @xmath517 denotes the harnack distance .
it is known that the @xmath198-th diameter @xmath518 of @xmath513 is close to the capacity @xmath519 , ( see fekete - szeg theorem @xcite , theorem 5.5.2 on page 153 ) , since @xmath520 .
hence @xmath514 is large , then @xmath521 and @xmath522 .
this way we have for large @xmath514 that @xmath523 let @xmath524 this @xmath247 is a polynomial with degree @xmath525 and @xmath526 is ( at least ) a double zero .
introduce @xmath527 where @xmath528 and this @xmath529 is a polynomial with degree @xmath525 . here , @xmath530 .
the error of interpolation on the @xmath531 plane is @xmath532 where @xmath528 .
it can be estimated from above as follows if @xmath533 @xmath534 where @xmath535 ( actually @xmath536 ) and we continue this estimate later . here
@xmath537 , @xmath538 can be estimated using that @xmath539 is a blaschke product with all poles in the unit disk , therefore @xmath540 and using the gonchar - grigorjan type estimate ( see theorem 1 in @xcite ) on @xmath446 for @xmath541 which implies that @xmath542 and , since @xmath543 has poles in @xmath442 , the maximum principle yields @xmath544 .
hence @xmath545 where @xmath198 is large enough ( depending on @xmath10 only ) .
we also use @xmath546 for some constant @xmath547 depending on @xmath10 and @xmath508 , since @xmath548 , and @xmath508 is fixed and is from positive distance from @xmath513 . finally , we estimate @xmath549 using ( when @xmath533 , @xmath538 ) , whence @xmath550 putting these estimates together , we can finish @xmath551 now substituting @xmath500 , we can write @xmath552 where @xmath553 and the constant in @xmath554 is independent of @xmath493 and @xmath198 and depends only on @xmath10 and @xmath508 only . finally , let @xmath555 this is a rational function with poles in @xmath442 with total order @xmath198 and one pole in the exterior of @xmath10 ( at @xmath556 ) with order @xmath557 .
moreover @xmath558 since @xmath559 .
we know that @xmath560 and using , @xmath561 therefore @xmath562 .
we use propostition [ prp : besharpness ] for @xmath493 , hence @xmath563 where actually the second term in the max is void ( @xmath493 has poles only inside the unit disk ) and by proposition [ prop : green_deriv_unitdisk ] , so @xmath564 in the last step we used that the first term in the max is @xmath565 and the second one is @xmath566 , hence if @xmath198 is large depending on @xmath10 , then the last equality holds .
this finishes the proof of theorem [ thm : analratsharp ] .
this research was partially done when the first author had a postdoctoral position at the bolyai institute , university of szeged , aradi v. tere 1 , szeged 6720 , hungary and was supported by the european research council advanced grant no .
he was partially supported by the russian science foundation under grant 14 - 11 - 00022 too ( theorem [ thm : analratsharp ] ) .
v. n. dubinin and s. i. kalmykov , _ a majorization principle for meromorphic functions _
sb . * 198 * ( 2007 ) , no .
12 , 3746 , english translation in sb .
( 2007 ) , no .
11 - 12 , 17371745 .
mr 2380804 ( 2009c:30074 ) v. n. dubinin , _ methods of geometric function theory in classical and modern problems for polynomials _ , uspekhi mat .
nauk * 67 * ( 2012 ) , no . 4(406 ) , 388 , english translation in russian math
. surveys 67 ( 2012 ) , no .
4 , 599684 . mr 3013845 a. a. gonar and l. d. grigorjan , _ estimations of the norm of the holomorphic component of a meromorphic function _ , math .
ussr - sb .
* 28 * ( 1976 ) , no . 4 , 571575 , russian original in mat .
99(141 ) ( 4 ) 634638 .
mr 0417417 s. i. kalmykov and b. nagy , _ on estimate of the norm of the holomorphic component of a meromorphic function in finitely connected domain _ , zap .
nauchn . sem .
s .- peterburg .
otdel . mat .
inst . steklov .
( pomi ) * 440 * ( 2015 ) , no .
analiticheskaya teoriya chisel i teoriya funktsii .
30 , 123137 .
a. l. lukashov , _
inequalities for the derivatives of rational functions on several intervals _
nauk ser . mat .
* 68 * ( 2004 ) , no . 3 , 115138 , english translation in izv .
68 ( 2004 ) , no .
3 , 543565 . mr 2069196 ( 2006j:26010 ) a. i. markushevich , _ theory of functions of a complex variable .
i , ii , iii _ , english ed . , chelsea publishing co. , new york , 1977 , translated and edited by richard a. silverman .
mr 0444912 ( 56 # 3258 ) g. v. milovanovi , d. s. mitrinovi , and th .
m. rassias , _ topics in polynomials : extremal problems , inequalities , zeros _ , world scientific publishing co. , inc . , river edge , nj , 1994 .
mr 1298187 ( 95m:30009 ) ch .
pommerenke , _ boundary behaviour of conformal maps _ , grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] , vol .
299 , springer - verlag , berlin , 1992 .
mr 1217706 ( 95b:30008 ) v. i. smirnov and n. a. lebedev , _ functions of a complex variable : constructive theory _ , translated from the russian by scripta technica ltd , the m.i.t . press ,
cambridge , mass . , 1968 .
mr 0229803 ( 37 # 5369 ) edward b. saff and vilmos totik , _ logarithmic potentials with external fields _ , grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] , vol .
316 , springer - verlag , berlin , 1997 , appendix b by thomas bloom .
mr 1485778 ( 99h:31001 ) j. l. walsh , _ interpolation and approximation by rational functions in the complex domain _ , fourth edition .
american mathematical society colloquium publications , vol .
xx , american mathematical society , providence , r.i .
mr 0218588 ( 36 # 1672b ) sergei kalmykov + department of mathematics , shanghai jiao tong university , 800 dongchuan rd , shanghai , 200240 , china + and + far eastern federal university , 8 sukhanova street , vladivostok , 690950 , russia + email address : [email protected] | borwein and erdlyi proved a bernstein type inequality for rational functions on the unit circle and on the real line .
here we establish asymptotically sharp extensions of their inequalities for rational functions on analytic jordan arcs and curves . in the proofs key roles are played by borwein - erdlyi inequality on the unit circle , gonchar - grigorjan type estimate of the norm of holomorphic part of meromorphic functions and totik s construction of fast decreasing polynomials .
keywords : rational functions , bernstein - type inequalities , borwein - erdlyi inequality , conformal mappings , green function classification ( msc 2010 ) : 41a17 , 30c20 , 30e10 |
the adiabatic loading of ultracold atoms and molecules in optical lattices represents a formidable opportunity to engineer strongly correlated states of quantum many - body systems with unprecedented control @xcite . on the one hand ,
known fundamental models for the physics of solid state systems can be literally implemented in optical lattices @xcite , and the experimental detection of their ground state or thermal equilibrium properties has the potential of responding to the lack of theoretical results due to fundamental limitations in the calculations , as those generally reported in fermionic systems or in frustrated quantum magnets . on the other hand , optical lattice systems offer the further advantage of controlling the hamiltonian parameters in real time , and this enriches the range of correlated phases that can be implemented , if one can guide the evolution of the system towards an off - equilibrium state which is not necessarily an eigenstate of a known hamiltonian .
an intense activity has recently been focused on the experimental implementation of disorder potentials in systems of trapped ultracold bosons , both experimentally @xcite and theoretically @xcite .
the introduction of tunable randomness in the system offers the possibility of realizing anderson localization of coherent matter waves of weakly interacting bosons @xcite and the opportunity of investigating the interplay between strong localization and strong interaction , _
e.g. _ in a deep optical lattice @xcite .
the fundamental model describing this rich phenomenology in presence of a lattice is the bose - hubbard model in a random potential @xcite , where , beside the conventional mott insulating and superfluid phases , a bose - glass phase appears , either associated with the fragmentation of weakly repulsive bosons into exponentially localized states , or with the localization of collective gapless modes for strong repulsion .
the disorder potential has so far been realized optically , either through laser speckles @xcite with or without an optical lattice , or , in optical lattices , through a secondary incommensurate standing wave @xcite . in the case of speckles ,
the typical length scale associated with the disorder potential is quite extended with respect to the correlation length of the bosons , so that classical trapping rather than quantum localization is responsible for the observed suppression of transport properties @xcite . on the other hand , the incommensurate superlattice of ref
. realizes a potential which is strongly fluctuating over the distance of a few lattice sites , but its _ pseudo_-disordered nature requires to take also into account gapped insulating phases at incommensurate fillings , competing with the bose glass @xcite .
an alternative proposal to create disorder in optical lattices involves the repulsive interaction with a secondary species of particles @xcite .
if the two species correspond to two different hyperfine states of the atoms , the use of state - dependent optical lattices @xcite allows one to first decouple the two species , and then to selectively suppress the hopping of one of them , freezing it in a given quantum state @xmath0 .
if the two species are subsequently brought into interaction , the mobile one evolves in a quantum superposition of all possible realizations of the random potential associated with the fock components of @xmath0 , and expectation values over the evolved state are therefore automatically averaged over the disorder distribution @xcite . in this proposal not only is the disorder potential strongly fluctuating over the length scale of a few lattice sites , but in principle one can also vary its statistics by preparing the frozen bosons in different states @xmath0 . in this paper
we investigate the above proposal in detail in the exactly solvable case of a one - dimensional gas of hardcore bosons on a lattice @xcite .
we consider both species of bosons ( the mobile one and the frozen one ) to be hardcore repulsive , which is not only experimentally feasible @xcite , but it has three fundamental theoretical advantages : 1 ) the state @xmath0 of the frozen bosons can be obtained exactly from the hamiltonian of the system before freezing , and disorder averaging can therefore be accurately performed ; 2 ) jordan - wigner diagonalization @xcite allows to calculate the exact real - time evolution of the mobile particles @xcite after they are brought into interaction with the frozen ones ; 3 ) and , most importantly from a conceptual point of view , localization phenomena of hardcore bosons are perfectly understood in terms of anderson localization of non - interacting fermions . while the observation of localization for interacting bosons is limited by screening of the disorder potential and the reduction of the healing length due to the interaction@xcite , many - body effects enter the system of non - interacting fermions only through fermi statistics .
in particular we focus on the case in which the frozen bosons are initially prepared in the superfluid ground state @xmath0 of the hardcore - boson hamiltonian at _ half _ filling with periodic boundary conditions .
the resulting disorder potential has the structure of a bimodal random on - site energy , with algebraically decaying correlations , but also with a very rich fourier spectrum dominated by short - wavelength components .
most importantly , such a potential is found to lead to anderson localization of all single - particle eigenstates ( apart from possibly a set of zero measure ) as expected for uncorrelated disorder @xcite .
the evolution of the mobile bosons in such a potential is found to invariably lead to a disordered steady state with exponentially decaying correlations and absence of quasi - condensation for a wide variety of realistic initial conditions . in particular phenomena of quasi - condensation in finite - momentum states
@xcite and fermionization @xcite , reported upon expansion of the hardcore bosons from a mott - insulating state and a superfluid state respectively , are completely washed out by the disorder potential .
hence we can conclude that this setup allows for a robust implementation of a localized state as the off - equilibrium steady state of the system evolution .
this paper is structured as follows : section [ system and method ] describes the system of two bosonic species , the exact diagonalization method for the study of real - time evolution , the monte carlo sampling of the disorder distribution , and the main features of the random potential ; section [ ground state properties ] is devoted to the study of localization of the single - particle eigenstates in the random potential ; section [ single trap ] investigates the evolution of the mobile bosons after interaction with the frozen ones when both species are prepared on a ring ; finally , section [ transport properties ] is dedicated to the study of the expansion of the mobile bosons in the random potential , starting from different initial confined states .
in this section we present the system of two bosonic species which is used to study the effect of a disorder potential ( subsection [ hamiltonian dynamics ] ) .
we then briefly describe the numerical procedure to exactly treat the equilibrium and out - of - equilibrium properties of the system ( subsection [ jordan - wigner transformation and implementation ] ) , and the sampling of the disorder distribution ( subsection [ disorder averaging ] ) .
finally , we examine the nature of the correlations in the disorder potential created by a frozen species of hardcore bosons in subsection [ characteristics of disorder ] .
the full system of two trapped interacting bosonic species in a one - dimensional optical lattice is described by the hamiltonian @xmath1 where @xmath2 is the hamiltonian of the bosons which will remain mobile , @xmath3,\end{aligned}\ ] ] @xmath4 is the hamiltonian for the bosons to be frozen , @xmath5,\end{aligned}\ ] ] and @xmath6 is the interaction hamiltonian @xmath7 here @xmath8 is the number of sites of the system , @xmath9 and @xmath10 are boson creation and annihilation operators and @xmath11 is the number operator for site @xmath12 .
symbols with the superscript @xmath13 are the corresponding operators for the frozen bosons that create the disorder potential . in eqs .
( [ h0]),([hf ] ) we also consider the possibility of both species being confined by a parabolic trap with different trapping strengths @xmath14 , @xmath15 . at times
@xmath16 the two species of bosons are not interacting with each other ( @xmath17 ) , and they are prepared in the factorized ground state @xmath18 of their respective hamiltonian with fixed numbers of particles @xmath19 and @xmath20 .
the ground state @xmath21 of the frozen particles can be decomposed in the fock basis @xmath22 where the sum extends over all fock states , @xmath23 at some time @xmath24 the frozen bosons are made immobile ( @xmath25 ) , and subsequently at @xmath26 the interaction between the two species is turned on ( @xmath27 ) .
furthermore , releasing the mobile bosons from their trap ( @xmath28 ) allows us to study their expansion properties .
the time evolution of the initially prepared state for @xmath29 is described by @xmath30 where the coefficients @xmath31 have acquired a phase factor which will become irrelevant , and the state @xmath32t\right\}{\bigl\lvert\phi\rangle } \ ] ] represents the time evolution of the initial state of the mobile bosons interacting with a single fock state @xmath33 of the frozen bosons , which determines the static external potential .
we have used the property that , for @xmath34 , @xmath35=0 $ ] .
hence , equation describes the parallel time evolution of the mobile bosons in a _ quantum superposition _ of different realizations of the disorder potential @xmath36 , each appearing with a probability @xmath37 .
remarkably , the time evolution of the expectation value of an operator @xmath38 acting only on the mobile bosons is automatically averaged over the disorder statistics @xcite : @xmath39 from here onwards we will restrict ourselves to the exactly solvable case of _ hardcore _ bosons , obtained in the limit @xmath40 for filling smaller than or equal to one .
it is convenient to incorporate the hardcore constraint directly in the operator algebra , passing to hardcore boson operators which _
_ anti__commute on the same site , @xmath41 , @xmath42 , and @xmath43 , @xmath44 .
making use of the jordan - wigner transformation @xcite , @xmath45 one can map the hardcore bosons operators onto spinless fermion operators @xmath46 and @xmath47 , obeying the same hamiltonian as the bosonic one , apart from a boundary term depending on the number of particles in the case of periodic / antiperiodic boundary conditions .
the fermionic problem is exactly solvable , and its eigenstates can be written in the following form @xmath48 where the @xmath19 columns of the matrix @xmath49 represent the first @xmath19 single - particle eigenstates . following the recipe of ref .
, from the matrix @xmath50 one can efficiently calculate the one - particle density matrix ( opdm ) @xmath51 and hence the momentum distribution @xmath52 which represents a fundamental observable in trapped atomic systems . at a more fundamental level
, the knowledge of the eigenvalues @xmath53 of the opdm , associated to eigenvectors @xmath54 also known as natural orbitals ( no ) , allows one to rigorously study condensation phenomena through the scaling of the maximum eigenvalue @xmath55 with the particle number @xcite . in absence of an external potential
the opdm of the hardcore bosons decays algebraically as @xmath56 , where @xmath57 , signaling off - diagonal quasi - long - range order in the system @xcite .
correspondingly the occupation of the @xmath58 momentum state , which coincides with the natural orbital with largest eigenvalue for a translationally invariant system , scales as @xmath59 with the particle number @xmath19 for any fixed density @xmath60 , namely it exhibits _ quasi - condensation_. remarkably , quasi - long - range order and quasi - condensation ( in the form of a @xmath61 scaling of the largest eigenvalue @xmath55 of the opdm ) survive also in presence of a trapping potential @xmath62 when the particle number is increased and correspondigly the trap strength @xmath14 is decreased so that the characteristic density in the trap @xmath63 is kept constant and smaller than a critical value @xmath64 ( @xmath65 for @xmath66 ) to avoid formation of a mott plateau in the trap center @xcite . finally , the exact solution of the fermionic hamiltonian allows one to calculate the real - time evolution of the fermionic wavefunction , and in particular of the @xmath67 matrix as @xmath68 with the single - particle time evolution operator given by @xmath69 making use of this approach , refs .
have shown that quasi - condensation is a robust feature of the system after expansion starting from an initially trapped quasi - condensed state , and it is even dynamically recovered when the initial state before expansion is a fully incoherent mott insulator state .
( [ expectation value ] ) shows that , ideally , the unitary evolution of the system explores all possible realizations of the disordered potential at once .
numerical calculations based on a matrix - product - state representation of the system state also enjoy this feature of `` quantum parallelism '' of the hamiltonian evolution @xcite by treating the disorder potential as a quantum variable in the system .
in this paper we use the more traditional approach of exactly calculating the ground - state properties and the hamiltonian evolution of hardcore bosons for a single realization of disorder at a time , averaging then over the disorder distribution through monte carlo importance sampling . accepting the overhead of disorder averaging
, this approach has the advantage that , unlike the method of ref . , the time evolution is exact for arbitrarily long times . according to eq . , the weights of the disorder configurations are defined by the coefficients of the fock - state decomposition for the initial state of the frozen bosons through @xmath70 . introducing the bosonic fock state @xmath71 and the @xmath72 matrix @xmath73 , after some algebra we get @xmath74 the energy eigenstates of noninteracting fermions with periodic boundary conditions , contained in the columns of the @xmath75 matrix ,
are given by plain waves .
thus it can be shown that the slater determinant of eq .
is a vandermonde determinant , which can be evaluated analytically .
the disorder weights finally become @xmath76.\ ] ] throughout the paper we will compare the effect of the disorder potential generated by the frozen superfluid state of @xmath77 bosons with that of the potential generated by @xmath77 _ fully uncorrelated _ frozen bosons @xcite , with the flat distribution @xmath78 even when equipped with the exact statistics of disorder as in eqs . and , it is generally hopeless to fully average the results of any calculation on the @xmath79 disorder configurations .
hence we opt for monte carlo importance sampling , where starting from an initially chosen random configuration @xmath33 , we propose a new one @xmath80 ( e.g. by swapping the occupation of two sites ) and accept it with metropolis probability @xmath81 .
the real - space properties are typically averaged over @xmath82 disorder realizations , obtained one from the other by updating @xmath83 sites , while the opdm is averaged over @xmath84 realizations ( due to the computational overhead of the opdm calculation ) .
this provides full convergence of the disorder - averaged quantities . a fundamental figure of merit for the proposed setup to introduce disorder in optical lattices is represented by the correlation properties of the disorder potential generated by the frozen particles .
as discussed in the introduction , the aspect of correlations represents one of the main weaknesses of the optical implementation of disorder through laser speckles @xcite .
in particular , we focus on the density - density correlation function @xmath85 where @xmath86 .
this correlation function for the frozen hardcore bosons is identical to the one for the corresponding fermions , and can be calculated exactly in the homogeneous system ( @xmath87 ) with periodic boundary conditions ; for @xmath88 , namely in the superfluid state of hardcore bosons , @xmath89^{-1 } \frac{\sin^2\bigl(\pi\frac{n^{\rm f}r}{l}\bigr)}{\sin^2\bigl(\pi\frac{r}{l}\bigr)}&\mbox{for } r\ge 1\\ \end{cases}\ ] ] hence , in the limit @xmath90 ( which is always satisfied in the thermodynamic limit ) the correlator decays algebraically like @xmath91 with a superimposed oscillation at twice the fermi wavevector @xmath92 .
the negative values of the correlator can be understood as resulting from an effective long - range repulsion between the hardcore bosons due to the kinetic energy , which enjoys as much free space around each boson as possible . from eq .
we deduce that the disorder potential created by the frozen bosons has slowly decaying correlations .
nonetheless it has fast oscillations on short distances , which are captured by the structure factor , namely the fourier transform of the correlator : @xmath93 this function can be evaluated to give for @xmath94 $ ] and @xmath95 @xmath96 the expression for @xmath97 follows from particle - hole symmetry .
the corresponding functions for an uncorrelated random potential generated by randomly displayed frozen particles are @xmath98 and @xmath99 .
we observe that the fourier spectrum of the frozen - boson potential is extremely broad , and it has a flat maximum for @xmath100 , reflecting the short wavelength oscillations of the potential .
hence we are in presence of a correlated but strongly fluctuating potential . in the following section we will see that such a potential generally leads to anderson localization of all single - particle states and to the suppression of quasi - condensation at all fillings . throughout the rest of the paper
, we will take the case of half - filling for the frozen bosons , @xmath101 .
in this section we investigate the properties of the eigenstates of the hamiltonian @xmath102 ( see equations and ) for the mobile bosons moving in the potential created by the frozen bosons . here
we consider an homogeneous system ( @xmath103 ) with periodic boundary conditions for both species of bosons .
we start with an analysis of the localization properties in real space and continue with an analysis of coherence in momentum space .
we investigate the localization properties in the frozen - boson potential through the _ participation ratio _ ( pr ) defined as @xmath104 where @xmath105 is the average particle density on site @xmath12 . in the case of a rectangular density profile
the participation ratio gives the support of the density profile , for exponentially localized states it is proportional to the decay length of the wavefunction , and for a gaussian - shaped density profile it is proportional to its standard deviation . .
the density of states is calculated for a system size @xmath106.,title="fig : " ] .
the density of states is calculated for a system size @xmath106.,title="fig : " ] the average pr of single particle eigenstates in the potential created by a half - filled system of frozen bosons at @xmath107 is shown in the upper panel of fig . [ pr eigenstates ] as a function of energy , for various system sizes .
the pr essentially becomes size independent for @xmath108 , clearly indicating the localization of all single - particle eigenstates , except for possibly a subset of zero measure ; we find a similar result for all strengths of the potential we investigated .
still a non - extensive number of extended eigenstates , not captured by this analysis , is in principle sufficient to suppress localization _
e.g. _ in transport experiments @xcite , so that further analysis is required to complete the picture on the localization properties of the frozen - boson potential ( see sec . [ transport properties ] ) .
the peak in the pr at the center of the band can be understood by the fact that the corresponding states have dominant @xmath109 components , which are also the dominant fourier components of the random potential for half filling .
+ we now move on to discuss the properties of the _ many - body _ ground state for hardcore bosons in the frozen - boson potential . making contact with the discussion of sec .
[ hamiltonian dynamics ] , such a ground state can be reached by adiabatically turning on the interaction @xmath110 between the two species after having prepared each of them separately and having quenched the hopping of the frozen bosons . in particular
we analyze the properties of the disorder - averaged opdm , which is translationally invariant , so that its eigenvalues correspond to the momentum distribution function ( mdf ) , and hence average condensation properties are studied in momentum space through the scaling of @xmath111 with the number of particles @xmath19 . of the opdm ( upper plot ) and the occupation of the zero momentum state @xmath112 ( lower plot ) for the ground state of a half - filled system of mobile bosons , interacting with the disorder potential .
the cases of a correlated potential resulting from the superfluid state , eq . , and that of a fully uncorrelated potential , eq . , are compared . here
the strength of disorder is @xmath107 .
the black lines are fits to @xmath113 for the first four data points.,title="fig : " ] of the opdm ( upper plot ) and the occupation of the zero momentum state @xmath112 ( lower plot ) for the ground state of a half - filled system of mobile bosons , interacting with the disorder potential . the cases of a correlated potential resulting from the superfluid state , eq . , and that of a fully uncorrelated potential , eq
. , are compared . here
the strength of disorder is @xmath107 .
the black lines are fits to @xmath113 for the first four data points.,title="fig : " ] the scaling analysis of the disorder - averaged occupations of the lowest natural orbitals @xmath55 is shown in fig .
[ scaling analysis ground state ] where , as @xmath19 grows , @xmath77 and @xmath8 are grown correspondingly such that @xmath114 . for small
system sizes the hardcore bosons show quasi - condensation behaviour @xmath115 , with @xmath57 within the error given by the simulation , but for larger system sizes @xmath55 saturates , hence revealing the absence of quasi - condensation in the thermodynamic limit . as shown in fig .
[ scaling analysis ground state ] , the same quantity for the potential generated by fully uncorrelated frozen particles gives a completely analogous picture .
this crossover can be qualitatively explained as a fragmentation effect .
the lowest natural orbital in each disorder configuration is localized , namely it does not scale with the system size and hence it can can only accommodate a finite number of particles , since the bosons repel each other .
as translational invariance is restored after disorder - averaging , the eigenvalues of the disorder - averaged opdm correspond to the momentum distribution function .
the peak at zero momentum , in which quasi - condensation appears in the case without disorder , is significantly reduced in presence of disorder . a scaling analysis of @xmath112 of the mobile bosons , shown in fig .
[ scaling analysis ground state ] detects a similar crossover from algebraic increase to saturation as for the disorder - averaged @xmath55 .
therefore , the consequences of fragmentation extends to this experimentally accessible observable , and we can conclude on the absence of quasi - condensation in the disorder - averaged opdm of the ground state of the system @xcite .
moreover the density of states , shown in the lower panel of fig .
[ pr eigenstates ] , reveals a continuous excitation spectrum at all energies .
hence , for a system of interacting bosons , the absence of quasi - condensation , together with the absence of gaps in the density of states , leads to classify the state of the system as a _ bose glass_. at the same time the jordan - wigner transformation translates this phase into an ideal anderson insulator of non - interacting fermions .
while the previous section investigated the state of the mobile bosons after an adiabatic on - turn of the interaction with the disorder potential created by the frozen bosons , in this section we consider the evolution of the mobile bosons after turning on that interaction suddenly .
we consider such a time evolution in the homogeneous system , _
i.e. _ @xmath103 , with periodic boundary conditions , mimicking the situation in which both species of particles are prepared in the same region of space before being brought into interaction . in this situation ,
the disorder - averaged real - space density of the mobile bosons remains uniform during time evolution , so that real - space localization effects are not visible .
nonetheless , even in a more realistic experimental scenario , in which both species of particles are kept in the same nonvanishing trap ( @xmath116 ) , the disorder - averaged density profile of the mobile species does not reveal marked localization effects , and it is even expanding during time evolution in order to reduce the overlap with the confined frozen bosons @xcite . with the particle number @xmath19 in a homogeneous system of mobile and frozen bosons ( @xmath103 ) at half - filling ( @xmath117 ) , with interaction strength @xmath107 .
the line corresponds to a fit @xmath118 to the first four data points . ]
the fundamental effect of disorder on the time evolution can be very clearly detected in momentum space , which can be measured in time - of - flight experiments .
the initial quasi - condensation peak in the mdf at @xmath58 decreases quickly during time evolution until a stationary regime is reached , in which the value of @xmath112 is oscillating with a small amplitude ( @xmath119 1% ) . the scaling analysis for the value of @xmath112 ,
time - averaged over the small oscillations , is shown in fig .
[ single.mdfscaling ] for a half - filled system of mobile bosons .
the occupation of the zero momentum state @xmath112 increases approximately like @xmath61 for small @xmath19 , but it then deviates from an algebraic increase and saturates for larger @xmath19 ; a completely analogous behavior is found for the case of uncorrelated disorder .
hence the scaling of @xmath112 reveals a crossover from quasi - condensation to fragmentation , with close similarity to the case of adiabatic evolution described in sec .
[ ground state properties ] ; in analogy to that case , we conclude that the steady state reached by time evolution realizes _ dynamically _ a bose glass .
in this section we discuss the expansion properties of initially confined mobile bosons in the potential created by the frozen particles .
the initial confinement for the mobile particles is provided by a tight parabolic trap ( @xmath120 ) , while the frozen particles can be imagined as prepared in a much shallower trap , whose effect is ignored for simplicity , so that the random potential they generate is the same as the one studied in the homogeneous case in sec .
[ characteristics of disorder ] .
in particular we imagine that at time @xmath26 the trap confining the mobile bosons is released ( @xmath28 ) and simultaneously the two species are brought into interaction ( @xmath27 ) .
after release from the trap , the mobile particles would expand forever in absence of disorder , so that a halt in presence of disorder explicitly shows localization effects .
subsection [ transport.localization ] deals with the evolution of the real - space properties during expansion , while subsection [ transport.quasi-condensation ] studies the evolution of the condensation properties of the system . for a single particle , initially in the ground state of a harmonic trap with @xmath121 , and interacting with a correlated disorder potential .
here the interaction strength is @xmath107 , and the system size is @xmath122 . ] . ] in this subsection we focus on the real - space properties of the time - evolved hardcore bosons in the frozen - boson potential , starting from a confined state in a parabolic trap .
snapshots from a disorder - averaged time evolution of a _ single _ particle starting from the ground state in the trap are depicted in fig .
[ transport.densities ] .
it is shown that the expansion reaches a _
localized _ steady state in the long - time limit ; in particular the spatial decay of the particle density undergoes a dramatic change from _ gaussian _ - as expected in a parabolic trap - to _ exponential _ - as expected in presence of anderson localization .
hence an anderson localized state is realized dynamically during expansion ; it is fundamental to stress that this steady state does not correspond to the ground state of the system , as the energy of the particle is conserved during expansion and hence it does not relax to the ground - state value .
the dynamical localization of the single particle wavefunction is fully captured by the time evolution of the pr , depicted in fig .
[ transport.pr frozen ] for various strengths @xmath110 of the disorder potential .
without disorder , the wavefunction spreads ballistically without changing its gaussian shape @xcite . in the presence of disorder ,
the pr saturates instead to a finite value , which decreases when increasing the disorder potential ; a fit to the final value of the pr , shown in fig .
[ transportprvaru ] , suggests that saturation in the pr takes place for any arbitrarily small value of the potential , as it would be expected for uncorrelated disorder , although exploring very small strengths of the disorder is numerically demanding as the steady state is reached for prohibitively large values of the pr . ) for @xmath19 particles , time - evolved in a system of size @xmath123 starting from the ground state of a harmonic trap ( @xmath121 ) and interacting with a correlated random potential or with a fully uncorrelated random potential .
the lines show a fit to the data points with an algebraically decaying function . ] ) of @xmath19 particles as a function of the particle number for various disorder strengths
. other parameters as in fig .
[ transportprvaru ] .
lower plot : total , kinetic , and potential energy of the @xmath124th single - particle eigenstate of the initial system in the trap.,title="fig : " ] + ) of @xmath19 particles as a function of the particle number for various disorder strengths .
other parameters as in fig .
[ transportprvaru ] .
lower plot : total , kinetic , and potential energy of the @xmath124th single - particle eigenstate of the initial system in the trap.,title="fig : " ] the time evolution of the participation ratio of many particles initially confined in a harmonic trap is similar to that exhibited by a single particle , and a saturation of the pr to a steady - state value is observed for the smallest value of @xmath110 that we could treat numerically ( fig .
[ transportprvaru ] ) .
+ nonetheless , upon increasing the particle number in the trap beyond a critical value @xmath125 ( see equation [ particle density trap ] with @xmath126 ) a mott plateau at unit filling starts forming in the center , and the properties of the initial trapped ground state change drastically .
it is then interesting to study whether the different initial conditions for the time evolution reflect themselves in the steady state of the system after the halt of the expansion .
the upper plot of fig . [ transport.varn ] shows the pr of the steady state as a function of the number of particles @xmath19 . the steady state pr increases with the number of particles for @xmath127 , showing a small peak for @xmath128 beyond which the increase with @xmath19 is much slower .
this feature can be understood in terms of the time evolution of the fermionic wavefunction eq . , whose real - space properties such as the pr are equivalent to those of the hardcore bosons , and whose time evolution is simply obtained through the time evolution of single - particle eigenfunctions . for @xmath129
the single particle wavefunctions that are populated for increasing @xmath19 are more and more confined to the sides of the trap , as the center has already a saturated density , and their energy is dominated by the trapping term , while the single - particle kinetic energy _ decreases _ with increasing @xmath19 beyond @xmath130 ( see lower plot of fig .
[ transport.varn ] ) . given that the trapping energy vanishes at @xmath26 after the trap release , the fermions occupying levels beyond the @xmath130-th expand with an energy which is less than that of the @xmath130-th level , so that they are expected to be localized by the disorder potential with a final participation ratio which is similar ( or even less than ) that of the @xmath130-th particle .
the time - evolved particle density of the many - body system is the simple sum of the squared time - evolved wavefunctions of the free fermions , so that the total participation ratio is not expected to increase significantly when adding particles beyond the @xmath130-th . for a system of @xmath131 hardcore bosons , initially in a trap with density @xmath132 , interacting with a correlated disorder potential . here
the system size is @xmath133 and the interaction strength is @xmath107 .
the momentum distribution for the corresponding spinless fermions is shown for comparison . ] in the previous section the real - space properties of initially confined hardcore bosons expanding in a disorder potential have been discussed .
this section addresses the condensation properties of the same system , motivated by the rich physical scenario offered by expanding hardcore bosons in absence of disorder @xcite .
subsection [ transport.harmonic ] is devoted to the time evolution of hardcore bosons initially confined in a superfluid state , while subsection [ transport.mott ] analyses the expansion from a mott insulator .
this subsection focuses on the expansion of the hardcore bosons initially confined in a shallow trap with characteristic density @xmath132 ( see eq . )
well below the critical value @xmath134 for the onset of a mott plateau in the trap center . with this initial condition quasi - condensation in the first no is present , and the mdf is peaked around @xmath58 at @xmath26 @xcite . in the absence of disorder
the mdf evolves towards that of the fermions ( which is a constant of motion ) , namely the bosons _ fermionize _ in momentum space , although they still quasi - condense in the lowest no which contains many momentum components @xcite .
furthermore , the opdm is decaying algebraically like @xmath135 , showing quasi - long - range order , whereas the phase of the opdm is oscillating at large distances leading to fermionization in momentum space .
the decay of the opdm is measured from the center of the trap , i.e. the center of the lowest no and the site with maximum occupation . from the system center @xmath136 at various times during evolution .
parameters as in fig .
[ transport.trapmdfu05 ] . ]
we now move on to the analysis of the evolution of the mdf during expansion from a shallow trap in the presence of disorder created by the frozen bosons .
the mdf of the mobile hardcore bosons is depicted in fig .
[ transport.trapmdfu05 ] at different times @xmath137 , compared to the mdf of fermions in the same system . in presence of disorder the mdf of the fermions
is no longer a constant of motion , but it is only slightly broadening in time due to its interaction with the frozen species of particles .
the effect of the disorder potential on the hardcore bosons is more significant .
the @xmath26 peak at zero momentum reduces its height , but , contrary to the expansion without disorder , it does not disappear and fermionization is not present . following ref .
, fermionization in absence of disorder is understood through the argument that , after a long - term expansion , the hardcore - boson system is dilute enough to be considered essentially non - interacting , and so it becomes equivalent to its fermionic counterpart . in presence of disorder , on the contrary , the expansion is stopped by localization , so that the extremely dilute limit is never reached and the hardcore bosons preserve their nature of strongly interacting particles .
furthermore , the interaction with a disorder potential leads to the loss of quasi - long - range order during time evolution , as shown in fig .
[ transport.trapopdmu05 ] by the decay of the opdm at various times .
the system finally reaches a steady state with an exponentially decaying opdm .
( recorded at time @xmath138 ) for varying number of particles in a system of size @xmath139 ; other parameters as in fig .
[ transport.trapmdfu05 ] .
the data for correlated frozen bosons are compared with those for fully uncorrelated ones .
the line corresponds to a fit to @xmath140 . ]
the loss of quasi - long - range order during expansion strongly suggests the loss of quasi - condensation in the system .
as the initial conditions break the translational symmetry , condensation properties are not captured by the scaling of the occupation of the zero - momentum state , and direct diagonalization of the disorder - averaged opdm is necessary to extract the scaling of the occupation @xmath55 of the lowest natural orbital .
@xmath55 decreases during time evolution , reaching a constant value in correspondence with the steady state observed in real - space .
the scaling analysis of @xmath55 is performed for this constant value in fig .
[ transport.trapscaling ] .
the scaling of @xmath55 deviates from the quasi - condensation behavior at sufficiently large particle numbers , revealing a crossover from quasi - condensation to fragmentation due to the localized nature of the lowest no .
fully uncorrelated disorder leads to a similar behavior , although the crossover appears to be much broader , and a quasi - condensation regime at low @xmath19 could not be identified .
hardcore bosons , initially in a perfect mott insulator , in a system of size @xmath122 without disorder.,title="fig:",width=302,height=188 ] hardcore bosons , initially in a perfect mott insulator , in a system of size @xmath122 without disorder.,title="fig:",width=302,height=188 ] hardcore bosons , initially in a mott insulator , interacting with the disorder potential created by frozen particles at half - filling . here
the interaction strength is @xmath107.,title="fig:",width=302,height=188 ] hardcore bosons , initially in a mott insulator , interacting with the disorder potential created by frozen particles at half - filling . here
the interaction strength is @xmath107.,title="fig:",width=302,height=188 ] as reported in ref . , hardcore bosons prepared in a perfect mott insulating state ( corresponding to an infinitely steep trap ) and subsequently time evolved exhibit the phenomenon of dynamical quasi - condensation at finite momentum .
indeed , from the initially flat mdf of the mott insulator two quasi - condensation peaks emerge at momenta @xmath109 during time evolution @xcite , as reproduced in fig .
[ transport.mottu0 ] .
strictly speaking quasi - condensation happens in two degenerate nos whose fourier transform is sharply peaked around @xmath109 ; the nos propagate at a velocity @xmath141 corresponding to the maximal group velocities @xmath142 for the single - particle dispersion relation @xmath143 at momenta @xmath109 .
the occupation of the degenerate lowest nos , @xmath55 , follows initially a universal power - law increase with time , independent of the number of particles @xmath19 ; at an @xmath19-dependent characteristic time @xmath144 , the two degenerate lowest nos begin to move in opposite directions , and @xmath55 starts to algebraically decrease , although its scaling with particle number shows the typical quasi - condensation behavior @xmath145 .
the two lowest nos clearly appear in the real - space densities as coherent fronts of the atomic cloud moving in opposite directions ( see fig . [ transport.mottu0 ] ) .
this aspect suggests the use of this setup to produce an atom laser @xcite .
we now consider the case of expanding hardcore bosons in the disorder potential created by frozen particles .
[ transport.mottu05 ] shows the density profile and the mdf for the expansion from a perfect mott insulator in the presence of disorder created by frozen particles .
initially peaks at momenta @xmath109 are emerging from the flat mdf at @xmath26 as in the case without disorder , and the hardcore boson cloud shows two outer fronts expanding ballistically in opposite directions .
nonetheless this initially coherent expansion is rapidly suppressed due to localization , which leads to a decrease and broadening of the momentum peaks at larger times up to a final steady - state mdf in which two broad peaks survive , and which corresponds to a localized state in real space as seen in subsection [ transport.localization ] .
a deeper analysis of condensation effects relies on the nos and their occupations .
[ transport.lambdavaru ] and [ transport.lambdavarn ] show the time evolution of the largest eigenvalue @xmath55 for different disorder strengths and different particle numbers .
initially @xmath55 is two - fold degenerate , corresponding to reflection symmetry at the center of the system , and it increases following a universal power law independent of the particle number , similarly to what is observed in absence of disorder but with a different disorder - dependent exponent . as in the case @xmath17 , at a characteristic time @xmath144 the two degenerate natural orbitals start to move into opposite directions , and correspondingly the time evolution of @xmath55 turns into a decreasing behavior ; @xmath144 strongly depends on the disorder strength ( fig .
[ transport.lambdavaru ] ) , while its dependence on the particle number @xmath19 becomes weaker for large @xmath19 , where @xmath144 is seen to approach an asymptotic value ( fig .
[ transport.lambdavarn ] ) . unlike the case @xmath17 , at a later stage the expansion of the system is stopped by disorder , and the degeneracy in the lowest nos is removed , going from two propagating ones to a single no localized in the system center .
hence the expanding system of hardcore bosons from a pure mott state shows a crossover from an incipient quasi - condensation regime at finite momenta to a localization regime .
it is natural to ask whether the system displays true quasi - condensation at any intermediate point in time . to address this issue
, we perform a scaling analysis of @xmath55 at the maximum - coherence time @xmath146 .
the results of this analysis are shown in fig .
[ transport.mottscaling ] , where @xmath55 exhibits a clear saturation for large particle numbers , and hence a crossover from quasi - condensation to fragmentation . repeating the same analysis for uncorrelated random disorder
we find a scaling of @xmath55 which is in qualitative agreement with the case of disorder generated by frozen particles ( the apparent quantitative agreement is accidental ) . for various disorder strengths .
other parameters as in fig .
[ transport.mottu05 ] . ] for various numbers of particles , and for disorder strength @xmath147 .
other parameters as in fig .
[ transport.mottu05 ] . ] during time evolution of @xmath19 hardcore bosons , initially in a mott insulator .
all parameters as in fig .
[ transport.lambdavarn ] .
the line corresponds to a fit to @xmath148 of the first four data points . ]
the original experimental proposal motivating this work is reported in ref . , where the use of state - dependent lattices was initially envisioned to create the quantum superposition of random potentials .
as shown in ref .
, two hyperfine states of the same atomic species can be loaded in the minima of two different polarization components of an optical lattice at a so called `` magic wavelength '' @xcite , at which each hyperfine state couples to one and only one polarization component .
hence , if the two polarization components are initially shifted by @xmath149 in space , the two species are non essentially interacting ( see fig . [ two species interaction ] ) .
the possibility of increasing drastically the intensity of one of the two polarization components would allow for a sudden quench of the hopping of one of the two species , preparing in this way the quantum superposition of random potentials .
the two species can be then brought into interaction at different strengths by shifting the spatial phase between the two polarization components of the lattice .
an adiabatic shift would transfer the mobile species to the bose - glass ground state in the random potential , as discussed in section [ ground state properties ] , whereas a sudden shift would give rise to an out - of - equilibrium bose - glass state as steady state after a transient evolution , as discussed in section [ single trap ] . as demonstrated in ref .
, the hardcore regime is achieved by using deep lattices for both polarization components and extremely dilute gases of both hyperfine states , so that a density of less than one atom per site is achieved in the lattice after loading .
as seen in section [ transport properties ] , expansion experiments in the disorder potential require to confine the two components with strongly different trapping frequencies , and then to release one of the two traps independently of the other . making use of the selective coupling of different hyperfine states to different polarization components of a magic - wavelength laser ,
it is possible to more strongly confine one of the two species through an optical dipole trap obtained by a tightly focused and polarized laser propagating transverse to the chain direction in the one - dimensional optical lattice . throughout all the previous sections we have seen that the onset of localization leaves very strong signatures on the momentum distribution , which is observed in standard time - of - flight experiments .
moreover expansion experiments in the disorder potential lead to exponentially localized steady states , whose density profile can be reconstructed by absorption images taken shortly after turning off all lasers @xcite . despite the averaging over the various tubes of the one - dimensional optical lattice @xcite ,
the extreme tails of the averaged particle density distribution should be dominated by the exponentially localized tails of the particle density in the central tubes of the one - dimensional optical lattice .
in this paper we have shown that a gas of one - dimensional hardcore bosons undergoes genuine quantum localization effects when set into interaction with a secondary species of bosons frozen in a massive quantum superposition of fock states .
each fock state can be regarded as a realization of a random potential , and the unitary evolution of the mobile species of bosons follows all possible paths related to the various disorder realizations _ in parallel_. physically relevant states in which one can prepare the frozen bosons , as for instance the superfluid state in the hardcore limit , realize a rapidly fluctuating disorder potential over the length scale of a few lattice spacings ; despite its power - law decaying correlations , the disorder potential is found to lead to the same localization effects as those observed in a _ fully uncorrelated _ potential . in the hardcore boson limit for the mobile species
these effects can be studied exactly in real time through jordan - wigner diagonalization , and we can numerically simulate realistic experimental setups for the dynamical preparation of localized many - body states . the equilibrium state of the hardcore bosons in the random potential is found to be a homogeneous bose - glass state with exponentially decaying correlations ; a similar state can be realized also dynamically after a sudden on - turn of the interaction between the two species . when the hardcore bosons are initially confined in a tight trap and then set free to expand in the random potential , for any non - vanishing disorder strengths the expansion stops and the system reaches an exponentially localized state . for any setup discussed in this work ,
we observe the absence of quasi - condensation and quasi - long - range order due to the disorder potential .
when the two species are confined in the same region and brought into interaction , the steady state of the system displays exponentially decaying off - diagonal correlations ; in the expansion experiments the disorder potential destroys the effects of fermionization ( when expanding from an initially dilute state ) and quasi - condensation at finite momentum ( when expanding from a mott state ) .
hence disorder created by a species of frozen hardcore bosons represents a very robust way to experimentally implement strongly fluctuating random potentials in optical lattices , and to realize fundamental localization effects of many - body systems .
we thank j. j. garca - ripoll and v. murg for useful discussions .
this work is supported by the european union through the scala integrated project .
99 d. jaksch , c. bruder , j.i .
cirac , c.w .
gardiner , and p. zoller , phys .
rev . lett . * 81 * , 3108 ( 1998 ) .
m. greiner , o. mandel , t. esslinger , t.w .
hnsch , and i. bloch , nature ( london ) * 415 * , 39 ( 2002 ) .
i. bloch , j. dalibard , and w. zwerger , cond - mat/0704.3011 .
m. lewenstein , a. sanpera , v. ahufinger , b. damski , a. sen de , and u. sen , advances in phyics * 56(2 ) * , 243 ( 2007 ) .
d. jaksch and p. zoller , annals of physics * 315 * , 1 ( 2004 ) .
e. jan , g. vidal , w. dr , p. zoller , and j.i .
cirac , quant .
comput . * 3 * , 15 ( 2003 ) .
lye , l. fallani , m. modugno , d.s .
wiersma , c. fort , and m. inguscio , phys .
lett . * 95 * , 070401 ( 2005 ) .
d. clment , a.f .
varn , m. hugbart , j.a .
retter , p. bouyer , l. sanchez - palencia , d.m .
gangardt , g.v .
shlyapnikov , and a. aspect , phys .
* 95 * , 170409 ( 2005 ) . c. fort , l. fallani , v. guarrera , j.e .
lye , m. modugno , d.s .
wiersma , and m. inguscio , phys .
lett . * 95 * , 170410 ( 2005 ) .
t. schulte , s. drenkelforth , j. kruse , w. ertmer , j. arlt , k. sacha , j. zakrzewski , and m. lewenstein , phys .
. lett . * 95 * , 170411 ( 2005 ) .
l. fallani , j. e. lye , v. guarrera , c. fort , and m. inguscio , phys .
lett . * 98 * , 130404 ( 2007 ) .
t. schulte , s. drenkelforth , j. kruse1 , r. tiemeyer , k. sacha , j. zakrzewski , m. lewenstein , w. ertmer , and j.j .
arlt , new j. phys .
* 8 * , 230 ( 2006 ) .
r. roth and k. burnett , j. opt .
b * 5 * , s50 ( 2003 ) .
r. roth and k. burnett , phys .
a * 68 * , 023604 ( 2003 ) .
b. damski , j. zakrzewski , l. santos , p. zoller , and m. lewenstein , phys .
lett . * 91 * , 080403 ( 2003 ) .
kuhn , c. miniatura , d. delande , o. sigwarth , and c.a .
mller , phys .
95 * , 250403 ( 2005 ) .
l. sanchez - palencia , d. clment , p. lugan , p. bouyer , g.v .
shlyapnikov , and a. aspect , phys .
. lett . * 98 * , 210401 ( 2007 ) .
d. clment , a.f .
varn , j.a .
retter , l. sanchez - palencia , a. aspect , and p. bouyer , new journal of physics * 8 * , 165 ( 2006 ) .
p. lugan , d. clment , p. bouyer , a. aspect , m. lewenstein , and l. sanchez - palencia , phys .
lett * 98 * , 170403 ( 2007 ) .
l. sanchez - palencia , phys .
a * 74 * , 053625 ( 2006 ) .
h. gimperlein , s. wessel , j. schmiedmayer , and l. santos phys .
95 , 170401 ( 2005 ) .
p. sengupta , a. raghavan , and s. haas new j. phys . 9 , 103 ( 2007 ) .
a. sanpera , a. kantian , l. sanchez - palencia , j. zakrzewski , and m. lewenstein , phys .
lett * 93 * , 4 ( 2004 ) .
fisher , p.b .
weichman , g. grinstein , and d.s .
fisher , phys . rev .
b * 40 * , 546 ( 1989 ) .
nevertheless , recent theoretical works has shown that laser - speckle potentials within experimental reach can possibly lead to anderson localization of weakly interacting bosons @xcite .
a. m. rey , i. i. satija , and c. w. clark , phys .
a * 73 * , 063610 ( 2006 ) .
u. gavish and y. castin , phys .
* 95 * , 020401 ( 2005 ) .
b. paredes , f. verstraete , and j.i .
cirac , phys .
lett . * 95 * , 140501 ( 2005 ) .
d. jaksch , h .- j .
briegel , j.i .
cirac , c.w .
gardiner , and p. zoller , phys .
lett . * 82 * , 1975 ( 1999 ) .
brennen , c.m .
caves , p.s .
jessen , and i.h .
deutsch , phys .
lett . * 82 * , 1060 ( 1999 ) .
o. mandel , m. greiner , a. widera , t. rom , t.w .
hnsch , and i. bloch , nature * 425 * , 937 ( 2003 ) .
e. lieb , schultz , and d. mattis , ann .
( ny ) * 16 * , 406 ( 1961 ) .
b. paredes , a. widera , v. murg , o. mandel , s. flling , i.j .
cirac , g.v .
shlyapnikov , t.w .
hnsch , and i. bloch , nature * 429 * , 277 ( 2004 ) .
t. kinoshita , t.r .
wenger , and d.s .
weiss , science * 305 * , 1125 ( 2004 ) .
m. rigol and a. muramatsu , phys .
a * 70 * , 031603(r ) ( 2004 ) ; m. rigol and a. muramatsu , phys . rev .
a * 72 * , 013604 ( 2005 ) .
m. rigol and a. muramatsu , phys .
lett . * 93 * , 230404 ( 2004 ) .
m. rigol and a. muramatsu , phys .
lett . * 94 * , 240403 ( 2005 ) .
e. abrahams , p.w .
anderson , d.c .
licciardello , t.v .
ramakrishnan , phys .
* 42 * , 673 ( 1979 ) .
k. ishii , suppl .
53 * , 77 ( 1972 ) ; d. j. thouless , phys
. rep . * 13 * , 93 ( 1974 ) .
o. penrose and l. onsager , phys rev . * 104 * , 576 ( 1956 ) .
korepin , a.g .
izergin , and n.m .
bogoliubov , quantum inverse scattering method and correlation functions , cambridge university press ( 1993 ) .
although the disorder - averaged largest eigenvalue @xmath150 and the disorder - averaged @xmath58 occupation @xmath112 show a similar scaling with system size , they have no simple relationship to each other ; indeed @xmath55 refers to localized natural orbitals , while @xmath112 is associated with an extended one .
the uncorrelated frozen bosons are considered to be in the state @xmath151 projected on the subspace with a fixed number of particles . this state can not straightforwardly be prepared in an optical lattice . | we investigate the behavior of ultracold bosons in optical lattices with a disorder potential generated via a secondary species frozen in random configurations .
the statistics of disorder is associated with the physical state in which the secondary species is prepared .
the resulting random potential , albeit displaying algebraic correlations , is found to lead to localization of all single - particle states .
we then investigate the real - time dynamics of localization for a hardcore gas of mobile bosons which are brought into sudden interaction with the random potential .
regardless of their initial state and for any disorder strength , the mobile particles are found to reach a steady state characterized by exponentially decaying off - diagonal correlations and by the absence of quasi - condensation ; when the mobile particles are initially confined in a tight trap and then released in the disorder potential , their expansion is stopped and the steady state is exponentially localized in real space , clearly revealing anderson localization . |
the galactic diffuse @xmath0-ray emission is believed to be mostly produced in interactions of cosmic rays with the matter and the radiation fields in the galaxy , the main production mechanisms being electron non - thermal bremsstrahlung , inverse compton scatterings off the radiation fields and pion decay processes in inelastic collisions of nuclei and matter .
although the standard production mechanisms of @xmath0-rays @xcite explain generally well the spatial and energy distribution of the emission below 1 gev , the model does not match egret observations of the @xmath0-ray sky above 1 gev.@xcite the discrepancy between the model and the observations is more pronounced towards the direction of the inner galaxy , where it amounts to 60 per cent of the observed emission .
many possible explanations have been proposed to account for the gev excess . @xcite
have proposed an hardening of the cosmic ray proton spectrum .
the average proton spectrum in the galaxy would have to be harder than locally measured .
@xcite have suggested a renormalization of the electron spectrum in order to fit egret data . @xcite
have argued that diffractive processes make the @xmath0-ray spectrum harder than the incident proton spectrum by 0.05 in power - law index , that the scaling violation produces more @xmath1 , and , that the inelastic @xmath2 cross section has a logarithmic increase with the incident proton energy , so that the combination of these effects could explain about 20 per cent of the `` gev excess '' .
@xcite claimed that the features of the egret excess are consistent with the hypothesis of a signal from dark matter annihilation .
recently @xcite has argued that the excess in the diffuse emission measured by egret and comptel could be explained by a combination of galactic source populations too dim to be detected by egret in addition to the standard production mechanisms . in any case a major difficulty when studying the diffuse emission is to disantangle the truly diffuse emission from that produced by unresolved sources . at tev energies ,
milagro has observed diffuse emission from a large region of the galactic plane from galactic longitude @xmath3 of @xmath4 @xcite . the different theories to explain the gev excess
make very different predictions about the tev flux .
recently @xcite have argued that no strong signal of pion decay is seen in the @xmath0-ray spectrum .
the pion decay mechanism would not be able to explain the excess above 1 gev .
therefore , @xcite claimed that the milagro measurements of the diffuse flux at higher energies revealed a new excess at tev energies and concluded that a different mechanism is required to explain the two excesses , either dark matter decay or contribution from unresolved egret sources .
an extrapolation of the egret sources within the range in galactic longitude of the milagro observation overestimates the diffuse tev flux measured .
however , there could be a population of sources undetectable by egret , but contributing to both the gev and tev excess diffuse emission .
recent observations by the tev observatory , hess , in fact point to a new class of hard spectrum tev sources .
hess is an array of telescopes using the imaging atmospheric cherenkov technique to detect @xmath0-rays above 200 gev .
hess scans the inner part of the galaxy in the region between @xmath5 and @xmath6 degrees in galactic longitude ( @xmath7 ) and between @xmath8 and @xmath9 degrees in galactic latitude ( @xmath10 ) with an angular resolution better than 0.1 degrees .
its average sensitivity at the energy threshold of 200 gev is 2 per cent of the crab flux within @xmath11 degrees in galactic latitude and decreases outside this range .
( the flux of the crab nebula is about @xmath12 for @xmath13 @xcite ) .
hess has previously reported the detection of 8 unknown sources close to the galactic plane .
@xcite recently the total number of discovered sources has been updated to 15 .
@xcite three sources have also been detected that are identified with the galactic center , the snr rx j1713.7 - 3946 and g0.9 + 01 .
all hess sources are located close to the galactic plane , within @xmath14 degrees of latitude .
@xcite the fluxes above 200 gev from all sources are fit to power laws , @xmath15 , whose photon indices @xmath16 range from 1.8 to 2.8 , with a relatively hard mean index of 2.32 and rms of 0.2 . in the inner galaxy region
observed by hess the existence of 91 snrs and 389 pulsars was already known at lower energies @xcite .
many of these snrs and pulsars within the inner galaxy might emit vhe @xmath0 rays but only a few @xmath0 ray sources were previously known . although the hess collaboration indicates a clear positional coincidence of its sources with a known snr or pulsar only in a limited number of cases , the distribution of galactic latitude of the eighteen vhe @xmath0 ray sources detected by hess agrees quite well with the distributions of all snrs catalogued by @xcite and of all pulsars catalogued by @xcite . assuming , for example , that snrs are a single class of counterparts with isotropic luminosity @xmath17 to the new hess sources and taking for them a simple radiative model
, @xcite found that the location of these sources favours a scale height of less than 100 pc , consistent with the hypothesis that these sources are either snrs or pulsars in a massive star forming region .
the hess detections of high energy @xmath0 rays from fifteen new sources has improved significantly the knowledge of both the spatial distribution and the spectra and fluxes of vhe @xmath0-ray galactic sources .
these hess results make it possible to estimate with unprecedented precision the contribution of unresolved point sources to the galactic diffuse emission measured up to 100 gev by egret and recently extended to tev energies by milagro.@xcite based on hess results , knowing the sensitivity and the field of view of an experiment , we estimate the number of expected sources and their expected vhe @xmath0-ray flux .
finally we evaluate the contribution of unresolved point sources to the diffuse @xmath0-ray emission for egret and milagro .
under the simplistic assumption that all sources have the same intrinsic energy luminosity @xmath18 for @xmath19 , the integral photon flux @xmath20 from a source depends upon the distance @xmath21 of the detector as @xmath22 where @xmath23 is the detector threshold energy and @xmath16 is the spectral index if the @xmath0-ray emission is a power law .
inverting eq .
( [ fluxlum ] ) @xmath24 where we have defined @xmath25 as @xmath26 if we assume that all sources are located in the galactic plane and their surface density , which does not have to be constant , is @xmath27 , then by differentiating eq.([df ] ) @xmath28 we obtain the number of sources per interval of galactic longitude @xmath29 and per interval flux @xmath30 @xmath31 as first case we assume that @xmath27 is constant . then integrating eq.([number ] ) over all source fluxes above hess sensitivity of about @xmath32 and over the range in longitude @xmath3 surveyed by hess the number of sources detected by hess fixes the density of sources in the galactic plane @xmath33 where the maximum flux detected by hess was @xmath34 . for a given source luminosity @xmath35 from eq.([intk ] ) @xmath36 for instance , assuming that the emission follows a power law with spectral index @xmath37 , the average spectral index of hess sources , if @xmath38 , @xmath39 , for @xmath40 , @xmath41 .
the average distance is 5.6 kpc for this luminosity .
this range of surface density is intermediate between the local density of snrs and pulsars . the region in the galaxy where point sources are likely located extends from @xmath42 kpc up to @xmath43 kpc @xcite .
assuming that the flux from each source depends only upon the distance from the detector @xmath44 and not on the type of source , hess total flux per steradian from resolved and unresolved sources is @xmath45 & = & \int_{d_{min}}^{d_{max } } \,\int_{l=-\frac{\pi}{6}}^{l=\frac{\pi}{6 } } \ , \frac{k \
, \sigma_{sources}}{d^2 } \ , dl \ ,
, dd \\[2 mm ] & = & 8.0 \times { 10}^{-9 } \quad { \mathrm { photons } } \quad { \mathrm{{cm}^{-2 } } } \ , { \mathrm { s^{-1 } } } \ , { \mathrm{{sr}^{-1 } } } \,.\end{aligned}\ ] ] @xmath46 in eq . [ fluxtotal ] the total integral flux is already divided by the hess field of view .
this flux is independent of @xmath18 .
the surface density of @xmath0-ray sources is not likely to be constant .
the distribution is probably similar to that of @xmath0-ray candidate sources such as pulsars or supernova remnants which are more prominent in the inner galaxy .
however , the ratio of @xmath0 emitting pulsars versus snrs is unknown .
the two extreme cases of all hess sources being either pulsars or snrs are considered .
@xmath27 in eq.([number ] ) will be either the surface density of snrs or the surface density of pulsars in the galactic plane , as observed in the radio wavelengths .
we do not consider the latitude distributions of pulsars and snrs because most point sources are located close to the plane of the galaxy , within a region ( @xmath47 ) which extends from @xmath42 kpc up to @xmath43 kpc @xcite . in terms of the heliocentric distance @xmath21 and the longitude angle @xmath3 the galactocentric distance , @xmath48 , is @xmath49 with @xmath50 @xmath51 , which is the distance from the earth to the galactic center .
the pulsar surface density @xmath52 , plotted in fig.1 , is fitted by the following shifted gamma function @xcite @xmath53}\ ] ] where @xmath54 and @xmath55 , @xmath56 @xmath57 , @xmath58 , @xmath59 and @xmath60 @xmath51 .
the snr surface density , plotted in fig.1 , is @xcite @xmath61 0 & \quad { \mathrm for } \quad { \mathrm r } > 16.8 \end{array } \right . \label{sigmasnr}\end{aligned}\ ] ] with @xmath62 @xmath57 , @xmath63 @xmath64 , @xmath65 and @xmath66 . integrating eq.([number ] ) over all pulsar fluxes above hess sensitivity , @xmath67 , and below @xmath68 , the greatest flux from a source detected by hess , and assuming a given source luminosity @xmath35 , the number of sources resolved by hess in the region ( @xmath7 ) fixes the maximum fraction of pulsars that are @xmath0-ray emitters @xmath69 & = & \frac{1}{2 } \,\int_{f_{min}}^{f_{max } } df \
, \frac{k}{f^2 } \ , \int_{l=-\frac{\pi}{6}}^{l=\frac{\pi}{6 } } d l \
, \sigma_{pulsar}(l , f ) = 18 \,.\end{aligned}\ ] ] @xmath70 from @xcite and @xcite and from @xcite and @xcite , respectively .
[ fig1],width=302 ] if hess population of sources consists entirely of pulsars , the total flux from resolved and unresolved sources divided over hess field of view is @xmath71 & = & \int_{d_{min}}^{d_{max } } dd \ , d \
, \frac{k}{d^2 } \ , \int_{l=-\frac{\pi}{6}}^{l=\frac{\pi}{6 } } d l \ , \sigma_{pulsar}(l , d ) \\[2 mm ] & = & 1.0 \times { 10}^{-8 } \quad { \mathrm { photons } } \quad { \mathrm{cm^{-2 } } } \quad { \mathrm { s^{-1 } } } \quad { \mathrm { { sr}^{-1 } } } \ , .
\end{aligned}\ ] ] @xmath72 if the all radio loud pulsars emit @xmath0-rays then the luminosity predicted in eq .
( [ sensitivity ] ) is @xmath73 ergs / s and the average distance would be 0.5 kpc . though the ratio between @xmath0-ray loud versus @xmath0-ray quiet pulsars is uncertain .
if we suppose that only five per cent of the radio loud pulsar emit @xmath0-rays , then the @xmath74 the source luminosity predicted would be @xmath75 ergs / s and the average distance for hess sources 1.8 kpc . if 1 per cent of radio loud pulsars emit tev @xmath0s then @xmath76 the source luminosity predicted would be @xmath77 ergs / s and the average distance for hess sources 3.5 kpc
if instead all hess sources belong to the snr population the total flux would be @xmath78 & = & \int_{d_{min}}^{d_{max } } dd \ , d \
, \frac{k}{d^2 } \ , \int_{l=-\frac{\pi}{6}}^{l=\frac{\pi}{6 } } d l \ , \sigma_{snr}(l ,
d ) \\[2 mm ] & = & 6.6 \times { 10}^{-9 } \quad { \mathrm { photons } } \quad { \mathrm{cm^{-2 } } } \quad { \mathrm { s^{-1 } } } \quad { \mathrm { { sr}^{-1 } } } \ , , \end{aligned}\ ] ] @xmath79 the source luminosity predicted for hess sources would be @xmath77 ergs / s and their average distance 3.5 kpc .
fig [ fig2 ] shows the values of luminosity @xmath18 and fraction of @xmath0-ray emitting pulsars or snrs , allowed by the condition that hess detects 18 sources . and fraction of @xmath0-ray emitting pulsars or snrs allowed by the condition that hess detects 18 sources .
the shaded region is not allowed .
[ fig2 ] , title="fig:",width=302 ] and fraction of @xmath0-ray emitting pulsars or snrs allowed by the condition that hess detects 18 sources .
the shaded region is not allowed .
[ fig2 ] , title="fig:",width=302 ] in the hess catalogue of fifteen new sources 8 sources have snrs as possible counterparts , and 4 of these new sources are potentially associated to pulsar wind nebulae .
so pulsars and snrs both occur but their surface densities are likely not so different . ) and by hess ( @xmath7 ) .[fig3],title="fig:",width=302 ] ) and by hess ( @xmath7 ) .[fig3],title="fig:",width=302 ]
milagro is a water cherenkov telescope , which has surveyed most of the northern sky at tev energies @xcite .
milagro has measured tev energy fluxes from the crab @xcite and the cygnus region @xcite and , recently the diffuse galactic @xmath0-ray emission up to tev energies .
milagro galactic diffuse emission is @xmath80 for @xmath81 tev in the region @xmath4 .
@xcite the detection of tev @xmath0-ray diffuse emission is of fundamental importance in solving the long standing problem of the production of high energy @xmath0 radiation in our galaxy .
in fact the milagro result seems to exclude an additional hard spectrum component continuing to above 10 tev to explain the 60 per cent excess of egret flux compared to @xmath82-decay production mechanism due to the local cosmic ray flux .
@xcite in order to discriminate among different production mechanisms of the diffuse radiation , it is important to estimate the amount of the measured diffuse emission which could arise from unresolved pointlike sources .
the milagro sensitivity to point sources is about @xmath83 and , in fact , it is of the order of the brightest integral source flux detected by hess .
however , part of the diffuse emission measured by milagro is emission from unresolved hess - like pointlike sources . the integral flux expected from the milagro field of view from hess - like sources assuming a pulsar shaped surface density and source luminosity @xmath84 is @xmath85 & = & \int_{0.3 \ , kpc}^{30 \ , kpc } dd \ , d \ , \frac{k}{d^2 } \
, \int_{l=\frac{2\pi}{9}}^{l=\frac{5\pi}{9 } } d l \ , \sigma_{pulsars}(l ,
d ) \\[2 mm ] & = & 2.4 \times { 10}^{-9 } \quad { \mathrm{{photons } } } \quad { \mathrm { s^{-1 } } } \,{\mathrm{{cm}^{-2 } } } { \mathrm { sr}^{-1 } } \ , , \end{aligned}\ ] ] @xmath86 while from hess - like sources assuming a snr shaped surface density and @xmath87 is @xmath88 & = & \int_{0.3 \ , kpc}^{30 \ , kpc } dd \ , d \ , \frac{k}{d^2 } \ , \int_{l=\frac{2\pi}{9}}^{l=\frac{5\pi}{9 } } d l \ , \sigma_{snr}(l ,
d ) \\[2 mm ] & = & 3.1 \times { 10}^{-9 } \quad { \mathrm { { photons } } } \quad { \mathrm { s^{-1 } } } \,{\mathrm{{cm}^{-2 } } } { \mathrm { sr}^{-1 } } \ , , \end{aligned}\ ] ] @xmath89 the fluxes in eq . ( [ eqn : terzabis ] ) and ( [ eqn : quarta ] ) are lower than hess flux from resolved and unresolved sources .
in fact as shown in fig.3 hess observes the inner region of the galaxy , whereas milagro field of view is more spread toward the outer regions in the galaxy , where the number of sources is substantially lower , so a lower contribution from unresolved sources is expected for milagro s region of the galactic plane than for hess .
milagro galactic diffuse emission is measured for a threshold energy of about 3.5 tev , whereas the fluxes from unresolved sources calculated in eq.([eqn : terzabis ] ) and ( [ eqn : quarta ] ) refer to hess threshold energy of 200 gev . in order to estimate the contribution of unresolved sources to milagro diffuse emission we correct the flux in eq.([eqn : terzabis ] ) and ( [ eqn : quarta ] ) for the different threshold energy .
all hess sources are fitted by a power law spectrum @xmath90 and the average spectral index @xmath37 .
the integral flux from unresolved pulsars if 5 per cent of radio loud pulsars emit @xmath0-rays corrected for milagro threshold becomes @xmath91 & \times & { ( \frac{e_{milagro}}{e_{hess}})}^{(-\gamma+1 ) } \\[2 mm ] & & = 5.5 \times { 10}^{-11 } \quad { \mathrm{{photons } } } \quad { \mathrm { s^{-1 } } } \,{\mathrm{{cm}^{-2 } } } { \mathrm { sr}^{-1 } } \end{aligned}\ ] ]
@xmath92 if all hess sources coincide with snrs @xmath93 & \times & { ( \frac{e_{milagro}}{e_{hess}})}^{(-\gamma+1 ) } \\[2 mm ] & = & 7.1 \times { 10}^{-11 } \quad { \mathrm { { photons } } } \quad { \mathrm { s^{-1 } } } \,{\mathrm{{cm}^{-2 } } } { \mathrm { sr}^{-1 } } \end{aligned}\ ] ] @xmath94 where @xmath95 and @xmath96 are milagro and hess energy threshold , respectively .
the contribution from unresolved hess - like sources to milagro diffuse emission is comparable to the diffuse emission itself . despite the uncertainty of this calculation , concerning especially the ratio bewteen @xmath0-ray loud pulsars and snrs , the result seems to indicate that unresolved point sources contribute significantly to milagro diffuse @xmath0-ray emission .
[ fig4 ] shows the fraction of the measured milagro flux which unresolved sources contribute to for different @xmath18 and for various total numbers of galactic snrs and pulsars .
the lines in blue represent the conditions for hess to detect 3 , 9 and 18 sources , respectively . and for different total number of galactic snrs and pulsars which are @xmath0-ray loud .
the lines in blue represent the conditions for hess to detect 3 , 9 and 18 sources , respectively .
[ fig4 ] , title="fig:",width=283 ] and for different total number of galactic snrs and pulsars which are @xmath0-ray loud .
the lines in blue represent the conditions for hess to detect 3 , 9 and 18 sources , respectively .
[ fig4 ] , title="fig:",width=283 ]
egret surveyed the @xmath0-ray sky in the energy region between 30 mev and 100 gev .
the third egret catalog lists 80 sources for latitudes less than @xmath97 degrees , of which 5 are pulsars , one is a solar flare and 74 are unidentified sources .
studies of the spectral characteristics @xcite and of the luminosity function @xcite of egret unidentified sources show as unlikely that even a small part of these sources are pulsars or snrs .
the contribution of these sources to egret flux was evaluated by egret collaboration and should be less than 10 per cent .
recently @xcite showed that the contribution of unresolved sources with luminosity @xmath98 , which are too dim to be detected by egret , could provide up to 5 , 10 per cent of egret emission .
also he showed that combinations of different source populations could explain the excess @xmath0-ray emission measured by comptel and egret at mev and gev energies , respectively .
the hess new population of sources does not seem to coincide with egret catalogue @xcite . from @xcite the egret point source sensitivity in the galactic plane
is about @xmath99 .
if we can believe that hess tev source fluxes continue down to gev energies , ten of these sources would have been detectable by egret if extrapolated without any cut - off to 100 mev .
the fact that only a few of hess new sources are spatially consistent with the 95 per cent positional error of unidentified egret sources ( hess j1640 - 465 , hess j1745 - 303 and hess j1825 - 137 for which a simple power law estrapolation of the spectrum to 1 gev gives egret flux ) suggests that the simple estrapolation of hess source fluxes to egret energies is probably not correct . these sources might have a cut - off close to 10 gev .
all these sources will be detectable by glast , which will have a flux threshold of about @xmath100 .
the sum of the integral fluxes of the three hess sources , which could coincide with egret unidentified sources , is @xmath101 at 1.4 gev , about one per cent of the egret diffuse flux . following the method introduced in chapter 2 we evaluate the contribution due to hess - like sources to egret diffuse emission @xmath102 at different energies @xmath103 above 0.50 gev and for the two cases in which either all snrs or 5 per cent of the radio loud pulsars contribute . in table
1 the first column gives the energy @xmath104 above which the integral flux @xmath102 is evaluated .
the measured egret integral flux for @xmath105 is given in the second column .
the contribution of hess - like pulsars or snrs is given in columns third and fourth , respectively .
as shown in table 1 the contribution from hess - like sources can account for about 10 per cent of the gev excess .
this contribution becomes more important at higher energies . at about 85 gev
the contribution from either hess - like pulsars or snrs is about 20 per cent of the diffuse flux . [
cols="^,^,^,^ " , ]
using hess new sample of galactic @xmath0-ray emitters we estimated the contribution of unresolved sources to the diffuse emission measured by egret and by milagro .
we find that the contribution of unresolved hess - like sources to the diffuse @xmath0 ray emission is large at energies above 100 gev , whereas it amounts to about 10 per cent of egret measurements of the diffuse flux at gev energies .
since pulsars and snrs are distributed mostly in the galactic plane , their unresolved fluxes will mainly contribute to the low latitude diffuse emission .
thus we expect that the total flux due to unresolved sources contributes less at higher latitudes . the main uncertainties of this calculation consist in assuming that all sources are monoluminous and the distribution of @xmath0-ray sources follows the distribution of either pulsars or snrs .
of the 1509 pulsars catalogued by the australian national telescope facility @xcite only a few are also @xmath0-ray emitters
. this could be due to strong selection effects as suggested by the existence of pulsars , such as geminga , which are @xmath0-ray loud pulsars , yet not observed in the radio .
however , both the snr and pulsar distribution give similar conclusions , as would any class of sources that follows a similar spatial distribution . to draw more certain conclusions about the high energy @xmath0-ray galactic sky , new observations are of fundamental importance .
veritas will be able to survey the same region of the sky as milagro .
milagro s recently improved sensitivity will better image the diffuse emission . finally , glast will investigate the window of energy between 10 mev to 300 gev , covering the energy gap left between egret and the ground based gamma - ray observatories . | the detection by the hess atmospheric cherenkov telescope of fifteen new sources from the galactic plane makes it possible to estimate the contribution of unresolved point sources like those detected by hess to the diffuse galactic emission measured by egret and recently at higher energies by the milagro collaboration . assuming that hess sources have all the same intrinsic luminosity ,
the contribution of this new source population can account for most of the milagro @xmath0-ray emission at tev energies and between 10 and 20 per cent of egret diffuse galactic @xmath0-ray emission for energies bigger than 10 gev .
also , by combining the hess and the milagro results , constraints can be put on the distribution and the luminosities of gamma ray emitters in the galaxy . |
support vector machines [ svm ; @xcite ] are a popular tool for classification .
two important aspects contributed a lot to this popularity .
first , support vector machines handle high - dimensional , low sample size data very well , in terms of computational efficiency as well as prediction quality .
therefore , they are well suited to tackle , for example , microarray data containing thousands of gene expression levels ( high dimensionality ) for a limited number of subjects ( low sample size ) ; see , for example , @xcite and @xcite .
second , support vector machines allow for incorporating kernel functions via the so - called kernel trick .
this way nonlinearity in the data can be handled , for example , using a polynomial or a gaussian kernel . moreover ,
nonnumerical data can be modeled by designing an appropriate kernel function using a priori biological information about the data at hand .
this strategy is reported to perform very well , for instance , in protein homology detection , for example , fisher svm [ @xcite ] , pairwise svm [ @xcite ] , spectrum kernel [ @xcite ] , mismatch kernel [ @xcite ] and local alignment kernel [ @xcite ] . for high - dimensional and complex data sets ,
the assumption of clean , independent and identically distributed samples is not always appropriate . in @xcite and @xcite , for instance , several samples are regarded as suspicious .
a potential drawback of support vectormachines is the sensitivity to an even very small number of outliers .
outlier detection is thus important and many approaches have been proposed in the literature .
although often useful , these methods come with some important drawbacks as well : * as discussed by @xcite , many techniques are limited to situations where the sample size exceeds the dimension , thus excluding modern high - dimensional data analysis .
* several types of outliers exist .
algorithms such as those proposed by @xcite , @xcite and @xcite focus on samples that are potentially mislabeled .
however , not every outlier is a mislabeled observation and vice versa : a sample can be correctly labeled yet behave in a completely different way than its group members .
such discrimination between several types of outliers is usually not provided . *
most algorithms basically provide a ranking of the samples according to potential mislabeling .
however , intuitively it is not always clear how many of the top ranked samples are serious outlier candidates .
automatic cut - off procedures often turn out too conservative ( not detecting all outliers ) or too aggressive ( pointing out good samples as outliers ) . *
the role of the kernel is highly undervalued .
some methods [ @xcite , @xcite ] do not use support vector machines or kernels at all .
@xcite use support vector machines , but restrict themselves to a linear kernel and even a constant regularization parameter , whereas optimization of hyperparameters through cross validation is preferred . in order to avoid some of these difficulties
, we propose an outlier map for svm classification .
outlier maps ( also called diagnostic plots ) are quite common in multivariate statistics , for example , for linear regressionand linear principal component analysis [ @xcite , @xcite ] .
the idea is to start from a robust method guaranteeing resistance to potential outliers .
based on this robust fit , appropriate measures of interest ( e.g. , residuals in regression ) are computed and plotted . in this paper a similar idea is developed for providing an outlier map which is easy to interpret , distinguishes different types of potential outliers , and works for any type of kernel .
on the @xmath0-axis of this map we put the stahel donoho outlyingness . in section [
sect : sd ] we explain how to compute this outlyingness measure in a general kernel induced feature space . on the @xmath1-axis of the outlier map we put the value of the classification function of a trimmed support vector machine .
more details on this robustified svm are given in section [ sect : sdsvm ] .
the main part of the paper is section [ sect : defom ] , where the outlier map is defined and illustrated in a simple two - dimensional example . in section [ sect : examples ] the outlier map is discussed in @xmath2 high - dimensional real life examples .
let @xmath3 be a data set of @xmath4-dimensional samples @xmath5 . in multivariate statistics the stahel
donoho outlyingness of sample @xmath6 [ @xcite , @xcite ] is defined by @xmath7 with @xmath8 a robust univariate estimator of location and @xmath9 a univariate estimator of spread .
popular choices are , for instance , the median for @xmath8 and the median absolute deviation ( mad ) for @xmath9 .
the set @xmath10 is a set of @xmath11 directions in @xmath12 . in practice , this set is often constructed by selecting directions orthogonal to subspaces containing @xmath4 observations if @xmath4 is sufficiently small .
another possibility is taking @xmath11 times a direction through @xmath13 randomly chosen observations .
this strategy works in any dimension @xmath4 and since we will extend the outlyingness to high - dimensional kernel spaces , this is the strategy of our choice .
the stahel
donoho outlyingness plays a crucial role in several multivariate robust algorithms , for example , covariance estimation [ @xcite ] and pca [ @xcite ] .
first we note that this outlyingness measure can be computed in an arbitrary kernel induced feature space .
let @xmath14 be @xmath15 elements in a set @xmath16 .
let @xmath17 be an appropriate kernel function @xmath18 with corresponding feature space @xmath19 and feature map @xmath20 such that the inner product @xmath21 between feature vectors in @xmath19 can be computed by @xmath17 : @xmath22 denote @xmath23 the matrix containing @xmath24 as entry @xmath25 .
this matrix is called the kernel matrix . a typical kernel method such as svm consists of applying a linear method in the feature space @xmath19 such that the computations only depend on pairwise inner products and thus on the kernel matrix [ @xcite ] .
we now show that the stahel donoho outlyingness ( [ eq : stdorig ] ) can be computed in such a manner .
let @xmath26 be the direction in @xmath19 through @xmath13 feature vectors @xmath27 and @xmath28 : @xmath29 the projection of a feature vector @xmath30 onto the direction @xmath26 is then @xmath31 since the squared norm of an element equals the inner product of the element with itself , we have that @xmath32 the vector @xmath33 denotes the vector with entry @xmath34 equal to @xmath35 , entry @xmath36 equal to @xmath37 and all other entries equal to @xmath38 .
then @xmath39 denote @xmath40 the vector containing the projections of all feature vectors onto the direction @xmath26 through feature vectors @xmath41 and @xmath28 : @xmath42 note that only the kernel matrix @xmath23 is needed and not the explicit feature vectors @xmath27 to compute the projections @xmath40 . from these projections
the stahel donoho outlyingness of a feature vector @xmath28 in @xmath19 can be calculated as follows : @xmath43 again @xmath8 and @xmath9 are univariate robust estimators of location and scale . from this point on we always take @xmath44 note that in ( [ eq : stdkern ] ) we have to check @xmath45 directions to find the maximum , where @xmath15 denotes the number of observations in the data set
. then all directions through @xmath13 observations are considered .
if @xmath15 is too large , a random subset of directions can be taken .
typically a few hundred is already enough to provide a good approximation [ @xcite ] . in our implementation
we use the full set if @xmath46 .
otherwise we select @xmath47 directions at random .
let us now turn to the typical svm setup .
let @xmath48 be a data set of @xmath49 training samples in some set @xmath50 and let @xmath17 be a kernel function @xmath51 .
let @xmath52 be the corresponding labels : @xmath53 if sample @xmath34 belongs to the negative group , @xmath54 if sample @xmath34 belongs to the positive group
. denote by @xmath55 the number of samples with label @xmath37 and @xmath56 the number of samples with label @xmath57 .
the following algorithm basically trims a fraction of the data with largest outlyingness and trains a standard svm on the remaining samples
. we will refer to this algorithm as sd - svm ( sd stands for stahel donoho ) : 1 .
set @xmath58 .
denote @xmath59 and @xmath60 ( @xmath61 denotes the largest integer smaller than @xmath62 ) .
trimming step : consider only the inputs with group label @xmath37 . compute the stahel
donoho outlyingness for every sample in this set using ( [ eq : stdkern ] ) .
retain the @xmath63 observations with smallest outlyingness .
denote this set of size @xmath63 as @xmath64 .
analoguously obtain the set @xmath65 containing the @xmath66 samples with group label @xmath57 with smallest outlyingness .
training step : train a standard svm on the reduced training set @xmath67 .
thus solve @xmath68\\[-8pt ] & & \mbox{subject to}\qquad 0\leq\alpha_i\leq c \quad\mbox{and}\quad \sum_{x_i\in t}\alpha_i y_i=0.\nonumber\end{aligned}\ ] ] the classifying function is given by @xmath69 to predict the group membership of a sample @xmath70 , one takes @xmath71 . note
that the computations in the training step are exactly the same as for an ordinary svm .
the only difference is that the reduced set @xmath72 containing the observations with smallest outlyingness is used , in order to avoid negative effects from possible outliers .
the regularization parameter @xmath73 in ( [ eq : sdsvmalpha ] ) is sometimes set to @xmath74 as a default value .
however , it is preferable to optimize the value of @xmath73 .
sd - svm is of course compatible with any type of model selection strategy : it suffices to add the model selection strategy to the training step ( step 3 ) of the algorithm outlined in section [ sect : sdsvm ] . in all the examples of this paper ,
10-fold cross - validation was used to optimize @xmath73 . to illustrate sd - svm ,
consider the following simple experiment : @xmath75 samples ( negative group ) are generated , each with @xmath76 independent standard normal components .
another @xmath75 samples ( positive group ) are generated with @xmath77 independent normal components with mean @xmath78 . in a second setup
the same data is used with additional outliers : @xmath2 samples are added to the negative group with @xmath77 independent normal components with mean @xmath79 . to the positive group , @xmath2 samples are added with @xmath77 independent normal components with mean @xmath80 . in both situations
sd - svm with a linear kernel is applied for several values of @xmath81 .
the fraction of misclassifications on @xmath82 newly generated test data is computed .
figure [ fig : sim ] shows boxplots over @xmath83 simulation runs . .
] in the case without outliers the number of misclassifications increases as @xmath84 decreases .
this is quite expected since a lower @xmath84 means more trimming , which is unnecessary in this case since all samples are nicely generated from two gaussian distributions .
thus , it is no surprise that a classical svm ( @xmath85 ) performs best .
however , a relatively small amount of outliers ( 8 out of 58 ) changes things completely ( right - hand side of figure [ fig : sim ] ) .
a classical svm ( @xmath85 ) is no better than guessing anymore ( more than 50@xmath86 misclassifications ) .
sd - svm with @xmath87 is not good enough either , since the trimming percentage is still smaller than the percentage of outliers .
only if @xmath84 is chosen small enough , good performance is obtained .
thus , a small @xmath84 provides protection against outliers at the cost of a slightly worse classification performance at uncontaminated data . for the outlier map it is most important to avoid the huge effects of outliers , whereas the small effect of unnecessary trimming is practically invisible .
therefore , a default choice of @xmath88 turns out to be a good choice for the construction of the outlier map , and we retain this choice throughout the remainder of the paper .
the following visualization is proposed : 1 .
make a scatterplot of the outlyingness and the value of the classifier @xmath89 .
thus , for @xmath90 , plot pairs @xmath91 where @xmath92 is the stahel donoho outlyingness of sample @xmath36 computed in the trimming step of the algorithm and @xmath93 can be calculated from ( [ eq : sdsvmf ] ) .
2 . plot the inputs with group labels @xmath57 as circles and those with group labels @xmath37 as crosses .
add a solid vertical line at horizontal coordinate @xmath38 .
consider a simple example in @xmath13 dimensions as follows : @xmath94 observations are generated from a bivariate gaussian distribution with mean @xmath95 and identity covariance matrix .
they have group label @xmath37 .
thirty observations are generated from a bivariate gaussian distribution with mean @xmath96 and identity covariance matrix .
they receive group label @xmath57 .
apart from these @xmath97 observations , @xmath98 more are added , representing several types of outliers : @xmath79 data points ( denoted 6163 ) are placed around position @xmath99 with label @xmath57 .
two observations ( denoted 64 and 65 ) with label @xmath57 are placed around @xmath100 .
one point ( denoted @xmath101 ) is placed at position @xmath95 with label @xmath57 . a two - dimensional view of the data is given in figure [ fig : fig1](a ) .
the solid line represents the sd - svm classification boundary with a linear kernel . despite the @xmath98 outliers in the data , sd - svm still manages to separate both groups quite nicely . c -dimensional classification problem .
the solid line is the sd - svm classifying line .
corresponding outlier map visualizing the two main groups and the different types of outliers.,title="fig : " ] + ( a ) + -dimensional classification problem .
the solid line is the sd - svm classifying line .
corresponding outlier map visualizing the two main groups and the different types of outliers.,title="fig : " ] + ( b ) + figure [ fig : fig1](b ) shows the corresponding outlier map . on the vertical axis one reads the stahel
donoho outlyingness .
observations @xmath102 and @xmath103 are positioned in the center of their respective group .
their outlyingness is indeed small .
observations further away from the group center have a larger outlyingness , for example , @xmath104 , @xmath79 and @xmath105 . on the horizontal axis the value of the classifying function @xmath89 as in ( [ eq : sdsvmf ] ) can be read .
the sign of this function determines the predicted group labels .
the vertical line at @xmath106 divides the plot in two parts : every point left of the line is classified into the negative group by sd - svm and every point on the right is classified into the positive group .
we can now see , for instance , that observation @xmath101 is a misclassification : it belongs to the positive group , but receives group label @xmath37 since it lies on the left of the vertical line in figure [ fig : fig1](a ) .
the absolute value of the @xmath1-coordinate in the diagnostic plot represents a distance to the classification boundary . in figure
[ fig : fig1](a ) it can be seen , for example , that observations @xmath79 and @xmath104 are almost equally distant from the negative group center , but observation @xmath79 is much closer to the classification line .
this information can be found in the outlier map in figure [ fig : fig1](b ) as well , since both have almost the same outlyingness ( vertical axis ) , but @xmath79 is much closer to the vertical line than sample @xmath104 ( horizontal axis ) .
the outliers in the data can be detected and characterized too .
observations 6163 are outlying with respect to the other data points in their group , which is clearly indicated by their large outlyingness
. however , both samples still follow the classification rule .
indeed , both are lying on the right side in figure [ fig : fig1](b ) .
samples 64 and 65 , on the other hand , are outlying with respect to the other observations in their group as well as with respect to the classification line : their outlyingness is large and the value of the classification function is negative , although it should have been positive to obtain a correct classification . finally consider observation @xmath101 .
its not extremely outlying with respect to the other data points in the positive group .
however , taking the negative group and the classification line into account , it seems to share more characteristics with the negative group than with its own positive group colleagues .
in the outlier map this is revealed by a moderate outlyingness and by its position almost in the middle of the left side of the vertical line .
the first example considers a data set by@xcite .
the data consist of microarrays from 128 different individuals with acute lymphoblastic leukemia ( all ) , publicly available in the all package in the software environment r. the number of gene expressions at each individual equals @xmath107 .
there are @xmath108 adult patients with t - cell all and @xmath109 with b - cell all .
figure [ fig : leuk ] presents the outlier map for svm with a linear kernel applied to this data set .
it turns out that the data is well classified and that there are no samples with a very large outlyingness .
both t - cell and b - cell form homogeneous groups as one would like when applying a linear svm .
thus , the outlier map immediately shows that the data is clean and one can safely proceed analysis without worrying about outliers .
the breast cancer data set from @xcite contains @xmath110 tumor samples that are either positive ( er@xmath111 ) or negative ( er@xmath112 ) to estrogen receptor .
the expression levels of @xmath113 genes are given for each sample .
for a linear kernel the corresponding outlier map is shown in figure [ fig : breast](a ) .
samples @xmath114 , @xmath115 and @xmath116 immediately catch the eye .
their outlyingness is unusually large . in @xcite samples @xmath114 and
@xmath115 were already rejected and taken out of the analysis due to failed array hybridization .
also sample @xmath116 was characterized as unusual .
it was the only sample in the er@xmath111 group for which the out of sample prediction was highly unreliable in the analysis performed by @xcite .
the samples @xmath103 and @xmath108 attract attention as well .
they have a large outlyingness and both are clearly misclassified .
it turns out that for this data the group membership er@xmath111 or er@xmath112 was determined not only by immunohistochemistry at time of diagnosis , but also by later immunoblotting . for samples @xmath108 and @xmath103 both methods returned different results .
@xcite show via statistical analysis that the initial labeling er@xmath111 for @xmath108 and er@xmath112 for @xmath103 is probably wrong and that the immunoblotting results are more appropriate .
this is clearly confirmed by the outlier map .
it is worth noting that the same data set was analyzed in@xcite , where a comparison was made between a proposed stability criterion , a simple leave - one - out criterion and the algorithm from @xcite .
however , none of these methods was able to detect the @xmath105 clear outliers discussed so far .
five more suspicious samples were indicated in @xcite : @xmath104 , @xmath117 , @xmath118 , @xmath119 and @xmath120 . in figure
[ fig : breast](b ) these samples are shown on a zoom - in from the full outlier map into the region @xmath121 on the vertical axis . except for @xmath104 , these samples are suspicious in the sense that they are not confidently classified , since the value of the classifying function is close to @xmath38 .
it is no surprise that these samples are found by the algorithms compared in @xcite , since those methods are designed to detect potentially mislabeled samples .
also note that some of these mislabeling detection algorithms pointed out samples @xmath122 and @xmath123 as suspicious , although these samples were not considered in @xcite . from the outlier map
it can be seen that @xmath122 and @xmath123 are indeed wrongly classified by sd - svm .
c are outlying but well classified .
samples @xmath108 and @xmath103 are slightly outlying with respect to their groups , but are clearly wrongly classified .
this suggests that they are mislabeled rather than erroneous , confirming the original analysis by west et al .
same plot , but zoomed - in at the region @xmath121 on the vertical axis for better visibility .
observations flagged by algorithms searching for mislabelings are shown @xmath124.,title="fig : " ] + ( a ) + are outlying but well classified . samples
@xmath108 and @xmath103 are slightly outlying with respect to their groups , but are clearly wrongly classified .
this suggests that they are mislabeled rather than erroneous , confirming the original analysis by west et al .
same plot , but zoomed - in at the region @xmath121 on the vertical axis for better visibility .
observations flagged by algorithms searching for mislabelings are shown @xmath124.,title="fig : " ] + ( b ) + the colon cancer data set from @xcite contains @xmath47 gene expression levels for @xmath118 tumor samples and @xmath125 normal samples .
the outlier map with a linear kernel is shown in figure [ fig : alon ] . in the tumor group t2 ,
t33 , t36 and t30 are misclassified .
sample t37 is classified correctly , but with low confidence : it is very close to the classification boundary . in the normal group n8 and especially n34 and n36 are the suspicious cases that behave differently from the other normal samples .
the 8 aforementioned samples plus sample n12 were identified as possible outliers in the original paper by @xcite for biological reasons . thus , 8 out of 9 true outliers can be identified on the outlier map , only leaving n12 undetected . however , in @xcite none of the methods that were compared could detect n12 . moreover , the stability criterion proposed by malossini et al . was unable to detect t37 and n8 too and incorrectly pointed at n2 and n28 as possibly suspicious samples .
also note the interesting sample t6 . from the outlier map
we see that this sample is classified correctly and with much confidence .
nevertheless , its outlyingness with respect to the other tumor samples is rather large .
this means that t6 behaves quite differently than the other tumor samples , but without distorting the classification . in malossini
most of the methods analyzed did not detect t6 at all .
again this is no surprise since methods such as the stability criterion of malossini et al .
specifically focus on mislabeled observations , whereas t6 is certainly not mislabeled .
only the outlier detection method of @xcite is able to detect t6 , but does rather poorly on the other samples detecting only 5 out of 9 true outliers .
the protein data set taken from @xcite contains @xmath126 protein sequences of the essentially ubiquitous glycolytic enzyme 3-phosphoglycerate kinase ( 3-pgk ) in three domains : archaea , bacteria and eukaryota . the data set is available in the protein classification benchmark collection at http://net.icgeb.org ( accession number pcb00015 ) .
we consider here classification task number 10 where the positive group consists of 35 eukaryota .
the negative group consists of 4 archaea and 40 bacteria . to classify these two groups of protein sequences , we use svm with the local alignment kernel [ @xcite ] .
default parameter values were used : gap opening penalty @xmath127 11 , gap extension penalty @xmath127 1 , scaling parameter @xmath127 0.5 .
the outlier map is shown in figure [ fig : protein ] .
one observes that the positive group of eukaryota is very heterogeneous as several clusters appear .
these clusters all have a biological interpretation , as the group of eukaryota contains several subgroups of different phyla .
for instance , observations 2931 are from the phylum of alveolata
. samples 1317 are the euglenozoa . note that 18 ( named q8srz8 ) , which belongs to the fungi , was clustered in the group of euglenozoa by pollack et al . ;
this is actually confirmed by the outlier map .
finally , samples 33 and 34 are outlying with respect to the positive group .
they form , together with 32 , the group of stramenopiles .
note that the different behavior of sample 32 from its fellow stramenopiles is again a confirmation of the analysis by pollack et al .
: their clustering method assigned 32 ( named q8h721 ) in the main group of eukaryota metazoa .
also , in the outlier map 32 is situated in the main group , whereas 33 and 34 form a separate cluster . in the positive group
the heterogeneity is less clear , although the 4 archaea ( 3639 ) do have the largest outlyingness compared to the other samples which are all bacteria .
an outlier map is proposed for support vector machine classification .
if the outlier map shows two homogeneous and well classified groups , one can safely proceed analysis without worrying about outliers .
however , in some situations this may not be the case and the outlier map can be a simple and useful tool to detect this . moreover , the outlier map can be drawn for any choice of kernel , including rather exotic ones such as used in protein analysis .
it can also be helpful to gain insight in the type of outliers , for example , whether outliers are mislabeled observations or not , or whether the outliers are isolated errors or rather a small subgroup of the group structure considered .
this is important to know how to proceed analysis .
if the outliers are truly erroneous observations , one should not take them into account to build a classifier , and one can manually discard them from the data set or apply a robust classifier . if the outliers are mislabeled observations , one probably should re - examine the labeling and change the label of the outlier if this seems indeed appropriate . if the outliers form a small subgroup of the data , one might reconsider the use of a binary classifier and turn to a more appropriate modeling technique . in any event
, the outlier map can be helpful for practitioners of svm classification to make such decisions .
alon , u. , barkai , n. , notterman , d. a. , gish , k. , ybarra , s. , mack , d. and levine , a. j. ( 1999 ) .
broad patterns of gene expression revealed by clustering of tumor and normal colon tissues probed by oligonucleotide arrays .
sci . _ * 96 * 64756750 .
chiaretti , s. , li , x. , gentleman , r. , vitale , a. , vignetti , m. , mandelli , f. , ritz , j. and foa , r. ( 2004 ) . gene expression profile of adult t - cell acute lymphocytic leukemia identifies distinct subsets of patients with different response to therapy and survival . _ blood _ * 103 * 27712778 .
furey , t. s. , cristianini , n. , duffy , d. , bednarski , w. , schummer , m. and haussler , d. ( 2000 ) .
support vector machine classification and validation of cancer tissue samples using microarray expression data
. _ bioinformatics _ * 16 * 906914 .
kadota , k. , tominaga , d. , akiyama , y. and takahashi , k. ( 2003 ) . detecting outlying samples in microarray data : a critical assessment of the effect of outliers on sample classification .
_ chem - bio .
j. _ * 3 * 3045 .
leslie , c. , eskin , e. and noble , w. s. ( 2002 ) .
the spectrum kernel : a string kernel for svm protein classification . in _ proceedings of the pacific symposium on biocomputing 2002 _
( r. b. altman , a. k. dunker , l. hunter , k. lauerdale and t. e. klein , eds . ) 564575 .
world scientific , hackensack , nj . leslie , c. , eskin , e. , weston , j. and noble , w. s. ( 2003 ) .
mismatch string kernels for svm protein classification . in
_ advances in neural information processing systems _ ( s. becker , s. thrun and k. obermayer , eds . ) * 15 * 14411448 . mit press , cambridge , ma .
li , l. , darden , t. a. , weinberg , c. r. , levine , a. j. and pedersen , l. g. ( 2001 ) .
gene assessment and sample classification for gene expression data using a genetic algorithm / k - nearest neighbor method . _
high throughput screen . _ * 4 * 727739 .
liao , l. and noble , w. s. ( 2002 ) . combining pairwise sequence similarity and support vector machines for remote protein homology detection . in
_ proceedings of the sixth international conference on computational molecular biology _ ( t. lengauer , ed . ) 225232 .
acm press , new york .
pochet , n. , de smet , f. , suykens , j. a. k. and de moor , b. ( 2004 ) .
systematic benchmarking of microarray data classification : assessing the role of nonlinearity and dimensionality reduction .
_ bioinformatics _ * 20 * 31853195
. pollack , j. d. , li , q. and pearl , d. k. ( 2005 ) .
taxonomic utility of a phylogenetic analysis of phosphoglycerate kinase proteins of archaea , bacteria and eukaryota : insights by bayesian analyses .
evol . _ * 35 * 420430 . west , m. , blanchette , c. , dressman , h. , huang , e. , ishida , s. , spang , r. , zuzan , h. , marks , j. r. and nevins , j. r. ( 2001 ) . predicting the clinical status of human breast cancer by using gene expression profiles .
_ * 98 * 1146211467 . | support vector machines are a widely used classification technique . they are computationally efficient and provide excellent predictions even for high - dimensional data . moreover , support vector machines are very flexible due to the incorporation of kernel functions .
the latter allow to model nonlinearity , but also to deal with nonnumerical data such as protein strings .
however , support vector machines can suffer a lot from unclean data containing , for example , outliers or mislabeled observations .
although several outlier detection schemes have been proposed in the literature , the selection of outliers versus nonoutliers is often rather ad hoc and does not provide much insight in the data . in robust multivariate statistics outlier maps are quite popular tools to assess the quality of data under consideration .
they provide a visual representation of the data depicting several types of outliers .
this paper proposes an outlier map designed for support vector machine classification .
the stahel donoho outlyingness measure from multivariate statistics is extended to an arbitrary kernel space . a trimmed version of support vector machines is defined trimming part of the samples with largest outlyingness . based on this classifier ,
an outlier map is constructed visualizing data in any type of high - dimensional kernel space .
the outlier map is illustrated on 4 biological examples showing its use in exploratory data analysis . . |
type ia supernovae ( sne ia ) are thought to be a thermonuclear explosion of a c+o white dwarf ( wd ) in a close binary system ( see * ? ? ?
* ; * ? ? ?
* ) . since the ignition is triggered when the wd reaches a mass close to the chandrasekhar limit , the properties of sne ia are expected to be homogeneous .
sne ia are believe to synthesize a large amount of @xmath2ni and other fe - group elements during the explosion , and thus they are one of the primary supply sources of fe - group elements in the universe . in addition , the radioactive decay of @xmath2ni results in an extreme luminosity of sne ia .
this large luminosity , and their relative homogeneity makes it possible to use them as a standardisable candles with which distances can be measured in the distant universe ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , the explosion mechanism of sne ia is still unclear .
although it is widely accepted that the explosion starts as a deflagration , i.e. , a subsonic burning flame @xcite , the subsequent evolution of the flame is under debate : the flame might remain subsonic ( deflagration model , e.g. , @xcite ) , or experiences a transition to the supersonic regime ( delayed detonation model , e.g. , @xcite ) .
these scenarios have been studied in detail with multi - dimensional simulations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
different explosion models predict different element abundance distributions and nucleosynthesis yields .
for example , deflagration models tend to leave more unburned elements than delayed detonation models , where the wd is burned almost entirely . three - dimensional deflagration models predict a mixed abundance distribution , which is not seen in one - dimensional deflagration models @xcite or in one- and multi - dimensional delayed detonation models @xcite .
thus , in order to place constrains on the explosion mechanism , it is important that the abundance distribution and the nucleosynthesis yields are derived from observed sne .
modelling a time series of optical spectra of extragalactic sne is a powerful method to extract the abundance distribution in sne ia . at early epochs ( photospheric phase , @xmath7 days
, hereafter @xmath8 denotes the time since the explosion ) , the optical spectra of sne ia exhibit a pseudo - continuum emission and p cygni profiles superposed onto it .
since the absorption features are formed by line scattering in the outer , optically thin layers , they carry information on the abundances in the outer layers . in addition , since the sn ejecta are expanding , the line - formation region moves inwards with time in the mass and velocity coordinate . following the time evolution of the spectra
, the abundances in different layers can thus be derived . at
late enough epochs ( nebular phase , @xmath9 days ) , a sn shows an emission - line spectrum .
most lines are forbidden lines of heavy elements .
since these lines are formed in the innermost layers of the ejecta , we can derive the abundances in the inner layers , which can not be studied with the photospheric - phase spectra . using the optical spectra , ( * ? ? ?
* paper i ) derived the abundance stratification in sn 2002bo .
sn 2002bo is a normally luminous type ia sn ( with a b - band light curve ( lc ) decline rate @xmath10 mag ) , but has higher line velocities , i.e. , larger doppler shifts at the absorption minima , than canonical sne ia @xcite .
paper i found that a large abundance of @xmath2ni and intermediate mass elements ( imes ) such as si and s is required in the outer layers , and pointed out that mixed - out @xmath2ni can explain the fast rise of the lc .
* paper ii ) studied sn 2004eo .
sn 2004eo has a transitional luminosity between normal and subluminous sne ia and has a relatively rapid decline rate of the light curve ( @xmath11 mag , * ? ? ?
a relatively small mass of fe - group elements was obtained for sn2004eo in paper ii , suggesting that this is responsible for the rapid decline of the lc . in this paper
, we study the abundance distribution in sn 2003du @xcite .
sn 2003du is a very normal type ia sn both photometrically and spectroscopically , and it can therefore be considered as a representative of normal sne ia .
sn 2003du is classified as low velocity gradient ( lvg ) according to the scheme of @xcite , or as core - normal according to the scheme of @xcite .
this sn is an extremely well - studied object at optical and near infrared ( nir ) wavelengths , with an extremely good time sampling , and one of the rare cases in which nir spectra at late epochs have been obtained @xcite .
this paper is organized as follows . in section [
sec : method ] , the methods of analysis are described .
the spectra at the photospheric and nebular phases are modelled in sections [ sec : early ] and [ sec : late ] , respectively . in section [ sec : lc ] , the derived abundance distribution is tested against the bolometric lc .
the derived abundance distribution and the integrated yields are shown in section [ sec : abun ] . in section [ sec : con ] we give conclusions . throughout the paper ,
the distance to sn 2003du is assumed to be @xmath12 mag @xcite .
the reddening in our galaxy and in the host galaxy are @xmath13 mag and @xmath14 mag , respectively @xcite .
we adopt a bolometric rise time of 19 days , which was derived from the lc synthesis @xcite and is consistent with the @xmath15-band rise time of 20.9 days . the mass and mass fraction of radioactive @xmath2ni [ @xmath16(@xmath2ni ) and @xmath17(@xmath2ni ) , respectively ] are expressed by the value at the explosion ( @xmath18 ) .
@xmath16(fe ) [ @xmath17(fe ) ] and @xmath16(ni ) [ @xmath17(ni ) ] only represent the mass [ mass fraction ] of stable fe and ni , respectively .
the photospheric - phase spectra are studied using the monte carlo spectrum synthesis code developed by @xcite .
assumptions made in the code are described in appendix [ app : early ] . for more details , see @xcite .
the code has been used for the modelling of various types of sne ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the code calculates a synthetic spectrum for given inputs : ( 1 ) a radial density profile , ( 2 ) a position of the photosphere in velocity ( i.e. , photospheric velocity @xmath19 ) , ) at the time of observations , the velocity can be used as a radial coordinate .
as the velocity is unchanged with time , it is more convenient than the radius itself . ]
( 3 ) an emergent luminosity @xmath20 and ( 4 ) element abundances above the photosphere . for the density structure ,
we use the deflagration model w7 @xcite .
the validity of this choice is discussed in section [ sec : abun ] .
the other parameters are optimized so as to give the best fit of the observed spectra .
the luminosity @xmath20 is most strongly constrained by the observed absolute flux .
the photospheric velocity @xmath19 is estimated primarily by using the overall colour of the observed spectra via the relation @xmath21 , where @xmath22 and @xmath23 are the effective temperature and the stefan - boltzmann constant , respectively .
the estimated velocity is further checked based on the reproduction of the observed line velocities ( i.e. , doppler shift at the absorption minimum of p cygni profiles , see fig .
[ fig : vel ] ) . with the estimated @xmath20 and @xmath19 ,
the abundance ratios above the photosphere are optimized to reproduce the observed spectra .
we start modelling from earlier spectra for abundance stratification because the photosphere recedes with time in the mass and velocity coordinates .
suppose that the photosphere recedes from @xmath24 to @xmath25 ( @xmath26 ) in the time interval from @xmath27 to @xmath28 ( @xmath29 ) .
first , the abundances at @xmath30 are optimized by fitting the observed spectrum at @xmath31 . at the later phase ( @xmath32 , @xmath33 ) , the inner part begins to make a contribution to the spectrum .
however , even at @xmath32 , the outer layers at @xmath30 still contribute to the spectrum .
thus , we keep the abundance ratios at @xmath30 and optimize the abundances at @xmath34 by fitting the observed spectrum at @xmath32 .
fitting the time series of spectra , we get an abundance distribution as a function of velocity .
a nebular - phase spectrum is modelled under different assumptions from those made for the photospheric - phase spectra .
the assumptions made in our analysis are described in appendix [ app : late ] . for more details , see e.g. , @xcite and @xcite .
this code has been used for the modelling of supernovae ( e.g. , * ? ? ?
the inputs are only the radial density profile and the element abundance distribution .
for the density structure , we use the w7 model as in the photospheric phase . for the abundance distribution ,
the results of the photospheric - phase modelling are used in the outermost layers ( @xmath35 ) . at intermediate layers ( @xmath36 @xmath37 ) ,
the abundance distribution is constrained both from photospheric - phase spectra and nebular - phase spectra .
the abundances in the inner layers ( @xmath38 ) are constrained only from the nebular - phase spectra .
the modelling is iterated until a consistent distribution is obtained .
the total mass and distribution of @xmath2ni are especially important for the modelling of nebular spectra because they determine the emergent luminosity and the energy deposition in each layer .
the total energy emitted by forbidden lines of fe is sensitive to the mass of @xmath2ni as well as to the total amount of fe ( i.e. , the sum of stable fe and the decay product of @xmath2ni ) , which affects in particular the ionization of fe .
thus , although @xmath2ni is mostly fe at the nebular epoch , stable fe and @xmath2ni products can be distinguished , albeit only indirectly .
rrrrrr epoch@xmath39 & @xmath8 & @xmath40 & @xmath19 & @xmath22 & @xmath41 + ( day ) & ( day ) & ( erg s@xmath42 ) & ( ) & ( k ) & ( k ) + -12.8 & 8.1 & 42.43 & 11500 & 8900 & 12700 + -10.9 & 10.0 & 42.69 & 10500 & 9500 & 13500 + -7.8 & 13.1 & 42.97 & 9800 & 10100 & 13800 + -5.8 & 15.1 & 43.08 & 9400 & 10400 & 13800 + [ 2.5pt ] -4.0 & 16.9 & 43.11 & 9050 & 10300 & 13400 + -1.9 & 19.0 & 43.17 & 8900 & 10000 & 12500 + 0 & 20.9 & 43.18 & 8500 & 9800 & 12000 + + 1.2 & 22.1 & 43.18 & 8200 & 9700 & 11800 + [ 2.5pt ] + 4.3 & 25.2 & 43.12 & 7700 & 9000 & 10700 + + 7.2 & 28.1 & 43.08 & 7200 & 8600 & 10000 + + 10.0 & 30.9 & 43.00 & 6600 & 8200 & 9400 + + 17.2 & 38.1 & 42.78 & 4000 & 8400 & 9800 + + @xmath39 days since @xmath15 maximum
twelve photospheric - phase spectra of sn 2003du presented by @xcite are modelled .
the input parameters ( @xmath20 and @xmath19 ) are summarized in table [ tab : param ] .
the effective temperature @xmath22 is calculated by the equation @xmath43 .
the blackbody temperature @xmath41 calculated with the code is also shown in table [ tab : param ] .
the blackbody temperature is defined by @xmath44 , where the luminosity @xmath45 in this equation is not the emergent luminosity but the luminosity emitted from the photosphere , including photons scattered backward into the photosphere @xcite .
thus , @xmath41 is always higher than the effective temperature ( @xmath22 ) . in fig .
[ fig : vel ] , the position of the photosphere is shown by red circles .
the observed line velocities of si ii @xmath466355 , si iii @xmath464560 , and s ii @xmath465640 are also shown for comparison @xcite .
the time evolution of the estimated photospheric velocity is similar to that of the si iii line velocities , which is the slowest among the three lines .
this indicates that the slowest line velocity is the best observational indicator of the photospheric velocity .
note that the estimated photosphere after maximum ( @xmath47 days ) is not very certain because the assumptions of the code become less reliable at later epochs ( see appendix [ app : early ] ) . in the following sections , the results of modelling at pre - maximum epochs ( from @xmath48 to @xmath49 days since @xmath15 maximum ) , maximum epochs ( from @xmath50 to @xmath51 days ) and post - maximum epochs ( from @xmath52 to @xmath53 days ) are presented .
days after the explosion ( around maximum ) . ] in fig .
[ fig : spec1 ] , the observed spectra at epochs from @xmath48 to @xmath49 days since @xmath15 maximum ( black ) are compared to the synthetic spectra ( red ) .
the observed spectra at these epochs show a relatively well defined pseudo - continuum and p cygni profiles of imes and fe - group elements . at these epochs ,
the photospheric velocity decreases from 11500 to 9400 . _ fe - group elements _ : the presence of the fe , co and ni lines clearly indicates that some amount of these fe - group elements are located in the outer layers .
especially , since only a small amount of @xmath2ni has decayed into fe at such early epochs , the presence of fe lines requires stable fe in the outer layers @xcite . the mass fraction of stable fe is found to be @xmath17(fe ) @xmath54 - @xmath55 .
_ calcium _ : in the spectrum at @xmath56 days from @xmath15 maximum , the ca iilines near 4000 ( ca ii h&k ) and 8000 ( ca ii ir triplet ) show an extremely high velocity ( @xmath57 ) . the origin of this line has been discussed , including the interaction with the circumstellar matter ( csm ) or high velocity blobs ( e.g. , * ? ? ? * ; * ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in this paper
, we do not consider the density enhancement with respect to the w7 model .
keeping the w7 density structure , a mass fraction as high as @xmath17(ca ) @xmath58 is required at @xmath59 .
the density at the high velocity layers at @xmath60 in delayed detonation models is higher than in the w7 model by a factor of @xmath61 ( see e.g. , fig . 11 of @xcite ) .
@xcite studied the influence of the density structure at the outermost layers .
they found that the required mass fraction of ca is decreased by a factor of @xmath61 in the delayed detonation models , but it is still higher than solar abundance ( see their table 3 ) .
the density enhancement may also possibly be caused by interaction with the csm @xcite . _ magnesium , silicon , and sulphur _ : since the strongest mg feature at 4200 is contaminated by fe lines , the mass fraction of mg is not as well constrained as those of si , s , and ca .
if we assume that the fe abundance is constrained from the features around 4700 , and we attribute any differences to mg , the mass fraction of mg is @xmath17(mg ) @xmath62 at @xmath63 .
sn 2003du has relatively narrow si ii lines . to reproduce the profile of the si ii @xmath466355 line
, the mass fraction of si is required to be @xmath17(si ) @xmath64 . the dominant element in the layers at
@xmath5 is not si , but more likely o. the distribution of si is discussed in appendix [ app : si ] .
the mass fraction of s is @xmath17(s ) @xmath65 around @xmath66 .
although the synthetic spectra for the two earliest epochs show stronger lines than the observations , the later spectra are nicely reproduced .
_ carbon and oxygen _
: there is no clear c line in the observed spectra except for the possible c ii @xmath466578 line at the emission peak of the strongest si ii @xmath466355 line @xcite .
we derive an upper limit for the c mass as @xmath16(c ) @xmath67 0.016 @xmath6 at @xmath68 ( details are presented in appendix [ app : c ] ) .
the small mass fraction and mass of c are consistent with the results of previous works @xcite .
the synthetic spectra show a clear o i @xmath467774 line around 7500 while the presence of such a strong o line is not that clear in the observed spectra .
although the abundance of o is not well constrained , considering the lower abundance of c , mg , si and s ( see above ) , we conclude that the dominant element in the outer layers is o. the mass fraction of o at @xmath69 is @xmath17(o ) @xmath70 , increasing towards the outer layers . in fig .
[ fig : spec2 ] , the observed spectra at the epochs from @xmath71 to @xmath51 days since @xmath15 maximum ( black ) are compared to the synthetic spectra ( red ) .
the element features present in the observed spectra are similar to those in the pre - maximum spectra . at these epochs ,
the photospheric velocity decreases from 9050 to 8200 .
the flux of the synthetic spectra at the line - free region at @xmath72 tends to be overestimated because of our assumption of blackbody emission ( see appendix [ app : early ] ) . _ fe - group elements _ : around maximum
, lines of fe - group elements are present , as in the pre - maximum epoch . a larger amount of @xmath2ni is required compared with the pre - maximum epochs to block the emission in the near - uv more effectively .
the mass fraction of @xmath2ni is @xmath17(@xmath2ni ) @xmath73 at @xmath74 . at maximum epochs
, the presence of fe lines does not directly indicate the presence of stable fe because of the high abundance of the decay product of @xmath2ni . at @xmath15 maximum ( @xmath75 days ) , about 10 % of @xmath2ni has already decayed into fe , and thus an fe abundance of @xmath76 is the product of the decay of @xmath2ni .
this is much larger than the abundance of stable fe required for the pre - maximum spectra , and thus we can not distinguish the contribution of stable fe .
_ silicon and sulphur _ : the si ii @xmath466355 line becomes narrower .
the si distribution at the outer layers derived from the line profile in the pre - maximum spectra ( appendix [ app : si ] ) nicely reproduce the maximum spectra . in the spectrum at maximum ,
the line ratio of two si ii lines ( @xmath465972 and 6355 , @xmath77(si ii ) in @xcite ) is also nicely reproduced , suggesting the estimates of the temperature and of the ionization are reasonable ( table [ tab : param ] , @xcite ) .
the s ii lines around 5500 are also reproduced nicely .
although the photospheric velocity at this epoch is @xmath19@xmath78 , the s iilines , as well as the si ii lines , are mainly formed at @xmath79 ( see fig .
[ fig : vel ] ) .
the `` detachment '' of the lines is caused by the ionization effect , e.g. , s iii is dominant near the photosphere while the fraction of s ii increases towards higher velocities ( see fig . 9 of @xcite ) . in fig .
[ fig : spec3 ] , the observed spectra at epochs from @xmath52 to @xmath53 days since @xmath15 maximum are compared to the synthetic spectra . at these epochs ,
the contribution from fe - group elements is larger than at earlier epochs .
the assumption of blackbody emission becomes unreliable , and results in an increasing overestimation of the flux at @xmath72 .
although the photospheric velocity decreases from 7700 to 4000 in our models ( table [ tab : param ] ) , these values are therefore not strongly constrained .
since the assumption that there is no energy deposition above the photosphere is also unlikely to hold at this epoch , we use the abundance distribution derived from the modelling of the nebular spectrum ( section [ sec : late ] ) , rather than optimize the abundances using the spectra at post - maximum epochs . even with this treatment , the synthetic spectra give reasonable agreement with the observed spectra .
_ fe - group elements _ : the @xmath2ni distribution derived from the nebular spectrum reproduces the ni , co and fe lines present in the observed spectra .
the ni ii , co ii and fe ii lines are responsible for the strong blocking of the uv flux . the layers below 8200 down to 4000 ( the photospheric velocity of the last spectrum , fig [ fig : spec3]d ) are rich in @xmath2ni , which has a mass fraction of @xmath17(@xmath2ni)=@xmath80 .
note that the fe lines at @xmath81 in the synthetic spectrum at @xmath82 days are too strong . the discrepancy is also seen in the model for sn 2004eo ( paper i ) , and it may result from our assumption of no energy deposition in the atmosphere ( see above ) . _
sodium , silicon , sulphur , and calcium _ : in the last spectrum ( fig .
[ fig : spec3]d ) , there is a strong absorption at 5600 in the observed spectrum , which is expected to be the na i d line . with our code , however , this line is not reproduced even if we set @xmath17(na ) = 0.1 .
this is already too high compared with the prediction by the explosion models : the yield of na is only @xmath83 ( e.g. , * ? ? ?
this is the result of the overestimate of na ionization by our monte carlo code , which is a problem also for other nlte codes ( * ? ? ?
* and references therein ) , and it may be caused by inaccuracies in the atomic data .
thus , we do not discuss total mass of na in the following section ( section [ sec : abun ] ) .
the strong si ii and s ii lines are reproduced well in the spectra at @xmath52 and @xmath84 days from @xmath15 maximum ( fig .
[ fig : spec3]a and [ fig : spec3]b ) .
these lines are formed in layers detached from the photosphere ( see fig .
[ fig : vel ] ) because of the high mass fractions of si and s , and the large fraction of si ii and s ii in layers above the photosphere .
the ca ii h&k and ir triplet lines in the synthetic spectra show too strong high velocity component .
they are caused by the high abundance at @xmath85 derived from the earliest spectra .
we believe that the constraints from the earlier spectra are more reliable because the validity of the assumptions in the atmosphere is more certain at those epochs .
days from @xmath15 maximum , black line , @xcite ) compared with the synthetic spectra ( color lines ) . the red solid line shows the best fit model while the blue dashed line shows the model without central region devoid of @xmath2ni . ] a spectrum at @xmath86 days since @xmath15 maximum @xcite is modelled to derive the abundances in the deepest layers .
a @xmath2ni distribution with a decreasing mass fraction toward higher velocities ( @xmath87 ) nicely fits the profile of the two strongest fe lines ( red line in figure [ fig : late ] ) .
this is consistent with the constraints from the spectra at the pre - maximum and maximum epochs .
the innermost layers are dominated by stable fe .
in particular , stable fe and ni dominate the abundance at velocities inside of about 3000 , accounting for a total mass of @xmath1 .
most of this is actually stable fe .
limits to the stable ni mass can be set by the weakness of the [ ni ii ] emission near 7400 . as for stable fe , this is required in the innermost regions in order to keep an ionization balance between fe ii and fe iii which allows the shape of the two strong , fe - dominated emissions at @xmath81 to be reproduced , as well as their ratio .
the bluer emission , near 4600 , is in fact dominated by fe iii , while the redder one , near 5200 , is dominated by fe ii .
if a central region devoid of @xmath2ni is absent ( blue dashed line ) , both fe features look very sharp , unlike the observations . in order to reduce the flux , the @xmath2ni distribution must be limited to a smaller region , and the overall higher density of the inner region leads to a lower fe ionization , so that the ratio of the two fe lines in the blue is no longer correct : the fe ii - dominated line ( near 5200 ) becomes too strong , as shown in the model in blue dashed line in figure [ fig : late ] . on the other hand ,
the presence of stable fe acts as a coolant , keeping the ionization degree to a level which yields reasonable line intensities ( red line ) .
the derived abundance distribution is tested against the bolometric lc @xcite .
sn 2003du is an extremely well - observed sn , and the bolometric lc was constructed with a small uncertainty , including the contribution of the nir @xcite . for the modelling
, we use a monte carlo lc code based on that developed in @xcite .
the code follows the emission and propagation of @xmath88-rays and positrons from @xmath2ni and @xmath2co decay .
both @xmath88-rays and positrons are treated with an effective opacity ( i.e. , the positrons are not assumed to deposit in situ ) .
the opacities adopted are 0.027 @xmath89 gm@xmath42 for the @xmath88-rays and 7 @xmath89 gm@xmath42 for the positrons @xcite . for optical opacity
, we use an analytic formula introduced by @xcite , which includes the composition dependence ( fe - group elements and other elements ) and a time dependence @xcite .
[ fig : lc ] shows the comparison between the observed bolometric lc ( black squares ) and the computed lc ( blue line ) . the input luminosities for the photospheric - phase spectra are also plotted ( red circles ) .
the model lc matches well with the observed lc from before maximum until @xmath90 year .
this suggests that the derived abundance distribution , especially that of @xmath2ni , is reasonable .
the bolometric lc derived from the synthetic spectra at epochs after maximum deviate progressively from the observational points , reflecting the increased production of spurious flux redwards of 6500 in the synthetic spectra ( see sect .
the abundance distribution derived from the modelling of the photospheric- and nebular - phase spectra is shown in figs . [
fig : abunmass ] and [ fig : abunvel ] .
these two figures show the mass fraction of each element as a function of mass and velocity , respectively . for comparison , the abundance distribution of a deflagration model ( w7 , * ?
* ) is also shown ( figs [ fig : abunmass]b and [ fig : abunvel]b ) .
note the line with `` @xmath2ni '' represents the mass fraction of radioactive @xmath2ni at the time of the explosion ( @xmath18 ) , and that `` fe '' and `` ni '' denotes only stable elements . at the innermost layers ( @xmath91 , @xmath92 ) , stable fe dominates ( section [ sec : late ] ) .
this is also suggested by the nir spectrum at late phase @xcite .
the existence of stable elements at the innermost layers is consistent with the explosion model .
these elements are synthesized via electron capture reactions at high densities , and thus are typically located in the innermost layers . above the layer of stable elements , there is a @xmath2ni -dominated layer ( @xmath93 , @xmath94 ) .
this is fairly consistent with the distribution of w7 .
however , in sn 2003du a moderate amount of @xmath2ni is distributed also in the outer layers , which is important for the blocking of uv light in the early spectra .
this was also true for sn2002bo ( paper i ) . at @xmath95 ( @xmath96 ) , there is @xmath97 of @xmath2ni .
the layers dominated by si and s are located above the @xmath2ni -rich layers . compared with w7
, si has a broader distribution in the mass / velocity coordinates . the moderate amount of si in the @xmath2ni -rich layers is required to explain the si absorption line in the spectra after maximum . even at the outer layers at @xmath98 ( @xmath99 ) , @xmath17(si ) @xmath100 is also required ( appendix [ app : si ] ) .
this is in contrast to the w7 deflagration model , where c+o remain unburned at @xmath101 . above the si and s - rich layers , the dominant element is o ( @xmath102 , @xmath103 ) .
although this is not directly constrained , a high abundance of si would not be consistent with the observed line profile ( appendix [ app : si ] ) and the mass fractions of c , mg , s , and ca are also not larger than o. the mass of unburned c is quite small , @xmath16(c ) @xmath104 at @xmath68 ( appendix [ app : c ] ) .
the small c mass indicates that the progenitor c+o wd was almost completely burned . in the outermost layers ( @xmath99 , @xmath98 ) only c - burning occurs , and elements heavier than o are not synthesized effectively .
note that there is no constraint on c above @xmath105 , and a mass fraction as high as @xmath106 can be allowed .
( see appendix [ app : c ] ) .
it is also possible that the c abundance in the outermost layers of the progenitor wd is already suppressed in the pre - sn stage . in the single - degenerate scenario ,
the wd grows by mass accretion from a companion star . in the case of weak
he flashes , their nucleosynthesis products can accrete onto the wd
. since the products of the he flashes are not always c and o but could also be heavier elements ( e.g. , * ? ? ?
* ) , the outermost layers of the wd do not necessarily consist only of nearly equal mass fractions of c and o. the heavier elements synthesized in the pre - sn stage can also be responsible for the high velocity features of the si and ca lines present in the earliest spectra @xcite .
the overall abundance distribution is not consistent with the almost fully mixed distribution seen in three - dimensional deflagration models @xcite .
but the boundaries between stable fe/@xmath2ni , @xmath2ni / ime , ime / o - rich layers are less clear than in the one - dimensional models .
this requires moderate degree of mixing during the explosion @xcite .
the sensitivity of the results on the parameters has been studied in paper ii . in the photospheric phase ,
the synthetic spectra are sensitive to the velocity at the inner boundary and the element mass fractions .
variations in the mass fractions of up to 25 % are possible , while the synthetic spectra are more sensitive to variations of the velocity : a change of more than 15% can noticeably affect the fits ( see paper ii ) . in the nebular phase ,
no given inner boundary is required and the constraints on the mass fractions are less uncertain . except for the elements shown in figures [ fig : abunmass ] and [ fig : abunvel ] , there must be some amount of ti and cr in the ejecta . for photospheric - phase spectra ,
these elements are responsible for the features or the flux level around @xmath107 .
the mass fraction of the sum of ti and cr is about @xmath108 ( papers i and ii ) .
however , since this wavelength range is not covered in most spectra used in this paper , we refrain from discussing the distribution and total yields of these elements . [ cols="<,>,>,>,>,>,>,>",options="header " , ] + @xmath109upper limit at @xmath110 .
@xmath111not constrained .
@xmath112not given .
+ @xmath113stable isotopes except for @xmath2fe from @xmath2co decay .
+ @xmath114only @xmath115fe .
+ @xmath116the sum of fe - group elements except for @xmath115fe , and thus , the upper limit for @xmath2ni ( see * ? ? ? * ) .
+ @xmath117stable isotopes @xmath118ni , @xmath119ni , @xmath120ni , @xmath121ni , and @xmath122ni .
+ @xmath123@xcite , @xmath124@xcite , @xmath125@xcite , @xmath126@xcite , @xmath127@xcite the ejected mass of each element in sn 2003du was calculated by integrating the abundance distribution assuming the w7 density structure .
results are summarized in table [ tab : mass ] .
for comparison , the nucleosynthetic yields of the following models are also shown : one - dimensional deflagration model ( w7 , * ? ? ?
* ) , one - dimensional delayed detonation models ( wdd1 and wdd2 , and @xmath128 for wdd1 and wdd2 , respectively . ] @xcite ) , three - dimensional deflagration model ( b30_3d_768 , * ? ? ?
* ) , two - dimensional detonating failed deflagration model ( y12 , * ? ? ? * ) , and two - dimensional off - center delayed detonation model ( o - ddt , * ? ? ?
* ) . in figure
[ fig : yield ] , the derived yields are compared with those of the one - dimensional models in the abundance ratio to fe , normalized by the solar abundances , i.e. , @xmath129 = { \rm log}_{10}[m / m({\rm fe } ) ] - { \rm log}_{10}[x / x({\rm fe})]_{\odot}$ ] . here , the yields after the radioactive decays are shown , i.e. , fe means the sum of stable fe and the decay product of @xmath2ni , and ni includes only stable ni . in sn 2003du , the ejected mass of neutron - rich elements is estimated to be @xmath1 . they are synthesized by electron capture reactions and mostly distributed in the innermost layers .
the mass of @xmath2ni is @xmath3 , which is the largest among all synthesized elements .
the mass of imes ( mg , si , s , and ca ) is @xmath130 in total .
the mass of o is estimated to be @xmath131 , and is predominantly located in the outer layers .
there is no sign of c , and the upper limit is @xmath16(c ) @xmath104 at @xmath68 . from these yields ,
the validity of the w7 density structure can be tested .
we use the formula for the kinetic energy of sne ia : @xmath132 \times 10^{51}$ ] erg @xcite , where @xmath16(@xmath2ni ) , @xmath16(stable fe ) and @xmath16(ime ) are the ejected mass of @xmath2ni , stable fe - group elements and the imes , respectively . substituting the element masses derived for sn 2003du
, we derive an ejecta kinetic energy of @xmath133 erg , which is in a good agreement with the kinetic energy of the w7 model ( @xmath134 ) .
thus , the element masses derived from the modelling under the assumption of the w7 density structure are self - consistent .
we first compare the integrated yields with those of one - dimensional models ( table [ tab : mass ] ) .
the total mass of stable fe - group elements estimated for sn 2003du is reasonably consistent with that of the one - dimensional explosion models .
the amount of neutron - rich stable elements in the central region is sensitive to the deflagration speed , which affects the efficiency of the electron captures .
thus , the good agreement implies that the deflagration speed in the explosion models is reasonable .
the mass of @xmath2ni in sn 2003du is also consistent with that of the one - dimensional deflagration and delayed detonation models . in the delayed detonation models ,
a larger amount of @xmath2ni suggests an earlier transition to a detonation .
thus , if the delayed detonation model is the correct scenario , a transition density close to that of the models wdd1 and wdd2 is reasonable for sn 2003du . the total mass of imes estimated for sn 2003du is smaller than in the delayed detonation models , while the mass of o in sn 2003du is larger .
this suggests that o burning in sn 2003du was not as strong as in the delayed detonation models .
however , the unburned c mass in the outermost layers in sn 2003du is smaller than in the deflagration models .
thus , the outermost part of the c+o wd is burned more intensely than in the deflagration model ( see also section [ sec : distribution ] ) .
next , we compare the results with multi - dimensional models .
the mass of @xmath2ni in sn 2003du is clearly larger than in the three - dimensional deflagration model @xcite .
the three - dimensional deflagration model also predicts more unburned c , which is not consistent with the small c mass derived for sn 2003du .
thus , also from the nucleosynthetic point of view , the three - dimensional deflagration model is unlikely to be a reasonable model for a normal sne ia such as sn 2003du .
it is interesting to compare the yields with the detonating failed deflagration model @xcite and the off - center delayed detonation model @xcite , because these models may explain the kinematic offset in the central region suggested by the nir spectra at late phases @xcite .
the detonating failed deflagration model @xcite predicts too much @xmath2ni compared with sn 2003du .
thus , this model does not seem a likely model for normal sne ia , as already cautioned by @xcite .
the yields of the two - dimensional off - center delayed detonation model @xcite are in good agreement with those derived for sn 2003du .
it is , however , noted that the model predicts @xmath2ni synthesized near the center , while the existence of a region free of @xmath2ni is suggested from our one - dimensional analysis of the nebular phase spectrum ( section [ sec : late ] ) . in this section , we compare the abundance distributions and integrated yields derived for sn 2003du with those derived for sne 2002bo ( paper i ) and 2004eo ( paper ii ) .
sn 2002bo is a normally luminous sn ( @xmath10 mag ) , but has high line velocities and a high velocity gradient at early phases @xcite .
sn 2004eo has a transitional luminosity between normal and subluminous sne ia , with a relatively rapid decline of the light curve ( @xmath11 mag , * ? ? ?
these three sne have in common an innermost layer of stable fe - group elements which are synthesized by electron capture .
the mass of this region @xmath135 is similar in the three sne studied thus far ( see also * ? ? ?
* ) , although the mass of @xmath2ni is different , 0.65 , 0.52 , and 0.43 @xmath6 for sne 2003du , 2002bo , and 2004eo , respectively .
the most significant difference between sne 2003du and 2002bo is in the outermost layers . in sn 2002bo si
reaches the outer layers ( paper i ) .
the abundance of si in the outer layers of sn 2002bo is larger than in sn 2003du .
this is the primary reason why si line velocities in sn 2002bo are higher than in sn 2003du , as demonstrated in appendix [ app : si ] .
@xcite also suggested that the difference between high - velocity and low - velocity sne ia is caused by the difference in the si abundance .
alternatively , the difference can also be caused by the difference in the density structure in the outer layers .
@xcite pointed that the difference in the velocity ( and/or velocity gradient ) is caused by viewing - angle effects in an aspherical explosion that has different density structures depending on the viewing angle .
sn 2002bo also has a more extended @xmath2ni distribution than in sn 2003du .
this explains the rapid rise of the lc of sn 2002bo ( paper i ) .
@xcite suggest that the rising time of the lc of sne with a high velocity gradient is generally shorter than that of sne with a low velocity gradient .
thus , the extended @xmath2ni distribution may be a common property of high - velocity sne .
interestingly , sne 2003du and 2004eo have a similar distribution of si , o , and c above @xmath136 ( @xmath137 ) .
this indicates that the different shapes of the lcs of these two sne result from the difference in the inner layers .
in fact , a difference between these sne is seen in the mass and mass fraction of @xmath2ni .
a smaller amount of @xmath2ni is responsible for the fainter luminosity and more rapid decline of sn 2004eo ( see also paper ii ) .
we have studied the element abundance distribution in sn 2003du by modelling the optical spectra at the photospheric and nebular phases . these are used to place constraints on the abundances at the outer and inner layers , respectively . since sn 2003du is a normal type ia sn , both photometrically and spectroscopically ,
the abundance distribution derived for sn 2003du can be considered as representative of normal type ia sne .
the abundance distribution and the integrated yields in sn 2003du can be summarized as follows .
the innermost layers are dominated by stable , neutron - rich elements , with a total mass of @xmath1 .
the yield is roughly consistent with the results of one - dimensional models .
this may imply that the deflagration speed assumed in the models is reasonable . above the stable elements ,
there is a thick @xmath2ni - rich layer .
the total mass of @xmath2ni is @xmath3 , which is consistent with the one - dimensional deflagration model and the one - dimensional delayed detonation models with transition density @xmath138 .
the @xmath2ni mass is larger than in the three - dimensional deflagration model and smaller than in the detonating failed deflagration model .
the layers dominated by si and s are located above the @xmath2ni -rich layers . the dominant element at @xmath4 ( @xmath5 ) is o. the progenitor c+o wd is almost entirely burned , and the mass of leftover unburned c is quite small ( @xmath104 at @xmath110 ) .
the small mass of unburned c is consistent with the delayed detonation model .
however , the mass of o derived for sn 2003du is larger than that in the delayed detonation model .
these results suggest that in the outermost layers c - burning is more intense than in the deflagration model , but o burning is not as strong as in the delayed detonation models .
the boundary between stable fe/@xmath2ni , @xmath2ni / ime , ime / o - rich layers is not as sharp as in the one - dimensional models , suggesting a moderate degree of mixing .
however , the strong mixing seen in the multi - dimensional deflagration model is not preferable .
this work has made use of the suspect supernova spectral archive .
is financially supported by fct portugal under program cincia 2008 .
this research has been supported in part by world premier international research center initiative , mext , japan .
99 natexlab#1#1 d. c. , lucy l. b. , 1985 , , 288 , 679 g. et al .
, 2007 , , 475 , 585 g. c. , sahu d. k. , jose j. , 2005 , , 429 , 667 t. s. , 1980 , phd thesis , aa(california univ . , santa cruz . ) , e. , bongard , s. , branch , d. , hauschildt , p. h. , 2006 , , 645 , 480 s. et al . , 2005 , , 623 , 1011 s. et al . , 2004 , , 348 , 261 d. et al . , 2006 , , 118 , 560 d. et al . , 2007 , , 119 , 709 e. , mazzali p. a. , benetti s. , danziger i. j. , turatto m. , della valle m. , patat f. , 1997 , , 328 , 203 j. i. , 1970 , , 149 , 111 n. et al .
, 2006 , , 369 , 1880 v. n. , khokhlov a. m. , oran e. s. , 2005 , , 623 , 337 g. et al . , 2007 , , 471 , 527 c. l. et al .
, 2004 , , 607 , 391 g. a. , 1997 , the physics and dynamics of planetary nebulae .
springer - verlag berlin heidelberg new york s. , mazzali p. a. , tanaka m. , hillebrandt w. , benetti s. , 2008 , , 389 , 1087 m .- a .
, hanawa t. , sugimoto d. , 1983 , , 35 , 1 k. , branch d. , fisher a. , baron e. , filippenko a. v. , 1999 , , 525 , 881 w. , niemeyer j. c. , 2000 , , 38 , 191 p. , khokhlov a. , 1996 , , 457 , 500 p. , gerardy c. l. , nomoto k. , motohara k. , fesen r. a. , maeda k. , ohkubo t. , tominaga n. , 2004 , , 617 , 1258 k. , brachwitz f. , nomoto k. , kishimoto n. , umeda h. , hix w. r. , thielemann f .- k . , 1999 , , 125 , 439 iv g. c. , fisher r. t. , townsley d. m. , calder a. c. , graziani c. , asida s. , lamb d. q. , truran j. w. , 2008 , , 681 , 1448 d. , rpke f. k. , woosley s. e. , 2009 , , 460 , 869 a. m. , 1991 , , 245 , 114 r. i. , castor j. i. , 1978 , , 220 , 902 r. et al . , 2005 , , 436 , 1021 b. , kirshner r. p. , filippenko a. v. , shields j. c. , foltz c. b. , phillips m. m. , sonneborn g. , 1991 , , 371 , l23 l. b. , 1971 , , 163 , 95 , 1999 , , 345 , 211 l. b. , solomon p. m. , 1970 , , 159 , 879 k. et al . , 2010 , , 466 , 82 k. , rpke f. k. , fink m. , hillebrandt w. , travaglio c. , thielemann f. , 2010 , , 712 , 624 k. , taubenberger s. , sollerman j. , mazzali p. a. , leloudas g. , nomoto k. , motohara k. , 2010 , , 708 , 1703 g. h. , hflich p. , gerardy c. l. , vacca w. d. , wheeler j. c. , robinson e. l. , 2009 , , 138 , 727 g. h. , hflich p. , wheeler j. c. , robinson e. l. , gerardy c. l. , vacca w. d. , 2006 , , 645 , 1392 s. , lundqvist p. , sollerman j. , kozma c. , baron e. , fransson c. , leibundgut b. , nomoto k. , 2005 , , 443 , 649 p. a. , 2000 , , 363 , 705 p. a. et al . , 2005 , , 623 , l37 p. a. , benetti s. , stehle m. , branch d. , deng j. , maeda k. , nomoto k. , hamuy m. , 2005 , , 357 , 200 p. a. , chugai n. , turatto m. , lucy l. b. , danziger i. j. , cappellaro e. , della valle m. , benetti s. , 1997 , , 284 , 151 p. a. , iwamoto k. , nomoto k. , 2000 , , 545 , 407 p. a. , lucy l. b. , 1993 , , 279 , 447 p. a. , lucy l. b. , butler k. , 1992 , , 258 , 399 p. a. , lucy l. b. , danziger i. j. , gouiffes c. , cappellaro e. , turatto m. , 1993 , , 269 , 423 p. a. , nomoto k. , cappellaro e. , nakamura t. , umeda h. , iwamoto k. , 2001 , , 547 , 988 p. a. , nomoto k. , patat f. , maeda k. , 2001 , , 559 , 1047 p. a. , rpke f. k. , benetti s. , hillebrandt w. , 2007 , science , 315 , 825 p. a. , sauer d. n. , pastorello a. , benetti s. , hillebrandt w. , 2008 , , 386 , 1897 c. a. , seitenzahl i. , townsley d. , jordan g. c. , truran j. , lamb d. , 2009 , , 693 , 1188 k. et al . , 2006 , , 652 , l101 k. , sugimoto d. , neo s. , 1976 , , 39 , l37 k. , thielemann f .- k . , yokoi k. , 1984 , , 286 , 644 k. , yamaoka h. , shigeyama t. , kumagai s. , tsujimoto t. , 1994 , in supernovae , bludman s. a. , mochkovitch r. , zinn - justin j. , eds . , p. 199
p. , phillips m. , baron e. , branch d. , hauschildt p. , 1995
, , 455 , l147
a. et al . , 2007 , , 377 , 1531 a. w. a. , duschinger m. , mazzali p. a. , puls j. , lennon m. , miller d. l. , 1996 , , 312 , 525 s. et al .
, 1999 , , 517 , 565 m. m. , 1993 , , 413 , l105
g. et al . , 2008 , , 388 , 971 p. a. , eastman r. g. , 2000 , , 530 , 757 t. , 2007 , , 657 , 942 t. , calder a. c. , lamb d. q. , 2004 , , 612 , l37 r. , hflich p. , kannappan s. j. , rykoff e. , rujopakarn w. , akerlof c. w. , gerardy c. l. , wheeler j. c. , 2006 , , 636 , 400 m. , hillebrandt w. , niemeyer j. c. , 2002 , , 391 , 1167 a. g. et al . , 1998 , , 116 , 1009 a. g. , press w. h. , kirshner r. p. , 1996
, , 473 , 88 f. k. , hillebrandt w. , 2005 , , 431 , 635 f. k. , niemeyer j. c. , 2007 , , 464 , 683 p. , lucy l. b. , 1992 , , 400 , 127 v. et al . , 2007 , , 469 , 645 m. , mazzali p. a. , benetti s. , hillebrandt w. , 2005 , , 360 , 1231 m. et al .
, 2008 , , 677 , 448 m. , mazzali p. a. , maeda k. , nomoto k. , 2006 , , 645 , 470 r. c. et al . , 2007 , , 654 , l53 r. c. , branch d. , baron e. , nomoto k. , li w. , filippenko a. v. , 2004 , , 601 , 1019 c. , hillebrandt w. , reinecke m. , thielemann f .- k .
, 2004 , , 425 , 1029 x. et al . , 2009 , , 697 , 380 x. et al . , 2008 , , 675 , 626 s. e. , kasen d. , blinnikov s. , sorokina e. , 2007 , , 662 , 487 m. et al . , 2009 , , 61 , 713
the spectrum synthesis code for the photospheric phases assumes a sharply defined spherical photosphere as an inner boundary . at the inner boundary ,
blackbody radiation is assumed .
since the temperature structure [ @xmath139 is unknown at first , the calculation starts with an assumed temperature . using this temperature , the ionization and excitation in the optically thin atmosphere above the photosphere
are calculated considering the dilution of the radiation .
the degree of dilution @xmath140 is defined by the equation @xmath141 in our code .
the excitation ( the population of an excited level @xmath142 ; @xmath143 for the ground state ) is computed as in @xcite and @xcite : @xmath144 where @xmath145 and @xmath146 are the statistical weight and the excitation energy from the ground level , respectively . for the ionization , the nebular approximation , which is similar to that made in the studies of stellar winds @xcite or planetary nebulae @xcite ,
are adopted with the modifications for sne @xcite : @xmath147 w \left ( \frac{t_e}{t_r } \right)^{1/2 } \left ( \frac{n_{i+1}n_e}{n_i } \right)^{*}_{t_r } , \label{eq : ion}\ ] ] where @xmath148 and @xmath149 are the electron density and the electron temperature , respectively . in the equation ,
@xmath150 and @xmath151 are the correction factors for the optically thick continuum shorter than the ca ii ionization edge and the fraction of recombinations that go directly to the ground state , respectively @xcite .
the first and second terms in the first square brackets on the right hand side represents the excitation from the ground and excited levels , respectively .
the last term on the right hand side is the ionization computed from the saha equation with @xmath152 . for the electron temperature , @xmath153 is crudely assumed @xcite . with the ionization and excitation states , line optical depths
are calculated under the sobolev approximation , which is a sound approximation in rapidly expanding envelope with large velocity gradients @xcite .
a number of photon packets are traced above the photosphere in spherical coordinates , taking into account electron scattering and line scattering . for line scattering , the effect of the line branching
is included @xcite .
the stream of the photon packets gives the flux at each radial mesh and a frequency moment is calculated by @xmath154 the radiation temperature is estimated from this frequency moment via @xmath155 here @xmath156 represents for the mean energy of the blackbody radiation [ @xmath157 . using this new temperature structure , the ionization and excitation are updated .
then , the next monte carlo ray tracing starts .
this procedure is continued until the temperature converges .
finally , the emergent spectrum is calculated using a formal integral @xcite .
it is known that these assumptions give a good approximation to the results of detailed nlte calculations @xcite . however , since no energy deposition by @xmath2ni and @xmath2co is assumed above the photosphere , the assumptions become unreliable as a large part of the @xmath2ni - rich layer becomes optically thin .
in addition , the assumption of a sharp photosphere becomes progressively invalid after maximum brightness .
thus , the quantities derived by fitting the spectra at later epochs ( especially after maximum brightness in sne ia ) are more uncertain .
the spectrum synthesis code for the nebular phases first calculates the energy deposition by @xmath88-rays and positrons produced in the decay of @xmath2co . for a given distribution of @xmath2ni , the energy deposition in each shell
is evaluated in a monte carlo scheme @xcite .
the deposited energy is spent in the ionization and in the heating of thermal electrons .
the latter is known to be dominant in fe - rich ejecta ( @xmath158 per cent , @xcite ) . for given ionization fractions in the ejecta ,
the heating of the thermal electrons ( @xmath159 ) is balanced with the radiative cooling ( thermal balance ) : @xmath160 where @xmath161 , @xmath162 and @xmath163 are the cooling via lines , free - free emission , and recombination , respectively .
since the contribution from the latter two are smaller than the line cooling ( @xmath164 and @xmath165 per cent , respectively , @xcite ) , we only consider cooling by lines .
the line cooling per unit volume by lines is given by @xmath166 where @xmath167 is the number density of the ion @xmath168 in the energy state @xmath169 , and @xmath170 and @xmath171 represents for the einstein a - coefficient and the escape probability of the transition @xmath172 . from this thermal balance ,
an electron temperature is evaluated in each shell .
the ionization fractions are determined balancing impact ionization by non - thermal electrons and recombinations ( ionization balance ) .
collisional ionization by thermal electrons can be neglected @xcite .
the ionization balance is solved for a given electron temperature . using this new ionization ,
the thermal balance is reconsidered , i.e. , the temperature and the ionizations are solved iteratively .
after the convergence , the emergent spectrum is computed by integrating the contribution from each shell considering the doppler shift .
the si ii line profile in sn 2003du is rather narrow and it is similar to that found in , e.g. , sne 2002er ( fig . [
fig : si ] , hvg , * ? ? ?
* ) , 2003cg ( lvg , * ? ? ?
* ) , 2005cg ( lvg , * ? ? ?
this is in contrast to the broad absorption seen in sne 2002bo ( fig .
[ fig : si ] , hvg , * ? ? ?
* ) , 2002dj ( hvg , * ? ? ?
* ) , 2006x ( hvg , * ? ? ?
* ; * ? ? ?
* ) , or a possible boxy profile in sne 1990n ( lvg , * ? ? ?
* ) , 2001el ( lvg , * ? ? ?
* ) , 2004dt ( hvg , * ? ? ?
* ) , 2005cf ( fig .
[ fig : si ] , lvg , * ? ? ? * ; * ? ? ?
lvg and hvg mean low velocity gradient and high velocity gradient , respectively .
these are the classifications of type ia sne based on the temporal evolution of the si ii line velocity @xcite .
note that high velocity features of ca ii lines are commonly seen in both groups at pre - maximum epochs @xcite .
we use the spectrum at @xmath56 days from @xmath15 maximum ( @xmath173 ) to constrain the distribution of si in the outer layers because this spectrum is reproduced better than the one at @xmath48 days ( fig .
[ fig : spec1 ] ) . to reproduce the observed profile ,
a moderately mixed - out si distribution is required .
the mass fraction of si is @xmath17(si)@xmath64 at @xmath174 and it is suppressed at the outer layers , @xmath17(si)@xmath175 at @xmath98 ( see figs .
[ fig : abunmass ] and [ fig : abunvel ] ) .
this model is shown in figures [ fig : spec1 ] and [ fig : si ] by the red line . in the right panel of figure
[ fig : si ] , two other models are shown . the orange line ( `` cutoff '' )
shows a model spectrum where the si abundance has been set to @xmath17(si)=0 at @xmath98 as in the w7 model ( fig .
[ fig : abunvel ] ) . in this model ,
the absorption of si at @xmath98 is clearly too weak , suggesting that some degree of burning took place in the outer layers @xcite .
the purple line ( `` homogeneous '' ) shows the model spectrum with a homogeneous si abundance with @xmath17(si)=0.3 at @xmath110 . in this case , the absorption at a high doppler shift is too strong .
this gives a similar profile to that in sne 2002bo and 2005cf .
in fact , such an extended si distribution was derived for sn 2002bo in paper i. 6355 shown in the doppler velocity measured from the rest wavelength of the si ii line ( @xmath176 ) .
_ left : _ the si line profile in sne 2003du ( at @xmath56 days from @xmath15 maximum , @xcite ) , 2002er ( @xmath177 days , @xcite ) , 2002bo ( @xmath178 days , @xcite ) , and 2005cf ( @xmath179 days , @xcite ) at similar epochs .
these spectra are normalized to sn 2003du at @xmath176 . _
right : _ the si line profile in sn 2003du at @xmath56 days from @xmath15 maximum compared with three synthetic spectra .
three model spectra are normalized to the best fit model at @xmath176 .
the red line shows the best fit model using the abundance distribution shown in figs .
[ fig : abunmass ] and [ fig : abunvel ] .
the orange line ( `` cutoff '' ) shows the model spectrum where the mass fraction of si , @xmath17(si ) , is set to be zero above 15000 .
the purple line ( `` homogeneous '' ) shows the model spectrum where the homogeneous si abundance with @xmath17(si)=0.3 is assumed at @xmath110 . ]
lines of carbon are not clearly seen in the optical spectra of sn 2003du . a marginal detection may be seen in correspondence of the emission peak of si ii @xmath466355 @xcite . the profile around the emission peak
is enlarged in figure [ fig : c ] and plotted versus doppler velocity measured from the c ii @xmath466578 line .
we use the spectrum at @xmath56 days from @xmath15 maximum ( @xmath173 ) to estimate an upper limit for unburned c since the emission peak of the si line is reproduced more nicely than in the spectrum at @xmath48 days .
we first divide the ejecta at @xmath110 into three parts , with boundaries at @xmath180 and @xmath181 . at @xmath182 ,
no constraints can be obtained on the c abundance because of the low density ( weakness of the line ) and the blending with the strong si ii line .
thus , we could set the upper limit of c as @xmath17(c)=0.5 at @xmath182 . in practice , since there are other elements such as si and ca in this velocity space , the mass fraction is set to be @xmath17(c)=0.4 at @xmath182 in the best fit model .
since the ejecta mass at @xmath182 is @xmath183 , the upper limit of the c mass at @xmath182 is @xmath184 .
the model spectrum with @xmath17(c)@xmath185 at @xmath186 shows a noticeable c ii line at the emission peak of the si ii line ( left panel of fig .
[ fig : c ] ) .
since the ejecta mass at @xmath186 is @xmath187 , the upper limit of the c mass in this layer is @xmath188 , which is smaller than the c mass allowed in the outermost layers ( @xmath184 at @xmath182 ) . at @xmath174 ,
the constraint on the c abundance is even stronger . the model spectrum with @xmath17(c )
@xmath189 gives a clear absorption trough of the c iiline ( right panel of fig .
[ fig : c ] ) .
thus , although the ejecta mass at @xmath174 is large ( @xmath190 ) , the c mass at these layers is less than @xmath191 .
thus , most of the c that is allowed must be located at @xmath182 .
summing up the upper limits to the c mass in different layers , we can safely conclude that the mass of c ejected in sn 2003du is @xmath192 at @xmath193 .
this is less than in the deflagration models ( @xmath194 ) , where @xmath17(c)=0.5 at @xmath35 ( see figs .
[ fig : abunmass ] and [ fig : abunvel ] , and table [ tab : mass ] ) .
note that all estimates are performed assuming the density structure of w7 model .
if a delayed detonation model was used , @xmath17(c ) in the outermost layers would be smaller @xcite .
6355 line shown in the doppler velocity measured from the c ii @xmath466578 .
_ left : _ the profile in sne 2003du at @xmath56 days from @xmath15 maximum compared with the model spectra with @xmath17(c)=0.01 ( red , best fit ) , 0.05 ( orange ) , and 0.4 ( purple ) at @xmath186 .
_ right : _ the profile in sn 2003du at @xmath56 days from @xmath15 maximum compared with the model spectra with @xmath17(c)=0.002 ( red , best fit ) and 0.008 ( orange ) at @xmath174 . | the element abundance distributions in the ejecta of type ia supernova ( sn ) is studied by modelling a time series of optical spectra of sn 2003du until @xmath0 year after the explosion . since sn 2003du is a very normal type ia sn both photometrically and spectroscopically , the abundance distribution derived for it can be considered as representative of normal type ia sne .
we find that the innermost layers are dominated by stable fe - group elements , with a total mass of @xmath1 , which are synthesized through electron capture . above the core of stable elements
there are thick @xmath2ni - rich layers .
the total mass of @xmath2ni is @xmath3 .
the si- and s - rich layers are located above the @xmath2ni - rich layers .
the dominant element in the outermost layers ( @xmath4 , @xmath5 ) is o , with a small amount of si .
little unburned c remains , with an upper limit of 0.016 @xmath6 .
the element distributions in the ejecta are moderately mixed , but not fully mixed as seen in three - dimensional deflagration models .
[ firstpage ] supernovae : general supernovae : individual : sn 2003du nucleosynthesis , abundances |
hybrid systems composed of plasmonic nanostructures and quantum emitters / absorbers ( e.g. , semiconductor quantum dots or other gain media ) are currently of much interest .
this is due to the plasmonic element s ability to interact strongly with light , leading to subwavelength control and confinement , and the possibility of the quantum element introducing nonlinear optical ( as well as more general quantum ) responses that may be enhanced by its proximity to the plasmonic element .
for example , theoretical predictions of certain fano resonance phenomena @xcite in such systems exist , and experiments are approaching the ability to measure such features @xcite and have already demonstrated novel lasing action @xcite and quantum coherences . @xcite the studies of such hybrid systems can also be viewed as steps in the emerging field of `` quantum plasmonics , '' @xcite which aims to realize quantum - controlled devices relevant to quantum sensing , single - photon sources , and nanoscale electronics
. features relevant to quantum information , such as the entanglement among the quantum dots ( qds ) , can also be achieved in hybrid plasmonic / qd systems .
this ability may seem surprising given the dissipative ( or lossy ) aspect of plasmonic structures .
however , interactions between quantum objects and a dissipative environment lead to the production of stable entangled states .
@xcite several pioneering theoretical studies have shown that dissipation - induced entanglement is relevant to systems of qds interacting with plasmonic nanostructures .
@xcite we previously explored methods for generating entanglement in qd - plasmon systems , using both systems in which only one qd is initially prepared in its excited state and the system evolves without external excitation and systems in which all the qds are initially in their ground states and the entire system is excited by an ultrafast laser pulse .
@xcite we showed that either a single or repeated optical pulse entangled the qds and that the amount of entanglement can be tuned by controlling the coupling of the qds to the plasmonic nanostructure .
furthermore , the whole system can be excited with a single pulse , without the need to individually address each subsystem .
this work and the present work allow for asymmetric coupling of the qds to the plasmonic system ; for example , one can imagine the qds to be configured to be at different distances from the plasmonic system or in some other way that can lead to asymmetry . to allow dissipation - induced entanglement to be an effective candidate for quantum information applications
, one must thoroughly understand how the entanglement is generated .
furthermore , constraining parameter sets in experimentally viable regions of the parameter space and knowing the sets associated degrees of entanglement are important for engineering such systems within any larger quantum information platform . in this paper
, we seek to determine system features that maximize the degree of entanglement between the qds . to accomplish this objective ,
we employ analysis based on solutions of limiting forms of the problem and optimization based on numerical solutions to the complete cavity quantum electrodynamics equations .
we show that for any number of qds , simple initial conditions and parameter guidelines generate systems where all pairs of qds share some degree of entanglement .
we consider a cavity quantum electrodynamics ( cqed ) model of a system of @xmath0 quantum dots in proximity to a plasmonic system . the underlying system s basis states are @xmath1 where @xmath2 represents the exciton of the @xmath3th qd and @xmath4 represents the plasmon energy levels . using a simplified notation @xmath5 , we can write the density operator as @xmath6 then our governing equation describing the cqed model is defined as @xmath7 - \frac{i}{\hbar } [ \hat{h_d } , \hat{\rho } ] + l(\hat{\rho}),\ ] ] where @xmath8 , @xmath9 , and @xmath10 are the operators for the hamiltonian , the driving term , and the lindblad , respectively . the hamiltonian @xmath8 for the coupled dot - plasmon system as @xmath11 defining the lowering and raising operator pairs for both the qds and the plasmon , @xmath12 and @xmath13 , in the usual manner as in @xcite , we have the isolated dot and plasmon hamiltonian terms @xmath14 respectively , and the dot - plasmon coupling terms @xmath15 equation represents the simplest possible dot - plasmon coupling term corresponding to a qd gaining ( losing ) a quantum of energy when the plasmon loses ( gains ) a quantum of energy . for the system exposed to a time - dependent electric field
@xmath16 , we have @xmath17,\ ] ] where @xmath18 and @xmath19 denote the transition dipole moments of the qds and plasmon , respectively .
we assume that the distance between the qds is large compared with the separation between qds and neighboring metal nanoparticles , so that direct through - space coupling among the qds can be neglected .
we also neglect retardation ; hence , our treatment is limited to systems with physical dimensions that are small compared with optical wavelengths .
the lindblad superoperator @xmath20 in describes the dephasing and dissipation effects .
we employ a previously developed @xcite extension of @xmath21 that is parameterized by the qd population decay @xmath22 , the qd dephasing rate @xmath23 , and the plasmon decay constant @xmath24 .
we consider time scales on the order of the inverse of these rates , so that there are no correlated fluctuations in the qds states and so that the use of the lindblad superoperator is justified .
although environmental dephasing is explicitly included for the qds , it is not necessary to do so for the plasmon because the dephasing that arises from its decay ( encoded in the corresponding term in @xmath21 ) is much larger in magnitude . as in ref . , the rotating wave approximation is applied .
we use wootter s concurrence @xcite to measure the entanglement of the system .
an alternative representation of the density operator is the density matrix @xmath25 with its elements defined by @xmath26 let @xmath27 be the reduced density matrix associated with one particular pair of qds , @xmath28 and @xmath29 , obtained by tracing the full density matrix @xmath25 over the plasmon quantum numbers @xmath30 and the quantum numbers for all other qds .
the @xmath31 pairwise concurrence is then given by @xmath32 where @xmath33 are the square roots of the eigenvalues of @xmath34 with @xmath35 ( in descending order ) .
the matrix @xmath36 corresponds to the spin - flipped density matrix @xcite @xmath37 where @xmath38 we define the `` dark '' evolution to be how a qd - plasmon system evolves given some initial qd excitation with everything else in the system initially in the ground state . in the limit of low total excitation energy
one can develop an exact analytical solution for the problem of an arbitrary number of quantum dots interacting with the plasmon if qd dephasing is neglected .
this procedure is discussed in appendix [ three - state - model ] .
we first discuss some predictions from this analysis for two qds and then briefly for larger numbers of qds . also of interest
is the case of pulsed excitation , where the system is initially cold and then subjected to a laser pulse .
we follow the discussion of the dark evolution with analysis of this pulsed case using simple rabi flopping ideas . in the case of two qds coupled to a plasmon , we are concerned with determining qd - plasmon coupling factors ( @xmath39 and @xmath40 ) that maximize entanglement . for two qds in particular , it is convenient to consider the two entangled qd states @xmath41 \ket{s}\ ] ] and @xmath42 \ket{s}\ ] ] in our calculations instead of the direct product of primitive qd states as in . for two qds ,
appendix [ three - state - model ] discusses in detail a three - state hamiltonian model involving the basis states @xmath43 , @xmath44 , and @xmath45 that neglects qd dephasing and spontaneous emission but allows for plasmon dissipation by introducing an appropriate complex diagonal matrix element to the hamiltonian matrix .
the initial state of relevance to the dark limit calculations is one with an excited qd1 , an unexcited qd2 , and a plasmon ; this state is represented by @xmath46 .
this initial state is interesting because although it is a separable , unentangled state , it is a nonstationary state of the full system that has been shown to evolve into a state with a possibly significant transient degree of entanglement .
@xcite with no plasmon dissipation ( @xmath47 ) and for short times @xmath48 , the probabilities of states @xmath49 , @xmath50 , and @xmath51 are given by the respective squares of @xmath52 , @xmath53 , and @xmath54 , where @xmath55 when @xmath56 , the limit as @xmath57 is of interest because the system can then reach a steady state in the populations .
appendix [ three - state - model ] shows that for the initial condition with one excited qd and the rest of the system unexcited , @xmath58 and @xmath59 where @xmath60 .
remarkably , these results are valid for any positive value of @xmath24 ( although it must be remembered that no qd dephasing has been allowed ) .
the concurrence in this asymptotic limit is simply @xmath61 and can be readily maximized to yield the optimum ratio of coupling strengths : @xmath62 or @xmath63 appendix [ three - state - model ] also develops an _ exact _ procedure for constructing the corresponding dark dynamics of @xmath0 qds interacting with a plasmon , without qd dephasing .
this system is then described by an effective @xmath64 complex effective hamiltonian model . for the scenario of one qd
initially excited , it can be used to get an idea of how the entanglement results scale with increasing @xmath0 .
we are also interested in the case when the system is initially unexcited and an optical pulse is used to generate transient entanglement . assuming
the pulse is relatively simple and resonant with the qds transition frequencies , a simple question to ask is what values of @xmath39 and @xmath40 will lead to the two - qd system being close to the @xmath65 state .
we know from previous work @xcite that such a system will evolve into a state with some degree of entanglement . on resonance ,
the qds undergo rabi oscillations as they are excited by the laser pulse .
the time for qd@xmath3 to undergo one rabi oscillation ( i.e. , to go from the ground state to the excited state and then back to the ground state ) is @xmath66 , where @xmath67 with @xmath68 being the amplitude of the sinusoidal electric field experienced locally by qd@xmath3 . the final term in is obtained by using the expression for @xmath69 derived in appendix [ local - field - enhancement ] , which relates this local electric field to the incident field @xmath70 .
( other phenomena , such as the purcell effect , are also occurring ; the rabi formulae discussed here should be construed as approximate indicators of the dynamics . ) in order to achieve a highly entangled state , one qd , say qd1 , must undergo @xmath71 rabi oscillations ( with @xmath72 ) so that it is left in the excited state .
the time for this process to occur is @xmath73 .
the other dot , qd2 , must undergo @xmath74 full oscillations so that it is left in its ground state .
the time for this process to occur is @xmath75 .
equating these two rabi times leads to the simple result that @xmath76 we see that the condition on the couplings for achieving one qd excited via pulsed excitation is not the same as the condition on the couplings for that excited state to evolve to an entangled state . nonetheless , for @xmath77 , predicts @xmath78 , in approximate accord with where @xmath79 .
although this restricts the parameter space for the pulsed case somewhat , many parameters ( especially those describing the pulse ) can still be varied freely . to find the set of system parameters that maximize the sum of the pairwise concurrences
, we employ a numerical optimization framework that samples the parameter space in a uniformly random fashion , evaluating the concurrence at each point .
the parameters in question include the @xmath0 qd - plasmon coupling coefficients ( @xmath80 , @xmath81 ) , environmental aspects such as the qd dephasing and plasmon decay constants ( @xmath82 and @xmath24 , respectively ) , and applied laser pulse features such as its fluence ( @xmath83 ) and duration ( @xmath84 ) .
( see sec .
[ sec : sim ] for definitions of the laser pulse parameters . )
since the sum of the pairwise concurrences is a nonconvex function of these parameters , several isolated local maxima are likely to exist .
our approach follows that in ref . , clustering evaluated points in the parameter space into basins of attraction for different maxima .
clusters are formed by using the points function values ( sum of pairwise concurrences ) and their proximity to points with better function values .
points that do not have a better point within a distance @xmath85 are considered the best points in their cluster .
the distance @xmath85 can be adjusted so a reasonable number of clusters are identified .
( one also can dynamically adjust @xmath85 as the parameter space is explored .
@xcite ) local optimization runs are then started from the best point in each cluster .
the local optimization problem of maximizing the sum of pairwise concurrences is solved by minimizing the figure of merit @xmath86 where @xmath87 depends on the system parameters being optimized over ( see ) .
this form is appropriate because the pairwise concurrences in are bounded above by 1 .
depending on the context , @xmath87 might be the maximum concurrence achieved over time or a long - time asymptotic limit . when viewed as a function of the parameters , defines a nonlinear least - squares problem .
we solve this problem with the practical optimization using no derivatives for sums of squares ( pounders ) algorithm .
@xcite for a system with @xmath0 qds , poundersiteratively builds local quadratic surrogates of each of the @xmath88 residual functions @xmath89 and combines this information in a master surrogate model . in each iteration of the algorithm ,
this surrogate model is minimized within a trust - region framework to generate candidate solutions .
[ numerical ] we consider the time evolution of the density operator in , with the choices of the parameters corresponding to a gold nanoparticle system interacting with qds in a polymer matrix with a dielectric constant @xmath90 = 2.25 ; these choices are similar to those originally used in our single plasmonic - qd system study .
@xcite for qd@xmath3 , we choose @xmath91 = 2.05 ev , assuming the qd and plasmon transition energies are equal .
we set the qd dipole moments to be @xmath92 = 13 d and the plasmon dipole moment to be @xmath93 = 4000 d. the qds are assumed to have the same spontaneous decay rate , @xmath94 = 190 nev . in some of our calculations we vary or consider several values for the qd dephasing rate , @xmath23 , and plasmon decay rate , @xmath24 .
consistent with our earlier work , base values are @xmath95 = 2 mev and @xmath96 = 100 mev .
we utilize coupling factors , @xmath97 , in the 030 mev range ; and unlike all the other qd parameters , we do allow qds to have different coupling constant values .
previous calculations show that a realistic approximation for the plasmon - qd coupling is approximately 10 mev for a system such as the one we study here .
@xcite other systems , such as silver nanoparticles or particles with different geometries , could exhibit larger coupling factors than does gold .
@xcite for calculations that include a laser pulse @xmath16 , we assume it has the form ( in the nonrotating frame ) @xmath98 , where @xmath99 and @xmath100 is a gaussian envelope function such that the full width at half maximum of @xmath101 is @xmath84 .
the pulse fluence is @xmath102 .
we formulate a density matrix equation from using , and we solve the density matrix equation consisting of a set of @xmath103 ordinary differential equations for the time - dependent complex amplitudes @xmath104 , with @xmath105 where @xmath0 is the number of qds and @xmath106 is the number of plasmon energy levels .
we solve these odes numerically using an efficient parallel solver that employs sparse matrix - matrix multiplication algorithms with either a runge - kutta or exponential time integration scheme .
we now detail our quantum dynamics results corresponding to a system of qds interacting with a plasmonic system as modeled in sec . [
sec : theoretical_methods ] .
we analyze such systems for both free evolution of some particular excitation ( what we refer to as `` dark '' evolution ) and in the presence of a laser pulse .
we first consider two qds ( qd1 and qd2 ) interacting with a plasmonic system under the assumption that the initial state @xmath107 has been prepared and evolves in the absence of any external pulses , that is , `` in the dark . '' unlike cases studied in previous work , @xcite the possibility of asymmetric dot - plasmon couplings ( @xmath108 ) can lead to new features in the time - dependent concurrence . when the qds are symmetrically distributed within the plasmonic system so that @xmath109 , the @xmath110 state is an eigenstate of the hamiltonian and decays with a relatively slow dephasing rate ( @xmath82 ) due only to the lindblad term in . with finite ( but still symmetric ) coupling , the @xmath111 state mixes with the @xmath112 state @xcite and is no longer an eigenstate of the hamiltonian .
the probability of being in @xmath113 is @xmath114 , leading to an increased initial decay of the @xmath113 state .
the plasmon decay ( @xmath24 ) damps out any additional oscillations of the @xmath113 population .
as shown previously , @xcite starting a system in the @xmath115 state then leads to a high degree of concurrence , since the @xmath111 state quickly decays , while the @xmath116 state undergoes a much slower decay . with @xmath117 ,
the maximum concurrence for such a symmetric system is therefore 0.5 .
we describe here a method to achieve larger degrees of concurrence by forcing the @xmath111 state to evolve into @xmath116 rather than into @xmath118 . by breaking the symmetry of the couplings between the two qds
, we mix the @xmath113 state with the @xmath119 state , through the @xmath120 state .
this approach follows the analysis in appendix [ three - state - model ] where a three - state model is discussed and solved analytically in certain limits .
when @xmath121 , no coupling occurs between @xmath113 and @xmath122 ; but when @xmath123 , the two states are indirectly coupled through the @xmath120 state ( to which both states are directly coupled ) .
starting in the @xmath124 state , and setting @xmath125 = 0 mev , lead to a cyclic evolution between a completely unentangled state and a highly entangled state .
0.5 0.5 we follow the convention that qd1 is the qd that is initially in its excited state .
if @xmath126 , then the @xmath122 state reaches a population approaching 1 ( fig .
[ cyclic_evolution(a ) ] ) ; if @xmath127 , then the @xmath113 state reaches a population approaching 1 ( fig .
[ cyclic_evolution(b ) ] ) .
the dynamics of these two examples are similar ; the @xmath122 ( resp .
@xmath111 ) state evolves into the @xmath118 state , which evolves into the @xmath111 ( resp .
@xmath116 ) state and back through the @xmath118 state into its initial state . where the @xmath111 ( resp .
@xmath116 ) state reaches its maximum , the concurrence does as well , reaching a value of nearly 1 . using the explicit three - state system described in in appendix [ three - state - model ] ,
we find the ratio @xmath128 gives an @xmath51 state population of unity , which also maximizes the concurrence for the @xmath126 case . a similar analysis can be done for the @xmath50 state , giving @xmath129 .
note that these optimal ratios for achieving large , instantaneous concurrences when there is no dissipation are different from those of sec .
[ analytical ] .
the latter concern either an asymptotic concurrence that can be reached in the case of dissipation or the couplings conducive to a pulsed laser generating a particular excited state that can then evolve to a state with significant concurrence . for short times @xmath48 , the approximation applies .
when @xmath126 in this case , the second term of @xmath130 is positive , which leads to a boost in the population of the @xmath131 state .
when @xmath127 , the second term is negative , and the population of the @xmath131 state initially declines . in both cases , @xmath132 initially declines ; but when @xmath127 , it reaches 0 much faster and then rises to nearly 1 .
both effects can be seen in fig .
[ cyclic_evolution ] . 0.5 0.5 + 0.5 0.5
when the lindblad terms describing dissipation and dephasing are added , the results of the two simulations ( @xmath133 ) = ( 12.5 mev , 25 mev ) or ( 25 mev , 12.5 mev ) in fig . [ cyclic_evolution ] become very different . in fig .
[ finite_gammas ] we consider simply adding plasmon dissipation ( @xmath56 ) to the simulations , while keeping the qd dephasing term @xmath82 at zero .
this case also has a closed - form solution ( see appendix [ three - state - model ] and sec .
[ analytical ] ) .
we see from figs .
[ finite_gammas ] and [ finite_gammas ] that the initial state evolves and begins to populate the first excited plasmon state , but the plasmon population quickly decays and the system reaches a steady state .
the steady - state concurrence for the case with @xmath126 , fig .
[ finite_gammas ] , is larger than the case with @xmath127 , fig .
[ finite_gammas ] .
this trend might be expected on the basis of the dynamics without plasmon decay , figs .
[ cyclic_evolution ] and [ cyclic_evolution ] , wherein a smooth rise of concurrence from 0 to 1 occurs over initial times for the case @xmath126 , fig .
[ cyclic_evolution ] , but a more complicated behavior involving a small local maximum in concurrence occurs for the case @xmath127 , [ cyclic_evolution ] . in a realistic system ,
the @xmath127 case will not create concurrences as large as those seen when @xmath126 ; the best case is that the plasmon decay is sufficiently large to stop the @xmath51 state from evolving into the @xmath50 state .
figure [ finite_gammas ] shows the maximum concurrence ( for each time trajectory ) for many different values of @xmath39 and @xmath40 . there is a clear area of large concurrence when @xmath134 in accordance with the expectation from in sec .
[ analytical ] .
small discrepancies with respect to can exist because this equation pertains to the asymptotic concurrence and we are considering the maximum concurrence achieved over a finite window of time .
note that the isolated qd population decay rates @xmath135 ( discussed in sec .
[ numerical ] ) are sufficiently small and are generally overwhelmed by the purcell decays that result from finite @xmath24 and plasmon - dot coupling factors , @xmath80 .
thus , the inclusion of decay in the qd populations has no significant effect on the results presented here for the other parameter values considered .
the qd dephasing terms @xmath23 , however , can have a more significant effect .
figure [ finite_gammas ] ( similar to fig .
[ finite_gammas ] , but with @xmath136 = 2.0 mev ) shows the maximum concurrence for many values of @xmath39 and @xmath40 .
naturally , the maximum concurrence is not as large as the @xmath136 = 0 mev case .
furthermore , the clear peak around the line @xmath134 has been distorted , although the line still has some significance .
including qd dephasing effects causes the qd populations to decay before significant entanglement can occur , unless the qds are strongly coupled . at small values of @xmath80 ,
the optimal point is far from the @xmath137 line ; but as the couplings are increased , the optimal points again fall upon the @xmath137 line . as mentioned above
, this derivation pertains to the asymptotic values of the concurrence , but the dephasing does not allow the system to approach that value without larger values of @xmath80 .
we have used the analytical solution for @xmath0 qds interacting with a plasmonic system with no qd dephasing ( appendix [ three - state - model ] ) to explore how the dark entanglement dynamics scales with increasing @xmath0 beyond @xmath139 . as noted in sec .
[ 2-dots - dark ] , introducing dephasing can lead to smaller concurrences and shifts in the optimal @xmath140 ratios , but our results should indicate what to qualitatively expect as @xmath0 increases .
as in our @xmath141 dark calculations , the initial condition corresponds to qd1 being initially excited . for the @xmath142 case , fig .
[ 3qdcontourplot ] shows a contour map of the asymptotic figure of merit as a function of @xmath143 and @xmath144 .
the results in this case do not depend on either @xmath39 or plasmon decay rate @xmath24 , provided that the latter is positive . ( the transient dynamics do depend on both @xmath39 and @xmath24 and can also be of interest . )
we see that the optimal concurrences are reached at @xmath145 , which is somewhat smaller than the @xmath146 ratio found for the @xmath141 case .
the optimal values of the concurrence are @xmath147 and @xmath148 . at 0.639 ,
the `` direct '' concurrence between the excited qd1 and each of the other @xmath149 dots is slightly smaller than the result for the two - qd system .
although we have not derived an explicit formula , we can evaluate the exact asymptotic dynamics of the @xmath0 qd case using the procedure described in appendix [ three - state - model ] .
we find that for the initial condition in question , two distinct concurrence values always exist : a major one ( @xmath150 ) , associated with all the qd pairs that involve the initially excited qd , and a smaller one ( @xmath151 ) , associated with all the indirectly excited pairs .
evaluation of the results for @xmath0 up to @xmath152 shows that @xmath153 , for @xmath154 ; that is , the major concurrence tends to zero , although it does so slowly , with an increasing number of qds . in this limit
, the minor concurrence decays somewhat faster , with @xmath155 .
also , the optimal concurrence figure of merit is achieved with just one unique ratio for all the couplings , @xmath156 .
we find that @xmath157 for @xmath154 .
the optimal value of @xmath40 becomes less than @xmath39 when @xmath158 , in contrast to the two- and three - qd systems , where @xmath127 .
this can be explained by looking at the relative fraction of qd pairs , @xmath159 , which have @xmath160 .
when @xmath161 , the fraction of qd pairs that have @xmath160 is greater than @xmath162 .
as @xmath0 becomes larger , more and more qd pairs have @xmath163 . when @xmath127 , there is a boost in @xmath150 , possibly at the cost of @xmath151
when @xmath0 is large , the solution from minimizing the figure of merit no longer favors boosting @xmath150 ; instead it increases the ( more numerous ) @xmath151 .
the state created by this mechanism , where all qds share bipartite entanglement ( possibly weakly ) with all other qds , is similar to a generalized w - state @xcite . in the w - state , all pairs of qubits have the same value of concurrence , and that value is as large as possible .
thus , the w - state is the optimal state , given our figure of merit . according to an idea known as the monogamy of entanglement ,
@xcite there is an upper bound on the possible sum of bipartite entanglement . when @xmath164 , each qubit pair can no longer be fully entangled . as @xmath0 increases , the maximum bipartite concurrence for each pair in the w - state decreases as @xmath159
this represents a fundamental limit on the entanglement that we can achieve in our system .
the decay of the concurrence with increasing @xmath0 in our qd - plasmon system is similar to that of the w - state .
furthermore , if we measure the initially excited qd , we can project onto a state where qd pairs share a small degree of entanglement with each other .
taking a ratio of the asymptotic value of @xmath151 in the projected state to the w - state concurrence shows that each qd pair will have only @xmath165 of the concurrence of the w - state .
while this may be a small fraction , it is a constant fraction with increasing @xmath0 , allowing a ( low - fidelity ) approximate w - state to be easily created for any number of qds .
the addition of decay in the qd populations will further decrease the fidelity ; but , as shown in the two - qd case , the entanglement still persists , although at a smaller value . , and
@xmath117 , as a function of the ratios of the qd / plasmon coupling parameters . ]
preparing a system in the initial state @xmath115 can create high degrees of entanglement , but it does not represent a simple experimental setup . a simpler setup is to prepare a system and excite it with a single optical pulse . introducing a laser pulse to a system increases the number of parameters that the system depends on and can have a large effect on the value of the concurrence .
@xcite for the two - qd system , the parameters varied include the laser fluence ( @xmath83 ) , laser duration ( @xmath84 ) , coupling strengths ( @xmath39 and @xmath40 ) , qd dephasing ( @xmath82 ) , and plasmon dephasing ( @xmath24 ) ; @xmath166 , @xmath167 , @xmath92 , @xmath93 , and @xmath22 remain fixed .
we also constrained the parameter values in a physically reasonable part of the parameter space ; see table [ constraints ] .
.constraints for optimization parameters .
[ cols="<,>,>",options="header " , ]
we provide a detailed explanation of the origins and optimization of bipartite ( or pairwise ) entanglement in two , three , and an arbitrary number of qds coupled to a plasmonic system .
we analyze systems with an initially excited state as well as initially unexcited systems excited by a laser pulse .
we vary the qd - plasmon coupling values ( which represent a qd s distance from the plasmonic system ) , as well the femtosecond pulse parameters and dephasing rates to explore entanglement generation . by utilizing the full density matrix master equation , we are able to study the entanglement ( via concurrence ) of many different systems . in the case of two qds ,
two mechanisms are identified as the source of the entanglement generation : the differing decay rates of the @xmath168 and @xmath131 states ( previously identified ) @xcite and a new mechanism involving an indirect coupling between the @xmath168 and @xmath131 states that leads to a boost in the @xmath131 state population . with no dephasing or decay ,
high degrees of entanglement can be generated by having near - unity populations of either the @xmath131 or @xmath168 states .
when plasmon decay is added , however , the entanglement generated from @xmath131 is much higher than that of @xmath168 . a simple analysis including plasmon decay but neglecting qd dephasing predicts that the asymptotic concurrence is maximized when @xmath146 in the dark case ; calculations show that this relation is still useful when qd dephasing is considered .
the two entanglement - generating mechanisms are most apparent when the system is initially prepared with equal amounts of @xmath50 and @xmath51 , which is most easily achieved by having one excited qd and the other qd in the ground state .
this dark case may be contrasted with attempts to generate entanglement in initially unexcited systems by using laser pulses . on the basis of optimization
, we find that only certain sets of parameters generate an analog to the dark case .
in particular , certain rational values of the ratio of the qd - plasmon couplings , @xmath169 , lead to results comparable to the dark case . these ratios can be understood by analyzing the underlying rabi flops of the component qds ; the target final state , after the pulse , consists of one qd excited and the other qd in the ground state . to achieve this
, one qd undergoes @xmath170 rabi oscillations , leaving it in an approximate excited state , while the other qd undergoes @xmath171 full oscillations , leaving it in an approximate ground state .
this method will work for any pair of qds , even if that pair is part of a larger system of @xmath0 qds . in the case of three qds
, we optimize the sum of the bipartite concurrences among all of the pairs . several local maxima corresponding to different sets of system parameters
are obtained , and we present two in this paper .
one ( `` solution 1 '' ) was analogous to the two - qd systems discussed above , with @xmath172 , while the other ( `` solution 2 '' ) had @xmath173 ; both exhibited the entanglement generation mechanisms described above .
the parameters for all three optimal systems are listed in table [ tab : summary ] .
the ratios of the @xmath80 values in the highly entangled pairs of these three - qd systems follow the simple rules derived from the rabi - flop analysis .
we also extended our results to @xmath0 qds , with some simplifying assumptions , such as no qd population decay and a single initially excited qd . for any number of qds , all pairs of qds will become entangled . however , since this mechanism relies on one qd being in the excited state and the rest of the qds being in the ground state , these mechanisms can strongly entangle only a fraction ( @xmath159 ) of the pairs of qds . using the simple rules laid out in this paper for a large number of qds results in the excited qd being strongly entangled with all other qds , but
all the ground state qds will be strongly entangled only with the excited qd and only weakly entangled with each other . since all qds share some amount of bipartite entanglement with all other qds , the resulting state is similar to a generalized w - state and , with a measurement of the excited qd ,
can be projected to a state where all pairs of qds share the same amount of bipartite concurrence ( though this projected state only has @xmath165 of the bipartite concurrence of a true w - state ) .
additionally , this procedure could generate certain types of cluster states .
a cluster state is a graph in which qubits are represented by nodes on a graph , and an edge between two nodes represents entanglement between the two qubits .
@xcite the w - state would be a cluster state represented by a complete , fully connected graph .
an important subclass of cluster states is the star state , where a central node is connected to all other nodes ( or , a single qubit is entangled to all other qubits but the other qubits are not entangled with each other ) .
a four - qubit star cluster can be used for universal quantum computing .
@xcite in our model , we have an approximate star cluster for @xmath0 qds , since the initially excited qd is strongly entangled with all other qds .
although we showed @xmath0 qd results only for a specific initial starting condition , we also show how this state can be prepared for @xmath0 qds from a single , ultrafast laser pulse .
the rules for the ratios of the coupling strengths based on the number of rabi flops can be used to define an appropriate set of parameters to prepare an approximate form of the specific initial starting condition studied .
further studies of such systems that better approximate the w - state are planned , as are studies of the entanglement between all qubits of the system ( rather than just pairs ) , which would be similar to the ghz - state , @xcite which represents entanglement where all of the qubits are mutually entangled with each other ( rather than just sharing bipartite entanglement with other qubits ) .
the w- and ghz - states represent two mutually exclusive examples of multipartite entanglements and allow entanglement to be used as a quantum information resource in different ways .
@xcite we will also apply the same ideas and methods to other systems , such as nitrogen vacancies in diamond and superconducting qubits .
this work was performed at the center for nanoscale materials , a u.s .
department of energy , office of science , office of basic energy sciences user facility and supported by the u.s .
department of energy , office of science , office of advanced scientific computing research ( both under contract no .
de - ac02 - 06ch11357 ) .
we thank ron shepard for helpful suggestions concerning some of the analytical analysis presented in this paper .
we thank todd pittman and jason kestner for helpful discussions .
the `` dark '' problem , namely , to determine the dynamics of @xmath0 qds and a plasmonic system that results from a given initial condition without any applied laser pulse , can be solved analytically if the initial condition is not too energetic and qd dephasing is neglected .
an example of such a system would be if there is just one quantum of excitation within the qd manifold and a cold plasmonic system . the analytical solution is made possible because under such conditions a time - dependent schrdinger equation involving an effective , non - hermitian hamiltonian can be employed and the latter can be represented by an @xmath64 matrix with a simple structure .
first we illustrate such a solution in detail for the case of @xmath141 .
we then present the general @xmath175 state solution . for two qds interacting with a plasmonic system ,
we wish to solve for the time evolution of @xmath176 satisfying @xmath177 where @xmath178 we refer to the three states @xmath43 , @xmath44 , and @xmath179 as the zero - order basis .
this limited basis is adequate for describing an initial condition that involves any superposition of these three states , such as the case of one qd being excited and the plasmonic system and other qd being cold . with the definitions of the basis states in the text , , , and , and the hamiltonian operator ,
, the corresponding @xmath180 hamiltonian matrix of the zero - order basis representation is @xmath181 the qd and plasmon transition frequencies are assumed to be equal , @xmath182 , the coupling between @xmath49 and @xmath50 is @xmath183 and the coupling between @xmath49 and @xmath51 is @xmath184 we assume no direct coupling between @xmath50 and @xmath51 . notice that in , we have added an imaginary part @xmath185 to the diagonal matrix element associated with @xmath49 . with @xmath186
this term represents the dissipative loss of the plasmonic system . introducing the more slowly varying coefficient vector @xmath187 leads to @xmath188 where @xmath189 the solution of is thus @xmath190 in the limit ( assuming no plasmon dissipation , @xmath191 ) , expanding the exponential and re - grouping terms ,
can be written more explicitly as @xmath192 where @xmath193 and @xmath194 note that for the initial condition corresponding to @xmath195 or @xmath196 , @xmath197 , the above exact solution ( for @xmath191 ) is such that @xmath52 is a purely imaginary number for all times and that @xmath53 and @xmath54 are purely real numbers for all times .
equation , can be approximated to various orders in time by expanding @xmath198 and @xmath100 defined in appropriately .
thus , with the initial condition @xmath199 , the approximate solution , accurate to second order in time , is @xmath200 or , in terms of @xmath39 and @xmath40 , @xmath201 of course , another way to obtain is to determine the eigenvalues and eigenvectors of @xmath202 , @xmath203 and @xmath204 , @xmath205 , and represent with them .
this procedure can be carried out exactly even when plasmonic dissipation is allowed ( @xmath206 ) .
the eigenvalues of @xmath207 are easily found to be @xmath208 and the associated ( unnormalized ) eigenvectors projected onto the zero - order basis are @xmath209 @xmath210 and @xmath211 the propagator may then be written as @xmath212 where @xmath213 the bra vectors we employ , @xmath214 ( as is most common ) , are defined to be the transpose of the complex conjugates of the coefficients representing their corresponding kets , @xmath215 .
thus @xmath216 , where @xmath217 , @xmath218 .
an expression such as , which involves an additional complex conjugate in the argument of the bra vector , implies that @xmath219 is the sum of the ( complex ) squares of the components of @xmath204 , as opposed to being the more familiar sum of the squares of the magnitudes of the components .
this necessary peculiarity arises from @xmath220 being symmetric but not hermitian .
( in particular the symmetry of @xmath220 implies for eigenvalues @xmath221 that @xmath222 , which ultimately leads to an expression for the unity operator involving a sum of @xmath223 terms instead of the more familiar sum of @xmath224 terms . )
we note that for @xmath225 , @xmath226 and @xmath227 _ always _ have negative imaginary components . as @xmath57 ,
only the @xmath228 contribution to survives because only @xmath229 has no decay ( or negative imaginary ) component .
if we initiate the system with one qd excited , then @xmath230 the asymptotic amplitude for @xmath231 is then @xmath232 where @xmath233 and we have used the fact that @xmath234 . in a similar fashion we find @xmath235 the asymptotic concurrence in this case is simply @xcite @xmath236 since the magnitude of @xmath237 in is always less than 1 when @xmath39 and @xmath40 are positive , @xmath238 , and one can ignore the outer absolute signs .
the concurrence then reduces to @xmath239 viewed as a function of @xmath237 , the maximum of is found to be at @xmath62 , and corresponds to @xmath146 , consistent with the results in the text . for this value of @xmath237 , @xmath240 . the three - state model above involving the states @xmath241 ,
@xmath44 , and @xmath179 is convenient because it led directly to simple analytical expressions for the asymptotic concurrence .
however , the same result can be obtained , with a little more work , by employing the basis @xmath241 , @xmath242 and @xmath243 .
in fact this approach is advantageous because it then is easily generalizable to @xmath138 qds .
assume we have @xmath0 qds , with each qd@xmath244 interacting only with the dissipative plasmon via a hamiltonian coupling term @xmath245 .
if the basis is taken to be @xmath246 , @xmath247 , @xmath248 , , then one has an @xmath64 hamiltonian matrix representation @xmath249 with @xmath250.\ ] ] the characteristic equation for the eigenvalues of @xmath251 is then @xmath252 , where @xmath253 .
it implies that there are @xmath149 degenerate eigenvalues @xmath254 and two complex eigenvalues , @xmath255 let @xmath256 denote the eigenvector corresponding to the @xmath244th eigenvalue , and let @xmath257 denote the @xmath258th component of this eigenvector .
one can easily see that the @xmath259 degenerate eigenvectors must all have @xmath260 ; that is , they contain no component in the basis state @xmath261 .
the remaining components must satisfy @xmath262 although one can easily solve for low @xmath0 in various ways , a systematic procedure for obtaining @xmath149 linearly independent and orthogonal eigenvectors is as follows .
notice that implies that each of the desired vectors @xmath256 must be orthogonal to the vector @xmath263 = @xmath264 .
thus one can initially set @xmath149 vectors with random coefficients and use a gram - schmidt procedure initiated with the vector @xmath263 , orthogonalizing all subsequent vectors against @xmath263 and previously generated vectors .
the final two eigenvectors for @xmath265 and @xmath266 are easily found to have the @xmath267 components @xmath268 and @xmath269 .
their @xmath270 components are @xmath271 and , finally , for the @xmath272 components , @xmath273 .
these two eigenvectors are orthogonal to each other and the previous @xmath149 eigenvectors associated with the degenerate eigenvalue , and we find it convenient to employ them in this way with normalization considerations entering into the propagator representation , .
with the systematic procedure above for evaluating all the eigenvectors , and introducing the time - dependent amplitudes @xmath274 corresponding to states @xmath275 within the basis @xmath276 , @xmath277 , , @xmath278 , one can use ( extended to @xmath174 states , of course ) to show @xmath279 where @xmath280 the probabilities for qds @xmath281 to be excited are @xmath282 , @xmath283 , @xmath284 . while obtaining bipartite concurrences
may appear arduous , if @xmath285 ( i.e. , no amplitude in the state corresponds to the plasmon excited with all qds cold ) and all the other amplitudes are real , one can show that the bipartite concurrences are simply @xmath286 . as with the three - state example
, we note that as @xmath287 , only the @xmath288 eigenvector contributions survive and one could use , setting the exponential to one and carrying the sum out to only @xmath289 , to evaluate the asymptotic populations .
to estimate rabi - flop frequencies for the qds , we need an estimate of the local electromagnetic field they experience , which is enhanced relative to the incident field due to the presence of the plasmonic system . to this end , we consider the interaction of one qd with a plasmonic system and employ a classical coupled dipole picture , as in ref . and associated supplementary material .
the time - dependent dipoles for the plasmon ( @xmath290 ) and qd ( @xmath291 ) , in the presence of an incident field with frequency @xmath292 satisfy the equations of motion @xmath293\ ] ] @xmath294.\ ] ] the parameters @xmath167 , @xmath295 , and @xmath24 are the same as those in the cqed model of sec .
[ sec : theoretical_methods ] .
the other parameters in these classical equations are related to those in the cqed model as follows : @xmath296 several comments are in order regarding these relations .
the relation for @xmath297 was derived in ref . .
the relations for @xmath298 and @xmath299 reflect exactly solving and in the limit of the dipoles not interacting ( @xmath300 ) and equating the resulting amplitudes of oscillation of the dipoles with the corresponding quantum expressions ( in the linear or low @xmath70 limit ) .
these expressions are twice as small as the previously inferred ones , which were less accurate because they were based on an approximate solution of the classical equations .
the classical decay factor @xmath301 is taken to be twice the corresponding quantum dephasing factor , @xmath82 .
this ensures that the full - width - at - half - maximum of the isolated qd spectrum , inferred from the classical expression with @xmath301 , is equal to the corresponding quantum result in the low @xmath70 limit .
we can identify the term @xmath302 in as the local electric field the qd experiences because of the plasmon , that is , @xmath303 for estimating @xmath304 , one can approximate @xmath290 by the expression that results from the exact solution of in the uncoupled ( @xmath300 ) and on resonance ( @xmath305 ) limits . this solution is readily obtained by complexifying the equation , that is , by replacing @xmath306 by @xmath307 , which leads to an equation that is easy to solve exactly .
the real part of the complex solution then solves the original , real equation .
thus , @xmath308 insertion of into leads to @xmath309 where @xmath310 where the expressions in have also been used . | a system of two or more quantum dots interacting with a dissipative plasmonic nanostructure is investigated in detail by using a cavity quantum electrodynamics approach with a model hamiltonian .
we focus on determining and understanding system configurations that generate multiple bipartite quantum entanglements between the occupation states of the quantum dots .
these configurations include allowing for the quantum dots to be asymmetrically coupled to the plasmonic system . analytical solution of a simplified limit for an arbitrary number of quantum dots and numerical simulations and optimization for the two- and three - dot cases are used to develop guidelines for maximizing the bipartite entanglements . for any number of quantum dots , we show that through simple starting states and parameter guidelines , one quantum dot can be made to share a strong amount of bipartite entanglement with all other quantum dots in the system , while entangling all other pairs to a lesser degree . |
of the @xmath6 150 distinct molecules detected thus far in interstellar clouds and circumstellar outflows , over 90@xmath7 contain only hydrogen and/or elements in groups iva , va , and via of the periodic table : viz .
carbon , nitrogen , oxygen , silicon , phosphorus , and sulfur . of the remaining 12 detected molecules ,
eight contain elements in groups ia , iia and iiia ( viz .
sodium , potassium , magnesium and aluminum ) and are found exclusively in circumstellar envelopes ( e.g. ziurys 1996 , and references therein ) with abundances that are set in cool stellar photospheres ; one contains iron ( feo , detected toward the sgr b2 interstellar gas cloud by walmsley et al .
2002 ) ; and the remaining 3 are interstellar molecules containing elements of the halogen group ( viia ) : hcl , hf , and cf@xmath0 .
although they are only minor constituents of the interstellar medium , halogen - containing molecules are of special interest because of their unique thermochemistry .
the basic behavior of any atom in the interstellar medium is determined by two or three key thermochemical considerations .
first , the first ionization potential ( ip ) determines whether the atom will be neutral or ionized within diffuse molecular clouds : an atom x with ip(x ) @xmath8 ip(h ) = 13.6 ev will be shielded by atomic hydrogen from the interstellar radiation field ( isrf ) and will be predominantly neutral , while an atom with ip(x ) @xmath9 ip(h ) will be predominantly ionized .
second , the dissociation energy of the neutral hydride , @xmath10(hx ) , determines whether x can react exothermically with h@xmath11 : atoms for which @xmath10(hx ) @xmath8 @xmath10(h@xmath11 ) = 4.48 ev can react exothermically with h@xmath11 to form hx , while atoms with @xmath10(hx ) @xmath12(h@xmath11 ) can not .
third ( and of importance mainly for atoms with ip @xmath9 13.6 ev ) , the dissociation energy of the hydride molecular ion , @xmath10(hx@xmath0 ) , similarly determines whether the ion x@xmath0 can react exothermically with h@xmath11 to form hx@xmath0 .
fluorine and chlorine are unique in regard to the second and third criteria , each in a different way : f is the only atom and cl@xmath0 is the only ion with an appearance potential @xmath9 13.6 ev that can react exothermically with h@xmath11 to form a diatomic hydride .
the uniqueness of these elements are presented graphically in figure 1 , which shows @xmath10(hx ) for each atom x in the first three rows of the periodic table .
the molecular binding energy of hx is largest for fluorine and decreases , nearly monotonically , as one moves leftwards and downwards in the periodic table . in figure 2 , @xmath10(hx ) and @xmath10(hx@xmath0 ) are shown for all the elements of which interstellar molecules have been detected .
red squares apply to atoms with ip @xmath9 ip(h ) and blue squares to those with ip @xmath8 ip(h ) . with the exception of the halogen elements , all the elements fall into two classes : \(1 ) _ oxygen and nitrogen , with ip @xmath8 ip(h ) , @xmath10(hx ) @xmath9 @xmath10(h@xmath11 ) and @xmath10(hx@xmath0 ) @xmath8 @xmath10(h@xmath11 ) . _
these elements are predominantly neutral in the diffuse ism .
their atoms do not react with h@xmath11 except at elevated temperatures that can drive endothermic reactions , and their chemistry is driven by cosmic ray ionization .
\(2 ) _ carbon , silicon , sulfur , and phosphorus , with ip @xmath9 ip(h ) , @xmath10(hx ) @xmath9 @xmath10(h@xmath11 ) and @xmath10(hx@xmath0 ) @xmath9 @xmath10(h@xmath11 ) . _
these elements are predominantly ionized in the diffuse ism .
their ions do not react with h@xmath11 except at elevated temperatures that can drive endothermic reactions .
chlorine and fluorine form two additional classes with one member each : \(3 ) _ fluorine , with ip @xmath8 ip(h ) and @xmath10(hx ) @xmath8 @xmath10(h@xmath11 ) .
_ fluorine is predominantly neutral in the diffuse ism , and can react exothermically with h@xmath11 to form hf .
\(4 ) _ chlorine , with ip @xmath9 ip(h ) ,
@xmath10(hx ) @xmath9 @xmath10(h@xmath11 ) and @xmath10(hx@xmath0 ) @xmath8 @xmath10(h@xmath11 ) .
_ chlorine is predominantly ionized in atomic clouds .
cl@xmath0 can react exothermically with h@xmath11 to form hcl@xmath0 .
the chemistry of interstellar fluorine has been considered in a previous study by neufeld , wolfire & schilke ( 2005 ; hereafter nws05 ) , while that of chlorine has been discussed in several studies , most recently those of schilke , wang & phillips ( 1995 ) and amin ( 1996 ) . in the present paper , we extend and update these previous investigations to account for recent developments in experimental and theoretical studies of relevant chemical processes .
our study is also motivated by the prospects for upcoming observations of halogen - containing molecules , particularly with the _ herschel space observatory_.
the key chemical pathways leading to the formation and destruction of chlorine - bearing molecules have been elucidated in the papers of jura ( 1974 ) , dalgarno et al .
( 1974 ) , van dishoeck & black ( 1986 ; hereafter vdb86 ) , blake , anicich & huntress ( 1986 ; hereafter bah86 ) , schilke , phillips & wang ( 1995 ) , federman et al .
( 1995 ) , and amin ( 1996 ) . in diffuse molecular clouds ,
the chemical network is initiated by reaction of cl@xmath0 with h@xmath11 : @xmath13 which is followed by dissociative recombination of hcl@xmath0 or a further h atom abstraction reaction to form h@xmath11cl@xmath0 : @xmath14 the @xmath4 molecular ion undergoes dissociative recombination to form either hcl or cl @xmath15 @xmath16 the branching ratio for dissociative recombination of h@xmath11cl@xmath0 has not yet been measured , but has been assumed in previous studies to be only 10@xmath7 for production of hcl ( reaction 3 ) .
this assumption was largely motivated by the small hcl abundances observed in diffuse clouds ( see 4 below . )
hcl is destroyed by photodissociation , by photoionization and by reactions with he@xmath0 , h@xmath17 and c@xmath0 . in regions where h@xmath11 is present , reactions ( 1 ) ,
( 2 ) and ( 4 ) can rapidly convert cl@xmath0 to cl .
the rate at which the reaction network is initiated ( reaction 1 ) is therefore limited by the availability of cl@xmath0 ions .
these are produced primarily by the photoionization of cl , following the absorption of ultraviolet photons in the @xmath18 wavelength range .
near cloud surfaces , the uv radiation field is dominated by starlight ( i.e. by the interstellar radiation field ( isrf ) or nearby ob associations ) , whereas in the deep interiors of dense clouds , the radiation largely comprises h@xmath11 lyman and werner band emissions excited by the secondary electrons produced by cosmic rays . here , proton transfer from h@xmath17 to cl , producing hcl@xmath0 , can also initiate the formation of cl - bearing molecules . taking as our starting point the reaction list and compilation of rate coefficients in the udfa ( formerly umist ) rate file ( woodall et al .
2007 ) , we have reviewed the rate coefficients critically and implemented the following changes : \1 ) for reaction of hcl with he@xmath0 , h@xmath17 and c@xmath0 , we have used the statistical adiabatic capture model ( sacm ; troe 1985 , 1987 , 1996 ; see nws05 for details of exactly how this was implemented ) to estimate the temperature dependence of the capture rate .
over the temperature range 10 300 k , we found the capture rate to be proportional to @xmath19 for rotationally - cold hcl ; accordingly we adopted a @xmath19 dependence for the assumed reaction rate .
similar considerations lead to a @xmath20 temperature dependence for the rate coefficient for the reaction of h@xmath11o with h@xmath11cl@xmath0 .
we also adjusted the t = 300 k rate coefficients in the udfa compilation , which were based upon measurements of the thermal rate coefficients , to account for the fact that hcl is primarily in j = 0 ( i.e. rotationally - cold ) under the astrophysical conditions of typical interest ; here again , we used the sacm model to estimate the necessary correction factors .
2 ) as in nws05 , we adopted a rate coefficient of @xmath21 for the dissociative recombination of diatomic molecular ions ( in the absence of experimentally - determined values ) . for the dissociative recombination of h@xmath11cl@xmath0
, we adopted a rate coefficient of @xmath22 , based upon preliminary results from a recent laboratory measurement ( geppert & hamberg 2009 ) .
because the branching ratios for the production of hcl and cl were not determined in this experiment , we followed previous authors in adopting - as our standard assumption - a branching ratio of 0.1 for the production of hcl .
as discussed in 4.2 below , we also investigated how the model predictions depend upon the assumed value of this branching ratio .
\3 ) we adopted the rate coefficients of pradhan & dalgarno ( 1994 ) for charge transfer from h@xmath0 to cl , and for the reverse reaction ( from cl@xmath0 to h ) .
these reactions were not included in the udfa compilation , but had been considered by bah86 , who assumed a large rate coefficient of 10@xmath23 for the reaction of h@xmath0 with cl and concluded that this process was among the subset of most significant reactions .
the calculations of pradhan & dalgarno ( 1994 ) , however , imply rate coefficients that are smaller than those guessed by bah86 by factors of @xmath24 and @xmath25 at temperatures of 300 k and 100 k respectively .
\4 ) we have also included the reaction of cl with h@xmath11 to form hcl and h , which is slightly endothermic ( by @xmath26 ev ) , along with its reverse reaction .
this reaction , which is not included in the udfa compilation , has been the subject of intensive investigation , both experimental ( e.g. kumaran , lim & michael 1994 ; hereafter klm94 ) and theoretical ( e.g. manthe , capecchi & werner 2004 ) .
here we adopted the rate coefficients recommended by klm94 ( their equations 10 , 11 and 6 ) , which were obtained by a fit to their own measurements together with those obtained in previous studies .
\5 ) in computing revised estimates of the hcl photodissociation and photoionization rates , we have used the photoabsorption cross - sections obtained by brion , dyke & cooper ( 2005 ) by means of dipole ( e , e ) spectroscopy using an electron energy loss spectrometer ( eels ) .
this technique is believed to offer significant advantages over direct uv absorption measurements for the case of strong absorption bands in which line saturation is important ( chan , cooper & brion 1991 ) for example , in the case of excitation to the m@xmath27 state by photons near 110.3 nm , brion et al . obtained an oscillator strength that was a factor of 4 larger than that obtained in the earlier uv absorption experiments of nee , suto & lee ( 1986 ) .
the continuous absorption cross - sections obtained by brion et al .
( 2005 ) are in good agreement with previous uv absorption measurements ( e.g. cheng et al .
2002 ) . to determine the photodissociation rate implied by the measured photoabsorption cross - sections
, the photodissociation probability must also be known . for radiation longward of 115 nm ,
the fluorescence yield has been measured ( nee et al .
1986 ) and found to be small , so the photodissociation probability is very close to unity . for absorption in the 105.5 115 nm region , the fluorescence at wavelengths longer than 115 nm
was observed to be weak , but the experiment of nee et al .
( 1986 ) did not exclude the possibility of strong fluorescent emission in the 105.5 115 nm region .
for the 97.3 105.5 nm spectral region , we are unaware of any experimental constraints on the fluorescent yield . in computing the photodissociation rate ,
we have assumed that the dissociation probability is unity .
this therefore constitutes a lower limit on the true value , and is in good agreement with the corresponding minimum value given by vdvhd . ] for all radiation longward of the ionization threshold at @xmath28 nm .
this leads to a photodissociation rate of @xmath29 for an unshielded molecule exposed to the mean interstellar radiation field given by draine ( 1978 ) , a value that is @xmath30 higher than that given by van dishoeck , van hemert & dalgarno ( 1982 ; hereafter vdvhd ) .
this difference is largely accounted for by the fact that the total oscillator strength shortward of 129 nm is now known to exceed significantly the simple estimate given by vdvhd .
we obtain an hcl photoionization rate of @xmath31 . in considering the effect of shielding upon the hcl photodissociation rate
, we used a method similar to that adopted by nws05 . here , we modeled the case of a semi - infinite slab that is illuminated isotropically , using the photodissociation region ( pdr ) model of le bourlot et al .
1993 and le petit et al .
2006 to determine how the uv radiation field varies with position .
we assumed the photodissociation rate to diminish as @xmath32 , where @xmath33 is the visual extinction in magnitudes behind the slab surface , @xmath34 is the exponential integral of order 2 , and @xmath35 is an adjustable parameter that we varied in order to obtain the best fit to the predicted photodissociation rate .
the uv dust albedo assumed here was 0.32 , the forward scattering function was 0.73 ( li & draine 2001 ) , and the best fit was obtained for @xmath36 .
\6 ) our estimate of the cl photoionization rate was based upon measurements of the ionization cross - section by ruscic & berkowitz ( 1983 ) , which are in good agreement with theoretical calculations of brown , carter & kelly ( 1980 ) .
two autoionizing resonances are present at wavelengths longer than the lyman limit , but make a negligible contribution to the overall photoionization rate .
we derived a photoionization rate of @xmath37 for an unshielded chlorine atom exposed to the mean interstellar radiation field given by draine ( 1978 ) , a value that lies a factor of @xmath38 below that adopted by van dishoeck ( 1988 ) .
because the ip of atomic chlorine is only slightly less than that of h , there is only a narrow wavelength range , 912 958@xmath39 , over which photoionization can occur .
radiation in this spectral region is subject not only to absorption by dust , but also to absorption by the h@xmath11 lyman and werner bands and the hi lyman lines . while the effect of hi absorption is small , that of h@xmath11 absorption can be significant .
we have considered dust and h@xmath11 absorption independently , by running the le petit et al .
( 2006 ) pdr model both with and without h@xmath11 absorption .
we have found that the reduction in the cl photoionization rate can be represented to good approximation by a factor @xmath40 where @xmath41 is the column density of h@xmath11 to the cloud surface .
the photoionization of hcl occurs over a similarly narrow range of wavelengths .
for that process , an identical treatment yields a reduction factor @xmath42 \7 ) using the cross - sections cited in ( 5 ) and ( 6 ) above , we have computed the rates of hcl photodissociation , hcl photoionization , and cl photoionization resulting from h@xmath11 lyman and werner band emissions excited by secondary electrons produced by cosmic rays . here
we used the same methodology adopted in nws05 .
\8 ) for the radiative recombination of cl@xmath0 , we adopt the value recommended by mazzotta et al .
dielectronic recombination is negligible at the temperatures of present interest
. our model does not include the photodissociation of molecular ions containing the halogen elements . with the exception of hcl@xmath0 , for which photodissociation cross - sections have been presented by pradhan , kirby & dalgarno ( 1991 ) and by andric et al .
( 2002 ) , we are not aware of any estimates of the relevant cross - sections .
the hcl@xmath0 photodissociation cross - sections imply an unshielded photodissociation rate of @xmath43 . in the region where hcl@xmath0 is most abundant
, photodissociation makes a negligible contribution - relative to dissociative recombination - to the hcl@xmath0 destruction rate ; we assume photodissociation to be similarly negligible for cf@xmath0 and h@xmath11cl@xmath0 .
the full reaction list for chlorine - bearing species is given in table 1 .
it contains 21 processes involving the following species : cl , hcl , cl@xmath0 , hcl@xmath0 , h@xmath11cl@xmath0 and ccl@xmath0 .
the chemical network for fluorine - bearing species has been discussed by nws05 .
the chemistry is initiated by reaction of h@xmath11 with f : @xmath44 once the h@xmath11 abundance becomes appreciable , hf becomes the dominant reservoir of fluorine ; it is destroyed slowly by photodissociation , and by reactions with he@xmath0 , h@xmath17 and c@xmath0 .
reaction with c@xmath0 leads to the formation of cf@xmath0 : @xmath45 the fluoromethylidynium ion , cf@xmath0 , is destroyed rapidly by dissociative recombination , but can nevertheless account for as much as @xmath46 of the gas - phase fluorine abundance .
it has recently been detected in the interstellar medium ( neufeld et al .
2006 ) . for completeness
, we have implemented five updates to the chemical network presented by nws05 . of the following changes ,
the first three of which were noted in the udfa reaction list ( woodall et al .
2007 ) , only the first is expected to have any significant effect on the model predictions : \1 ) for the rate coefficient for dissociative recombination of cf@xmath0 , we now adopt the recent laboratory measurement of novotny et al .
the adopted rate coefficient lies factors of 4 and 1.5 below that assumed by nws05 at temperatures of 300 and 10 k respectively .
\2 ) we have corrected an erroneous value adopted by nws05 for the rate coefficient for reaction of f with h@xmath11o .
the correct value is @xmath47 .
this was stated correctly in the text of nws05 , but a value 10 times too large was erroneously given in the reaction list table ( and erroneously implemented in our calculation ) .
\3 ) for the reaction of h@xmath11 with f@xmath0 , we include a channel that produces h@xmath48 in addition to that which produces hf@xmath0 .
\4 ) for the cosmic ray induced photodissociation of hf , we include the effects of h lyman alpha radiation .
this results in a 50@xmath7 increase in the adopted photodissociation rate .
\5 ) for the radiative recombination and dielectronic recombination of f@xmath0 , we adopt the rate coefficients of badnell ( 2006 ) and badnell et al .
( 2003 ) .
the full reaction list for fluorine - bearing species is given in table 2 .
it contains 18 processes involving the following species : f , hf , f@xmath0 , hf@xmath0 , h@xmath11f@xmath0 , cf@xmath0 and sif@xmath0 . as in nws05 , we have used a modified version of the pdr models of kaufman et al .
( 2006 ) , kaufman et al .
( 1999 ) , wolfire et al .
( 1990 ) , and tielens & hollenbach ( 1985 ) to investigate the chemistry within diffuse and dense molecular clouds .
the modifications to these models have been largely described in nws05 , with the exception of the following : \1 ) for the radiative and dielectronic recombination of c@xmath0 , we adopt the recently - computed rate coefficients of badnell et al .
( 2006 ) and badnell et al .
( 2003 ) .
\2 ) in treating the depletion of fluorine and chlorine that results from freeze - out onto grain mantles within dense molecular clouds , we use equation ( 5 ) of nws05 , but with the hf or hcl photodissociation rate augmented so as to include the effects of h@xmath11 lyman and werner band emissions excited by the secondary electrons produced by cosmic rays . as in nws05 , we assume a gas - phase fluorine abundance of @xmath49 relative to h nuclei in diffuse molecular clouds . for chlorine ,
we assume an analogous abundance of @xmath50 , the average value observed in three sightlines that intersect diffuse molecular clouds : @xmath3 oph ( federman et al . 1995 ) , hd 192639 ( sonnentrucker et al . 2002 ) and hd 185418 ( sonnentrucker , friedman & york 2006 ) . in our standard models , as in nws05
, we assume a _ primary _ cosmic - ray ionization rate of @xmath51 per h _
nucleus_. recent observations of h@xmath17 have led to the postulate of substantially larger rates in diffuse clouds ( e.g. indriolo et al .
2007 ) . to investigate the effect of enhanced cosmic ray ionization rates upon the chemistry of interstellar halogens , we have also considered models for which the assumed @xmath52 is increased by a factor 10 .
our standard models also compute the thermal balance and equilibrium gas temperature as a function of distance into the cloud . to investigate
the possible effects of enhanced temperatures ( such as might result from the passage of interstellar shocks ) , we have also considered a series of models in which the temperature is set to some depth - independent value ( larger than that computed by our self - consistent treatment of the thermal balance ) .
for our standard models , we have followed previous authors in adopting a branching ratio of 0.1 for the production of hcl following dissociative recombination of h@xmath11cl@xmath0 . to study the dependence upon this branching ratio , we have also considered diffuse cloud models in which values of 0 , 0.3 and 1.0 were adopted .
we have constructed a series of one - sided models in which ultraviolet radiation is incident upon a semi - infinite slab , with the specific intensity constant over the incoming hemisphere .
each model is characterized by two key parameters : the density of h nuclei , @xmath53 , assumed constant throught the slab ; and the incident radiation field , @xmath54 , expressed in units of the mean isrf given by draine ( 1978 ) . for one - sided slab models ,
we have considered all combinations of @xmath55 , @xmath56 , @xmath57 , @xmath58 , and @xmath59 and @xmath60 , @xmath61 , @xmath56 , and @xmath57 . in figure 3 , we present example results for a single such combination : @xmath62 , and @xmath63 . here , we plot the relative abundances of h@xmath11 , cl , cl@xmath0 , hcl , hcl@xmath0 , h@xmath11cl@xmath0 , hf , and cf@xmath0 as a function of the depth into the cloud .
ccl@xmath0 fails to appear on this plot because its abundance is too low .
each abundance is normalized relative to the relevant _ gas - phase _ elemental abundance ( i.e. hydrogen for the case of h , chlorine for the case of cl - containing molecules , fluorine for the case of f - containing molecules ) .
the depth is measured in terms of the visual extinction , @xmath33 , in magnitudes , along a perpendicular ray to the cloud surface .
figure 4 shows the gas temperature and the f and cl gas - phase elemental abundances as a function of depth . in figures 3 and 4 ,
solid lines show the results for an assumed primary cosmic ray ionization rate of @xmath64 per h nucleus ( standard case ) , while dashed lines show the results for an assumed primary cosmic ray ionization rate of @xmath65 per h nucleus .
the effect of the increased cosmic ray ionization rate is negligible for @xmath66 , because the ionization of cl is dominated by the isrf near the cloud surface . at larger depths below the cloud surface ,
the predicted abundances of @xmath4 and @xmath67 are significantly enhanced if a value of @xmath65 is adopted for @xmath52 ; however , at these depths , the observed h@xmath17 abundances do _ not _ support an enhanced value for the cosmic ray ionization rate ( indriolo et al .
2007 ) . by integrating the molecular abundances into a depth @xmath68 mag ,
we have obtained the total column densities for each case . in figure 5 , we present the column densities of hcl , hcl@xmath0 , h@xmath11cl@xmath0 , hf , and cf@xmath0 . here , we have plotted the molecular column densities as a function of @xmath69 .
results are shown for four different values of the radiation field , @xmath70 , @xmath61 , @xmath56 , and @xmath57 .
we have also constructed two - sided models , in which a slab of finite thickness is illuminated from both sides .
while the one - sided models are primarily applicable to pdrs in which a hot star or star cluster irradiates a nearby molecular cloud , the two - sided models are applicable to diffuse or translucent molecular clouds that are exposed to the isrf . in figure 6 , we present the predicted column densities as a function of the total @xmath71 across the slab .
results are shown for clouds with density @xmath72 exposed to radiation with @xmath54 = 1 and 10 . in figure 7 , we present the same model predictions but with @xmath41 in place of @xmath71 as the independent variable .
figure 8 shows how the hcl and ccl@xmath0 abundances are enhanced at elevated gas temperatures .
here we show the results of one - sided models for slabs of density @xmath72 exposed to radiation with @xmath54 = 1 and 10 .
in addition to the standard model in which the temperature is determined by considerations of thermal balance , we show the abundance profiles for models in which the temperature is arbitrarily set to 150 , 200 , 300 or 400 k. even temperatures as low as 150 k enhance the hcl abundance by driving the slightly endothermic reaction of cl and h@xmath11 . the ccl@xmath0 abundance is increased accordingly , a result of the subsequent reaction of hcl with c@xmath0 .
as expected , the results shown in figures 3 7 indicate qualitatively - different behavior for the f- and cl - bearing species .
the results for f - bearing molecules are qualitatively the same as those obtained previously by nws05 : once h@xmath11 becomes abundant , hf becomes the dominant reservoir of gas - phase fluorine , with cf@xmath0 the only other significant molecule containing fluorine . because of the lower dissociative recombination rate assumed for cf@xmath0 in the present work , the predicted cf@xmath0 abundances are somewhat larger .
the chemistry for cl - bearing molecules is somewhat more complex , by contrast , and the computer - generated network diagrams shown in figure 9 , together with figures 10 - 13 in the online materials , help elucidate the key pathways . in figures 9 13
, circles represent the key cl - bearing species , with a color scale indicating their relative abundance .
the color map extends from cyan to blue to purple to red ( highest value ) , with grey circles denoting species that account for less than 10@xmath73 of the gas - phase chlorine abundance .
arrows indicate the key reactions linking the various species : again , the colors of the arrows indicate the reaction rate per unit volume .
the results apply to a one - sided model with @xmath74 and @xmath75 , and the different figures show the behavior at various depths into the cloud .
the depths were chosen to illustrate several different regimes .
\1 ) @xmath76 : at the cloud surface , there is no shielding and the abundance of molecules is small . the cl@xmath0/cl ratio is @xmath77 , being determined by the balance between photoionization and recombination . for the latter process , radiative recombination and charge transfer recombination with h although endothermic by @xmath6 0.6 ev are of roughly equal importance .
\2 ) @xmath78 : here , h@xmath11 self - shielding has greatly increased the h@xmath11 abundance .
reaction with h@xmath11 is now the dominant destruction mechanism for cl@xmath0 , resulting in the formation of hcl@xmath0 .
however , the e / h@xmath11 abundance ratio is still sufficient for dissociative recombination to dominate the destruction of hcl@xmath0 , although the competing reaction with h@xmath11 produces a small amount of h@xmath11cl@xmath0 .
the dissociative recombination of hcl@xmath0 leads to atomic cl , which becomes the dominant gas - phase reservoir of chlorine . beyond @xmath79 ,
the abundances of the hcl@xmath0 and h@xmath11cl@xmath0 ions start to fall , along with that of @xmath80 .
\3 ) @xmath81 : the e / h@xmath11 abundance ratio has now dropped sufficiently that hcl@xmath0 reacts primarily with h@xmath11 , rather than electrons , forming h@xmath11cl@xmath0 .
hcl is produced by dissociative recombination of h@xmath11cl@xmath0 , although the dominant formation process for hcl is the slightly endothermic reaction of h@xmath11 and cl ; this reaction is still possible at the @xmath82 k temperature at this point within the pdr . at @xmath81 , cl and hcl account
respectively for @xmath83 and @xmath84 of elemental chlorine in the gas phase .
\4 ) @xmath85 : the gas temperature drops below the level at which hcl can be formed by reaction of h@xmath11 with cl , and the hcl abundance is lower than it was at @xmath81 .
the cl@xmath0/cl ratio is now only @xmath86 , the photoionization rate of cl having dropped dramatically .
thus the formation of hcl@xmath0 is dominated by reaction of h@xmath17 with cl , not h@xmath11 with cl@xmath0 .
the h@xmath17 molecular ion is produced by the effects of cosmic rays ; these ionize h@xmath11 , forming h@xmath48 , which then transfers a proton to h@xmath11 to form h@xmath17 .
\5 ) @xmath87 : as at @xmath88 , hcl@xmath0 is produced primarily by reaction of h@xmath17 with cl , and reaction of hcl@xmath0 with h@xmath11 then forms h@xmath11cl@xmath0 . at this depth , the electron fraction is so low that dissociative recombination of h@xmath11cl@xmath0 is no longer dominant .
h@xmath11cl@xmath0 transfers a proton to a neutral species of higher proton affinity than hcl ( e.g. co or h@xmath11o ; this process is denoted by the arrow labeled
n ) , yielding hcl . at this point ,
the external uv is almost completely attenuated , and the hcl destruction rate is very small .
hcl now accounts for several @xmath89 of the gas - phase chlorine abundance , with atomic chlorine accounting for essentially all of the remainder .
our results in this regime are in good agreement with those obtained previously by schilke et al ( 1995 ) . h@xmath11cl@xmath0 and ccl@xmath0 are also present , but their abundances are just a few @xmath90 of gas - phase elemental chlorine .
hcl reacts with h@xmath17 to form h@xmath11cl@xmath0 , although this rapidly leads to reformation of hcl by proton transfer to other neutral species .
several other destruction processes are of comparable importance for hcl destruction : cosmic ray induced photodissociation , and reaction with the positive ions c@xmath0 and he@xmath0 . in this regime ,
the rates of hcl formation and destruction all scale linearly with the cosmic - ray ionization rate , with the result that the hcl abundance is almost independent of the latter .
although the diagrams shown in figure 9 13 apply to just a single model , similar regimes exist for other values of @xmath54 and @xmath69 .
although the relevant regions move in or out ( to larger or smaller @xmath33 ) as the assumed @xmath91 ratio is increased or decreased , the same qualitative behavior is observed . in examining similar diagrams for the entire set of models
, we have not identified any important reaction pathway that is not apparent in figures 9 13 . to date
, hf , cf@xmath0 , and hcl are the only halogen - bearing molecules to have been detected in the interstellar medium . in regard to hf , our results are identical to those presented in nws05 : we predict hf to be the dominant gas - phase reservoir of fluorine within both diffuse and dense molecular clouds ; we expect the _ herschel space observatory _ to detect widespread absorption in the hf @xmath1 transition .
however , the abundances we predict for cf@xmath0 lie a factor @xmath92 above the prediction of nws05 , a direct consequence of the smaller dissociation rate adopted for cf@xmath0 in the present study .
as noted by neufeld et al .
( 2006 ) , the cf@xmath0 column densities inferred from observations of the orion bar were already a factor of @xmath93 below the predictions of nws05 , and the discrepancy between theory and observation is now increased to more than an order of magnitude .
this disagreement may indicate that we have overestimated the rate coefficient for reaction of c@xmath0 and hf .
we are not aware of any laboratory studies of this reaction ; the rate we adopt is simply the capture rate , and thus an upper limit on the true reaction rate .
laboratory measurements of this key reaction would be very desirable .
hcl has been observed in both diffuse and dense molecular clouds . in diffuse clouds ,
ultraviolet absorption studies have led to upper limits , and in one case , a tentative detection toward @xmath3 oph ( federman et al .
the latter result , obtained using the goddard high resolution spectrograph on hst , yielded an hcl column density of @xmath94 . given an atomic chlorine column density of @xmath95 for this sight - line ( federman et al .
1995 ) , the corresponding @xmath96 ratio is @xmath97 ( uncertain by a factor @xmath38 ) , a value in excellent agreement with predictions of the diffuse cloud models presented by vdb86 .
very similar results are obtained from our updated treatment of cl chemistry . in figure 14 , we show predictions from a series of models with @xmath54 ranging from @xmath98 to 10@xmath99 .
all models apply to a slab with parameters appropriate to @xmath3 oph , viz .
density @xmath100 and total visual extinction of 0.8 mag .
the horizontal axis shows the h@xmath11 column density , for which the measured value is @xmath101 along this sight - line , while the vertical axis shows @xmath96 ( red curve ) , @xmath102 ( orange ) and @xmath103 ( blue ) .
squares located along the curves show the results for @xmath104 to 10@xmath99 from right to left .
the h@xmath11 column density along the @xmath3 oph sight - line requires a @xmath105 , in agreement with previous studies , and the predicted @xmath96 is in good agreement with the observations . the almost exact agreement between the present model and that of vdb86 in regard to the predicted hcl abundance appears to result from the fortuitious cancellation of two changes in the photochemical network : the photodissociation rate for hcl and the photoionization rate for cl are both increased by a factor @xmath38 . as in the vdb86 models ,
the predicted cl@xmath0 column density lies substantially below the observations .
federman et al . ( 1995 ) attributed this discrepancy to the presence of a significant cl@xmath0 column density within hii regions along the sight - line .
one important caveat must be noted .
the results shown in figure 14 were obtained by assuming a branching ratio of only 10@xmath7 for the production of hcl following dissociative recombination of h@xmath11cl@xmath0 .
this value was initially invoked primarily to _ explain _ the relative low hcl abundance derived from upper limits obtained with the _ copernicus _ satellite .
models that we obtained for other branching ratios ( 0 , 0.3 and 1.0 ) indicate that the hcl column density is linearly proportional to the branching ratio for values greater than 0.1 with a branching ratio of zero ( i.e. with no production of hcl from dissociative recombination of h@xmath11cl@xmath0 ) , a non - zero but negligible hcl column density ( @xmath106 ) results from the endothermic reaction of h@xmath11 with cl . in dense clouds
, hcl has been unequivocally detected in several sources by means of submillimeter observations of the @xmath1 emission line .
first detected using the kuiper airborne observatory toward orion ( blake , keene & phillips 1985 ) and then sgr b2 ( zmuidzinas et al .
1995 ) , the @xmath1 transition has also been observed using the ground - based caltech submillimeter observatory ( cso ) under good atmospheric conditions .
cso observations have led to additional detections toward orion a , mon r2 ( salez , frerking & langer 1996 ) and several other sources ( phillips et al .
2009 ) , as well as mapping observations toward omc-1 ( schilke , phillips & wang 1995 ) .
typical column densities derived for these dense clouds lie in the few @xmath107 to few @xmath108 range , in reasonable agreement with the predictions of our model for regions at high density exposed to strong uv radiation ( fig .
as noted in the observational studies cited above , depletion plays an important role in limiting the fractional abundance of hcl .
our study identifies two additional cl - bearing species that are potentially detectable : the molecular ions hcl@xmath0 and h@xmath11cl@xmath0 .
these ions are isoelectronic with oh and h@xmath11s respectively .
they are most abundant near cloud surfaces , where the photoionization rate for hcl is highest , and show column densities that are an increasing function of @xmath91 .
thus , pdrs subject to strong uv irradiation present attractive targets for searches for hcl@xmath0 and h@xmath11cl@xmath0 at millimeter and submillimeter wavelengths .
for example , in a pdr with @xmath74 and @xmath75 , column densities of @xmath109 and @xmath110 are predicted for hcl@xmath0 and h@xmath11cl@xmath0 .
these values apply to sight - lines perpendicular to the illuminated surface ; in edge - on pdrs , they can be significantly enhanced by limb - brightening .
the rotational spectrum of hcl@xmath0 has been measured using laser magnetic spectroscopy ( lubic et al . 1989 ) .
the lowest - lying rotational transition , the @xmath111 multiplet near 207.6@xmath112 m , lies in a wavelength range accessible to the hifi instrument on the _ herschel space observatory_. the rotational spectrum of h@xmath11cl@xmath0 has also been the subject of a laboratory study at high spectral resolution ; here , the measurements of araki et al .
( 2001 ) indicate that the lowest - lying rotational line of ortho - h@xmath11cl@xmath0 , the @xmath113 transition , lie near 189.2 and 188.4 ghz respectively for the @xmath114 and @xmath115 isotopologues .
( these transitions are split into 6 and 4 hyperfine components , covering @xmath116 and @xmath117 mhz respectively . )
this spectral region lies in the wing of the strong 183 ghz telluric water absorption line , but is nevertheless observable from ground - based observatories under favorable but by no means unusual
atmospheric conditions . in one of the earliest studies of interstellar chlorine chemistry ,
the possibility of detecting hcl@xmath0 in _ diffuse _ molecular clouds was discussed by jura ( 1974 ) , who considered the @xmath118 band that is accessible in the near - ultraviolet region . given an oscillator strength of @xmath119 for the strongest vibrational band ( pradhan , kirby & dalgarno 1991 ) , and given the abundances predicted in figure 17 , our results for @xmath3 oph would imply equivalent widths of only @xmath120 , well below the limit of detectability .
finally , we note that , unlike cf@xmath0 , the ccl@xmath0 ion is typically predicted to have an extremely low abundance .
although ccl@xmath0 is produced rapidly by reaction of hcl and c@xmath0 , these two species show very little overlap : the hcl abundance is small in the surface layers where c@xmath0 is abundant , unless the temperature is increased by shock heating , and the c@xmath0 abundance is small in the deep cloud interiors where hcl is relatively abundant .
enhanced ccl@xmath0 abundances are possible when shock heating is present along with uv irradiation .
we are very grateful to w. d. geppert and m. hamberg for informing us about recent laboratory results on h@xmath11cl@xmath0 dissociative recombination , in advance of their publication .
we thank paule sonnentrucker and paul bryans for several useful discussions .
we gratefully acknowledge the support of a grant from the nasa herschel science center s theoretical research / laboratory astrophysics program .
van dishoeck , e. f. 1988 , rate coefficients in astrochemistry .
proceedings of a conference held in umist , manchester , united kingdom ,
september 21 - 24 , 1987 .
editors , t.j .
millar , d.a .
williams ; publisher , kluwer academic publishers , dordrecht , boston , 1988 .
isbn # 90 - 277 - 2752-x .
lc # qb450 .r38 1988 .
, 1988 , 49 ll @xmath121 & @xmath122 @xmath123 + & @xmath124 @xmath125 + @xmath126 & @xmath127 @xmath123 + & @xmath128 @xmath125 + @xmath129 & @xmath130 + @xmath131 & @xmath132 + @xmath133 & @xmath134 + @xmath135 & @xmath136 + @xmath137 & @xmath138 + @xmath139 & @xmath140 + @xmath141 & @xmath142 + @xmath143 & @xmath144 + @xmath145 & @xmath146 + @xmath147 & @xmath148 + @xmath149 & @xmath150 + @xmath151 & @xmath136 + @xmath152 & @xmath153)]\,{\rm s}^{-1}\ , $ ] + & + @xmath154 + @xmath155 & @xmath156)]\,{\rm s}^{-1}\ , $ ] + & @xmath157^{1/2})$ ] + & + @xmath158 + + @xmath159 & @xmath160)]\,{\rm s}^{-1}\ , $ ] + & @xmath161^{1/2})$ ] + & + @xmath162 + @xmath163 & @xmath164 + @xmath165 & @xmath166 + @xmath167 & @xmath168 + @xmath169 & @xmath170 + ll @xmath171 & @xmath172 + & @xmath173 \ , cm^3 \,s^{-1}$ ] + @xmath174 & @xmath175 + @xmath176 & @xmath177 + @xmath178 & @xmath179 + @xmath180 & @xmath181 + @xmath182 & @xmath183 + @xmath184 & @xmath185 + @xmath186 & @xmath187 + @xmath188 & @xmath138 + @xmath189 & @xmath190 + @xmath191 & @xmath142 + @xmath192 & @xmath142 + @xmath193 & @xmath194 + @xmath195 & @xmath194 + @xmath196 & @xmath197 + @xmath198 & @xmath199 + @xmath200 & @xmath201 + @xmath202 & @xmath203)]\,{\rm s}^{-1}\ , $ ] + & + @xmath204 + fig . 1 dissociation energies , @xmath205 ( at 0 k in ev ) , for the neutral diatomic hydrides ( derived from values recommended in the nist chemistry web book and in armentrout & sunderlin 1992 , chen , clemmer & armentrout 1991 , 1993 ; binning & curtiss 1990 ; and huber & herzberg 1979 ) fig . 2 dissociation energies of hx and hx@xmath0 for x = h , c , n , o , f , si , p , s , and cl . atoms with ip @xmath9 ip(h ) appear in blue ; those with ip @xmath8 ip(h ) appear in red .
the dashed lines indicate the dissociation energy of h@xmath11 ( 4.48ev ) .
3 abundance profiles for a one - sided slab model with @xmath206 and @xmath207 .
solid lines show the results for an assumed primary cosmic ray ionization rate of @xmath208 per h nucleus ( standard case ) .
dashed lines show the results for an assumed primary cosmic ray ionization rate of @xmath65 per h nucleus .
4 profiles of temperature and f- and cl - depletion factor , for a one - sided slab model with @xmath206 and @xmath207 .
solid lines show the results for an assumed primary cosmic ray ionization rate of @xmath208 per h nucleus ( standard case ) .
dashed lines show the results for an assumed primary cosmic ray ionization rate of @xmath65 per h nucleus .
fig . 8 effect of enhanced temperatures on the abundances of hcl ( red curves ) and ccl@xmath0 ( green curves ) . results of one - sided models for slabs of density @xmath72 exposed to radiation with @xmath54 = 1 ( left panel ) and 10 ( right panel ) .
in addition to the standard model in which the temperature is determined by considerations of thermal balance ( solid curve in each set ) , we show the abundance profiles for models in which the temperature is arbitrarily set to 150 k ( dotted ) , 200 k ( dashed ) , 300 k ( dot - dash ) , and 400 k ( triple - dot - dash ) fig .
14 predictions from a series of models with @xmath54 ranging from @xmath98 to 10@xmath99 .
all models apply to a slab with parameters appropriate to @xmath3 oph , viz .
density @xmath100 and total visual extinction of 0.8 mag .
the horizontal axis shows the h@xmath11 column density , for which the measured value is @xmath101 along this sight - line , while the vertical axis shows @xmath96 ( red curve ) , @xmath102 ( orange ) and @xmath103 ( blue ) .
squares located along the curves show the results for @xmath104 , 1 , 10 and 10@xmath99 from right to left . | although they are only minor constituents of the interstellar medium , halogen - containing molecules are of special interest because of their unique thermochemistry . here , we present a theoretical study of the chemistry of interstellar molecules containing the halogen elements chlorine and fluorine .
we have modeled both diffuse and dense molecular clouds , making use of updated estimates for the rates of several key chemical processes .
we present predictions for the abundances of the three halogen molecules that have been detected to date in the interstellar medium : hf , cf@xmath0 and hcl . as in our previous study of fluorine - bearing interstellar molecules ,
we predict hf to be the dominant gas - phase reservoir of fluorine within both diffuse and dense molecular clouds ; we expect the _ herschel space observatory _ to detect widespread absorption in the hf @xmath1 transition .
our updated model now overpredicts the cf@xmath0 abundance by a factor @xmath2 relative to observations of the orion bar ; this discrepancy has widened because we now adopt a laboratory measurement of the cf@xmath0 dissociative recombination rate that is smaller than the estimate we adopted previously .
this disagreement suggests that the reaction of c@xmath0 with hf proceeds more slowly than the capture rate assumed in our model ; a laboratory measurement of this reaction rate would be very desirable .
our model predicts diffuse cloud hcl abundances that are similar to those predicted previously and detected tentatively toward @xmath3 oph .
two additional species are potentially detectable from photodissociation regions : the @xmath4 and @xmath5 molecular ions .
ortho-@xmath4 has its lowest - lying transition in the millimeter spectral region observable from the ground , and the lowest rotational transition of hcl@xmath0 is observable with _ herschel _ s hifi instrument . |
the magnetic properties of the cubic crystal cu@xmath0oseo@xmath1 are of great interest for several reasons .
first of all , having the space group @xmath7 without centre of inversion , this crystal becomes a cubic helimagnet below the critical temperature of about 58 k. moreover , it is the first cubic crystal beyond the class of itinerant magnetics with @xmath8 crystal structure @xcite , for which a - phase , associated with skyrmion lattice @xcite , has been recently observed @xcite .
secondly , being an insulator , cu@xmath0oseo@xmath1 has magnetoelectric properties @xcite , which provides a new physical significance in comparison with the @xmath8 crystals .
the interconnection between magnetization gradients and electric polarization makes this crystal potentially applicable for data storing devices and spintronics @xcite . and
thirdly , but not finally , the more complex than @xmath8 structure ( 16 magnetic copper atoms in two non - equivalent positions in the unit cell ) makes it an interesting object for studying the spin textures .
most of the known twisted magnetics possess symmetry lower than cubic .
their strong anisotropy orients magnetic helices in special crystallographic directions .
as opposed to them , in cubic magnetics without centre of inversion , the binding between magnetic and crystal structures can be so subtle , that the helix can be easily reoriented along any direction without essential change of its pitch and energy .
nevertheless , at the microscopic level , this seeming freedom is achieved by correlated tilts of discrete magnetic moments tightly bound with atoms in the crystal .
besides , the easier seems to be the phenomenology of the isotropic system as compared with anisotropic one , the more complex appears its description in terms of discrete spins , particularly when the magnetic helix is oriented by the field in some arbitrary direction . as it has been shown recently for the helimagnetics with @xmath8 structure ( mnsi - type ) , this freedom of the helix rotations
is achieved by extra tilts ( or canting ) of the local magnetic moments , which always accompany global spiralling , but are not described in the phenomenological theory . moreover ,
the canting still remains , when the spiralling disappears in strong magnetic field . as a rule , in papers discussing the spin canting some class of canted antiferromagnetics
is implied @xcite , where small relative tilts of magnetic moments belonging to different sublattices result in appearing of weak magnetization .
the phenomenon is called weak ferromagnetism , and it serves as direct evidence of the canting , because the tilt angles of individual spins are directly proportional to the observed spontaneous magnetization .
the canting in ferromagnetics is less elaborated , and it has been known for many decades that it is more difficult to observe small changes of strong magnetization than small magnetization of weak ferromagnetics . here , a refined experiment would be useful , e.g. on observation of forbidden bragg reflections of antiferromagnetic origin @xcite .
repeating small rotations ( cantings ) of spins belonging to neighbouring atomic planes can result in a global twist of crystal magnetization .
therefore , in chiral ferromagnetics , the twist and the canting are often mistaken for the same thing .
only recently , it was suggested how to distinguish between canting and twist in ferromagnetics @xcite .
it was shown that the canting should always appear when there are more than one magnetic atom in the unit cell .
it leads to coexistence of several spin helices within the helimagnet , which differ by phases and rotational planes ( figure [ fig : canting ] ) .
it was shown by example of cubic helimagnets with @xmath8 structure that the canting makes contribution to the magnetic energy comparable with that of global spiralling .
therefore , it should be taken into account when calculating propagation number of spin helices from the first principles , i.e. starting from the heisenberg microscopical model of ferromagnetism .
even more intriguing is the fact that , if the dzyaloshinskii moriya ( dm ) vectors are perpendicular to the bonds between magnetic atoms ( which is predicted by many quantum mechanical models ) , both twist and canting in mnsi - type crystals are determined by the same components of the dm vectors .
the study of the cu@xmath0oseo@xmath1 crystal , with its structure different from @xmath8 , also can help us to distinguish between general regularities of twisted cubic magnetics and particularities of the mnsi - type crystals .
the twist and the canting ( by example of a one - dimensional crystal with two magnetic atoms in the unit cell ) .
the canting of spin sublattices leads to the combination of two helices with the same propagation vector and a constant phase shift.,width=264 ] in this paper , a general theory of the spin canting in cubic helimagnetics is developed , which unifies the cases of the mnsi - type and cu@xmath0oseo@xmath1 crystals . a microscopical description of the spin structures is made for the case of weakly non - collinear spins .
the latter means that the angle between neighbouring spins is close to 0 or to @xmath9 , i.e. both the spiralling and canting are small .
it is just the case of the majority of itinerant magnetics with @xmath8 structure ( exemplified by mnsi ) and twisted ferrimagnetic cu@xmath0oseo@xmath1 .
we confine ourselves only to the case of the twist induced by the dm interactions ; the case of the twist owing to the frustrated exchange is beyond the scope of our consideration . in section [ sec :
balance ] , the definition of canting is given .
then , the transition to continuous approximation is performed ( section [ sec : continuous ] ) , and the density of the magnetic energy is calculated ( section [ sec : energy ] ) . in section [ sec : collinear ] , it is shown which simplifications can be made in the case , when , in the absence of the dm interaction , all the spins of the system remain collinear .
we assert that two different kinds of the canting exist in the twisted magnetics .
the first of them is connected with the gradients of the magnetization , and it arises even in the absence of the dm interaction ( in this case the magnetization gradients can be induced by an external influence , e.g. from a surface ) . in order to take this kind of canting into account , it is convenient to use the method of fictitious `` exchange '' coordinates ( section [ sec : collinear ] )
. then , simple expressions can be found for the macroscopic constants @xmath4 and @xmath5 , and the helix propagation number @xmath10 ( section [ sec : phenomenological ] ) .
the second kind of canting is induced by the dm interactions .
it can be described in terms of the `` tilt '' vectors , whose symmetry is found to be determined by the symmetry of the crystal ( section [ sec : collinear ] ) .
the canting of this kind still remains even in the ferromagnetic state , when the helical spin structure is unwound by the external magnetic field .
this residual canting can be measured , using neutron and x - ray diffractions ( see discussion in section [ sec : discussion ] ) . in sections [ sec :
mnsi ] and [ sec : cu2oseo3 ] , the theory is applied to the cases of mnsi and cu@xmath0oseo@xmath1 , respectively .
in the classical heisenberg theory of magnetics , the energy of a magnetic structure is written as @xmath11 \right ) - \mu_\mathrm{b } g_i \mathbf{h } \cdot \mathbf{s}_i \right\ } , \ ] ] where @xmath12 enumerates all the magnetic atoms , the inner sum ( @xmath13 ) is taken over close neighbours of @xmath12th atom , the coefficient @xmath14 is needed because each bond is included twice in the sum , @xmath15 is the magnetic moment of @xmath12th atom ( @xmath16 is the direction of classical spin , @xmath17 ) , @xmath2 are the isotropic exchange constants , whereas the antisymmetrical exchange is characterized by the dzyaloshinskii
moriya ( dm ) vectors @xmath3 ( @xmath18 , @xmath19 ) , @xmath20 is an external magnetic field .
the equation of @xmath12th spin balance can be easily found copying out the part of ( [ eq : e1 ] ) associated with the @xmath12th spin , @xmath21 \right ) + \mu_\mathrm{b } g_i \mathbf{h } \right\ } \cdot \mathbf{s}_i \equiv -\boldsymbol{\varepsilon}_i \cdot \mathbf{s}_i\ ] ] ( notice the absence of @xmath14 in comparison with ( [ eq : e1 ] ) ) .
it is evident that @xmath22 is minimal , when @xmath23 being of relativistic origin , the dm interaction is considerably weaker than the isotropic exchange , @xmath24 therefore it is convenient to develop a perturbation theory with a small parameter of the order of @xmath25 . in the case of twisted magnetics
we can also neglect the influence of the external magnetic field on canting , supposing that @xmath26 where @xmath27 is the field of full `` unwinding '' of the magnetic helix .
indeed , it has been found in @xcite that the energy of canting is of the order of @xmath28 , while the energy of the interaction of canting with an external magnetic field ( @xmath29 ) is determined by the canting of the second order and being proportional to @xmath30 .
it follows from the finding of @xcite that , in ferromagnetics , the canting of the first order has antiferromagnetic character and does not contribute in this part of the magnetic energy .
notice that , accordingly to the elementary phenomenological theory , the field does not affect the helix pitch , which is in a good agreement with experimental data .
neglecting the dm interaction and external magnetic field , equation ( [ eq : si ] ) determines an equilibrium system of spins @xmath31 with @xmath32 the isotropy of the exchange interaction results in that both the structure energy and spin balance remain unchanged , if all the spins rotate as a unit by arbitrary angle relative to immovable atomic structure , or , which is practically the same , if the atomic structure rotates preserving initial spin directions .
the latter understanding is important for a cubic crystal structure , where equivalent atomic positions connect to each other by rotational symmetry transformations .
it is obvious that a spin balance can be achieved , when @xmath33 are equal for all the equivalent positions ( figure [ fig : equiv ] ) .
here we suppose that the twist is induced only by the dm interactions , and leave the possibility of the pure exchange spiralling outside the scope of our consideration .
notice that such pure exchange twist reveals a threshold behaviour , i.e. it can appear only at some values of the exchange constants .
three magnetic atoms ( green ) in positions connected by rotational symmetry elements , and their nearest magnetic environments . in absence of non - symmetric dm interaction ,
a balance is achieved , when the spins in the equivalent positions ( equally coloured circles ) have the same directions.,width=264 ] the non - isotropic dm interaction breaks the symmetry of @xmath33 , because dm vectors of equivalent bonds change ( rotate ) with the bond directions .
therefore spin @xmath16 deviates from @xmath33 by a small angle of order of @xmath25 .
taking into account that @xmath34 , this spin change , hereinafter referred to as `` canting '' , in the first approximation is perpendicular to @xmath33 . introducing correction @xmath35 of first order on @xmath25 , we find @xmath36 where @xmath37 and @xmath38 are parallel and perpendicular to @xmath33 components of @xmath35 , correspondingly ; @xmath39 is the canting of @xmath12th spin in the first approximation , and @xmath40 is proportional to @xmath28 change of the spin along @xmath33 ( figure [ fig : circle ] ) .
deviation of spin @xmath41 from the equilibrium state @xmath42 in presence of non - isotropic dm interaction . the tilt angle @xmath43 .
in a simple approximation the deviation can be decomposed into two parts , @xmath44 and @xmath45 .
, width=264 ]
in cubic crystal without centre of inversion , a constant direction of magnetic moment density vector is energetically unfavourable , and magnetic structure becomes twisted . besides , the strength of the twist is characterized by the same ratio @xmath25 as the canting does . if the magnetic moment density changes slowly along the crystal , a transition can be performed from discrete spins to smooth continuous spin functions @xmath46 , whose number coincides with the number of magnetic atoms in the crystal unit cell .
then , the magnetic energy can be rewritten in integral form @xcite , @xmath47 \right ) - \mu_\mathrm{b } g_i \mathbf{h } \cdot \hat{\mathbf{s}}_i \right\ } , \end{aligned}\ ] ] where the sum ( @xmath12 ) is now taken over all the magnetic atoms in the unit cell , and @xmath48 is a bond directed from @xmath12th to @xmath13th atom ; both @xmath49 and @xmath48 are measured in cubic cell parameters @xmath50 ; the values of the spin functions and their spatial derivatives in the integrand are taken at the same point . using ( [ eq : ei ] ) , the vector @xmath51 can be written in continuous approximation , @xmath52 \right ) , \ ] ] where we neglect the magnetic field and replace @xmath53 and @xmath54 with continuous functions @xmath55 and @xmath56 . here , we use condition ( [ eq : d / j ] ) and make the expansion , assuming that not only canting in the first approximation but also the global gradients of the magnetic structure ( e.g. the helix propagation number ) are determined by small parameter @xmath25 , i.e. @xmath57 .
then , @xmath58 '' designates vector projection on the plane perpendicular to @xmath59 . because @xmath60 and @xmath61 , then each summand in the sum over @xmath13 is a vector perpendicular to @xmath62 .
equation ( [ eq : u1 ] ) shows that the canting is determined by both the dm interaction and spatial derivatives of @xmath63 .
let us average out ( [ eq : u1 ] ) using the crystal symmetry .
for any cubic ( or tetrahedral ) point group @xmath64 where index `` eq '' means averaging over equivalent bonds and positions .
thus , we obtain @xmath65 ) , that condition ( [ eq : meanu ] ) is satisfied , when @xmath66 , \ ] ] with @xmath67 $ ] being the change of spin @xmath59 , induced by rotation of all the spins as a unit by a small angle @xmath68 . as is mentioned above
, the rotation does not change the isotropic exchange energy and results in new functions @xmath59 , also satisfying ( [ eq : eps0 ] ) .
therefore it is possible to include @xmath69 into the definition of @xmath59 and assume without loss of generality that @xmath70 in this case , @xmath59 is the normalized average spin over all the positions in the unit cell equivalent to the @xmath12th one .
let us write out the contributions to the energy ( [ eq : e2 ] ) up to the second order terms on @xmath25 .
the zero order energy is @xmath71 the 1st order contribution @xmath72 \right\}\end{aligned}\ ] ] becomes zero in the equilibrium due to ( [ eq : meanbd ] ) and ( [ eq : meanu3 ] ) .
it is convenient to divide the 2nd order contribution into two parts : the first one , associated with the gradients of the magnetic moment @xmath63 only , @xmath73 \right\ } , \end{aligned}\ ] ] and the second one , depending on local cantings , @xmath74 + [ \hat{\boldsymbol{\mu}}_i \times \hat{\mathbf{u}}^{(1)}_j ] \right ) \right\ } .\end{aligned}\ ] ] the expression for @xmath75 can be averaged over equivalent bonds , using equation @xmath76 for the vectors @xmath77 and @xmath78 transformed with the elements of the cubic ( tetrahedral ) point group .
then @xmath79 \right\ } .\end{aligned}\ ] ] notice that all the summands corresponding to equivalent bonds are equal , and , therefore , we can calculate them only once , multiplying then by the bond multiplicities .
changing summation order and using relations @xmath80 and @xmath19 , equation ( [ eq : en2u1 ] ) can be rewritten as @xmath81 \right\ } .\end{aligned}\ ] ] using ( [ eq : eps0 ] ) , ( [ eq : u2par ] ) and ( [ eq : u1 ] ) , the latter expression can be transformed to @xmath82
in frustrated magnetic structures , the spins @xmath59 can be non - collinear . besides , in a number of practically important cases , including ferro- , antiferro- , and some ferrimagnetic orders , all @xmath59 can be aligned along a line , and several considerable simplifications can be made .
indeed , assume that @xmath83 here , @xmath63 is a unit vector directed along the local magnetization .
then all @xmath84 belong to the plane perpendicular to @xmath63 , and we can discard the index `` @xmath85 '' in ( [ eq : u1 ] ) : @xmath86 \right\ } , \end{aligned}\ ] ] where is used that @xmath87 the system ( [ eq : xideal ] ) contains several independent vector equations ( whose number coincides with the number of non - equivalent magnetic positions in the unit cell ) and can be considered as a condition imposed on the atomic coordinates .
indeed , the energy ( [ eq : e1 ] ) does not depend explicitly on the atomic coordinates , as opposed to expression ( [ eq : e2 ] ) , where they are included in vectors @xmath48 .
obviously , the coordinates should disappear after minimization of the energy , therefore we can choose them arbitrarily , e.g. using condition ( [ eq : xideal ] ) .
it is shown in @xcite , that the possibility of intentional choice of ideal atomic coordinates is connected with the ambiguity of transition from discrete spins to continuous density of magnetic moment .
notice that , because ( [ eq : xideal ] ) contains only exchange interaction parameters , these fictitious coordinates depend only on @xmath2 .
so , hereinafter we will refer to the positions as `` exchange '' ones . in section [ sec : discussion ] the sense and properties of the exchange coordinates are discussed in details . the atomic `` shift '' from real to fictitious positions results in disappearing of the part of the canting induced by derivatives of @xmath63 : @xmath88 \right ) .\ ]
] now the canting can be found in form @xmath89 , \ ] ] where `` tilt '' vectors @xmath90 possess the following properties .
( i ) the vector @xmath90 corresponds to the symmetry of the @xmath12th atomic position , e.g. the tilt vector of an atom in position @xmath91 of the space group @xmath7 is directed along 3-fold axis of symmetry passing through the atom , whereas the tilt vector of an atom in position @xmath92 has three independent components .
( ii ) the tilt vectors in equivalent positions are connected to each other by the corresponding transformation of the point group .
now the system ( [ eq : u1 - 3 ] ) can be rewritten as @xmath93 the system ( [ eq : rho ] ) contains as many independent vector equations as many non - equivalent magnetic atoms are in the unit cell .
the energy density expressions are also simplified in the case of collinear mean spins : @xmath94 @xmath95 .\end{aligned}\ ] ] using equation @xmath96 \cdot [ \boldsymbol{\rho}_j \times \hat{\boldsymbol{\mu } } ] \rangle_\mathrm{eq } = \frac23 \boldsymbol{\rho}_i \cdot \boldsymbol{\rho}_j , \ ] ] we find @xmath97 or , using ( [ eq : rho ] ) , @xmath98 notice that in the latter expression as well as in ( [ eq : en2mu-3 ] ) the summands corresponding to equivalent bonds are equal . in the given approximation ,
expression ( [ eq : en2u5 ] ) is nothing but a constant ( negative ) addition to the magnetic energy .
however , it was found in @xcite that this contribution had view @xmath99 , with @xmath100 being a saturation magnetization .
this term is important for describing of double twisted magnetic structures , e.g. the skyrmion lattices ( a - phases ) and hypothetical cubic textures similar to the blue phases of liquid crystals .
indeed , it is not clear so far why the a - phases are energetically preferable in some ranges of temperatures and magnetic fields . in order to solve this problem , in the phenomenological theory ,
they often add to the energy a correction term in the form @xmath101 . yet
a trouble emerges here , namely , the magnetization in some areas can exceed the saturation value @xmath100 .
the form @xmath99 , resulting from the heisenberg model , resolves this difficulty . because , as is seen from ( [ eq : rho ] ) , @xmath90 are independent of spatial derivatives of @xmath63 , the energy @xmath102 does not contain the derivatives as well .
on the other hand , @xmath103 does not depend on @xmath90 , and this part of the energy can be rewritten in the following form conventional for the macroscopic phenomenological theory @xmath104 , \ ] ] where @xmath4 and @xmath5 are the phenomenological constants expressed now through the parameters of the microscopical heisenberg model : @xmath105 @xmath106 and @xmath63 is a unit vector oriented along the local magnetic moment density @xmath107 .
parameters @xmath4 and @xmath5 determine spiralling of the spin structure . in particular
, the minimization of ( [ eq : en2mu-4 ] ) gives as a result the helix with propagation number @xmath108 and vector @xmath63 rotating in the plane perpendicular to the helix axis .
the absence of canting in expression ( [ eq : en2mu-4 ] ) does not mean that the canting does not affect the spiralling . as a matter of fact
, the contribution of canting is taken into account , when choosing the fictitious atomic coordinates with condition ( [ eq : xideal ] ) .
the simple additive expressions ( [ eq : calj ] ) and ( [ eq : cald ] ) can appear only because the contributing bonds @xmath48 are the functions of the exchange coordinates .
notice that , due to the dependence of vectors @xmath48 on the exchange constants , both @xmath4 and @xmath5 are non - linear functions of @xmath2 . in the case of @xmath109 non - equivalent bonds with parameters
@xmath110 , @xmath111 and @xmath112 independent coordinates of magnetic atoms ( @xmath113 for mnsi and @xmath114 for cu@xmath0oseo@xmath1 ) , the phenomenological constants have the following views : @xmath115 , @xmath116 , where @xmath117 and @xmath118 are some homogeneous polynomials of @xmath110 with degrees @xmath119 and @xmath112 , correspondingly , and @xmath120 is a homogeneous polynomial linear by the components of vectors @xmath111 and being of degree @xmath112 with respect to @xmath110 .
indeed , it follows from equation ( [ eq : xideal ] ) , which is a system of @xmath112 linear equations with respect exchange coordinates , with both coefficients and constant terms being linear functions of @xmath110 .
consequently , the intuitively obvious for dimensional reason expression @xmath121 transforms to @xmath122 .
let us illustrate the theory developed by a simple example of the mnsi - type crystals with the @xmath8 structure .
the cubic crystal mnsi has the space group @xmath7 .
its unit cell contains four magnetic manganese atoms in the position @xmath91 ( at threefold axes ) with @xmath124 . in accordance with @xcite
, we define four magnetic shells by the non - equivalent bonds @xmath125 , @xmath126 , @xmath127 , and @xmath128 , with corresponding exchange constants @xmath129@xmath130 and dm vectors @xmath131@xmath132 .
the first atom @xmath133 has 6 bonds of each kind : @xmath134 , @xmath135 , @xmath136 , @xmath137 ( the symbol @xmath138 means possible cyclic permutations of the coordinates ) .
the corresponding dm vectors can be obtained from @xmath131@xmath132 by the same sing changes and cyclic permutations of the components .
we assign @xmath139 for all the manganese atoms in the equivalent positions .
then , condition ( [ eq : xideal ] ) for the atom @xmath133 can be rewritten as @xmath140 or , using the symmetry 3 of the position , @xmath141 and @xmath142 in accordance with @xcite .
the 4th neighbours do not influence the exchange coordinate , because the bond @xmath143 connect the atoms belonging to the same magnetic sublattice . from ( [ eq : calj ] ) and ( [ eq : cald ] )
the macroscopic parameters can be found , @xmath144 @xmath145 after substitution of the exchange coordinate @xmath146 into the @xmath78 vectors , the macroscopic exchange parameter can be expressed through the exchange constants of the bonds , @xmath147 which coincides with the result from @xcite . the atom @xmath133 , belonging to the 1st magnetic sublattice of the crystal , has two atoms of each other sublattice ( 2 , 3 , 4 ) in the 1st , the 2nd and the 3rd magnetic coordination spheres , and six atoms of the same sublattice ( 1 ) in the 4th coordination sphere .
therefore , condition ( [ eq : rho ] ) on tilt vectors can be written for the atom @xmath133 as @xmath148 with @xmath149 . using symmetry equation @xmath150 we find @xmath151 in accordance with @xcite .
tilt vectors @xmath152 , @xmath153 and @xmath154 of other manganese atoms in the unit cell can be obtained from @xmath155 by the corresponding symmetry transformations . as is seen from ( [ eq : rho-3 ] ) , the 4th magnetic neighbours , belonging to the same sublattice , do not influence tilt vectors .
the cubic crystal cu@xmath0oseo@xmath1 has the space group @xmath7 , the lattice constant @xmath157 .
its unit cell contains 16 magnetic copper atoms : four cu - i in the position @xmath91 with coordinates @xmath158 ( at threefold axes ) and twelve cu - ii in the general position @xmath92 with coordinates @xmath159 @xcite . in the unit cell ,
16 copper atoms form four almost regular tetrahedra , which , in their turn , do a bigger one , resembling the first - order model of the sierpinski fractal tetrahedron ( figure [ fig : unitcell ] ) .
all the bonds shown by the rods in the picture have approximately the same length . the nearest magnetic environment is determined by four types of non - equivalent bonds , but following @xcite we take into account an additional bond with the atoms of the second magnetic environment .
the examples of these five non - equivalent bonds are represented by the arrows in figure [ fig : unitcell ] . in table
[ 5bonds ] , the coordinates and the energetic parameters are listed of five non - equivalent bonds , radiated from the cu - ii atom in position @xmath160 .
copper atoms in the unit cell of cu@xmath0oseo@xmath1 : four cu - i ( green ) in positions @xmath91 and twelve cu - ii ( indigo ) in positions @xmath92 of the space group @xmath7 .
the arrows show five non - equivalent bonds providing main contributions into the isotropic exchange and the dm interactions.,width=264 ] @clll @xmath161 & @xmath162, & @xmath163 , mev & @xmath164 , mev + @xmath165 & 3.054 & @xmath166 & @xmath167 + @xmath168 & 3.049 & @xmath169 & @xmath170 + @xmath171 & 3.226 & @xmath172 & @xmath173 + @xmath174 & 3.304 & @xmath175 & @xmath176 + @xmath177 & 6.349 & @xmath178 & @xmath179 + the magnetic neighbours of atom cu - i are nine copper atoms in positions @xmath92 .
three of them are listed in table [ bondsi ] .
because the atom in position @xmath180 is on the 3-fold axis @xmath181 $ ] , the remaining six bonds can be obtained by cyclic permutations of the coordinates of the vectors .
@cccc @xmath182 & @xmath2 & @xmath3 & @xmath183 + @xmath184 & @xmath185 & @xmath186 & @xmath187 + @xmath188 & @xmath130 & @xmath189 & @xmath190 + @xmath191 & @xmath192 & @xmath193 & @xmath194 + the magnetic environment of atom cu - ii contains seven copper atoms : four in positions @xmath92 and three in positions @xmath91 ( table [ bondsii ] ) .
@cccc @xmath195 & @xmath2 & @xmath3 & @xmath183 + @xmath165 & @xmath129 & @xmath196 & @xmath197 + @xmath198 & @xmath129 & @xmath199 & @xmath200 + @xmath168 & @xmath185 & @xmath201 & @xmath202 + @xmath171 & @xmath203 & @xmath204 & @xmath205 + @xmath206 & @xmath203 & @xmath207 & @xmath208 + @xmath174 & @xmath130 & @xmath209 & @xmath210 + @xmath177 & @xmath192 & @xmath211 & @xmath212 + all 16 copper positions in the unit cell are characterized by `` sense '' numbers @xmath213 and tilt vectors @xmath90 determining canting .
owing to ferrimagnetic order , the spins of the copper atoms in positions @xmath91 are opposite to those in positions @xmath92 and to the summary magnetic moment @xcite , so we can assign the values @xmath214 and @xmath215 .
the tilt vectors of the atoms in positions @xmath180 and @xmath160 are chosen as @xmath216 and @xmath217 , correspondingly .
tilt vectors of other copper atoms can be obtained from these two vectors using symmetry transformations of the point group 23 .
let us use condition ( [ eq : xideal ] ) for the atoms in positions @xmath180 and @xmath160 , in order to find exchange coordinates of the positions .
then , the system of linear equations can be easily obtained and solved : @xmath218 with @xmath219 @xmath220 general solution is rather combersome . using the values of @xmath129@xmath192 from table [ 5bonds ] , we find @xmath221 just these ideal coordinates , rather than the real ones , should be used in calculations of the phenomenological constants @xmath4 , @xmath5 by formulae ( [ eq : calj ] ) , ( [ eq : cald ] ) .
the calculation with the data from table [ 5bonds ] gives @xmath222 mev , @xmath223 mev .
therefore , the helix propagation number @xmath224 ( @xmath225 ) .
the positive value of @xmath226 means that the magnetic helicoid is expected to be right - handed in the crystals with the given set of atomic positions . in the enantiomorphs with opposite values of all atomic positions
the helicoid should be also opposite ( left - handed ) .
this prediction is rather easy for experimental proof ( the corresponding procedure is well developed for mnsi - type crystals @xcite ) . in order to find the tilt vectors , we write system ( [ eq : rho ] ) for the atoms in positions @xmath180 and @xmath160 , @xmath227 @xmath228 using the data from table [ 5bonds ] , we find @xmath229 the canting angles are @xmath230 ( @xmath231 ) , @xmath232 ( @xmath233 )
. the tilt vectors ( [ eq : rho2 ] ) determine spin cantings @xmath234 of the first approximation , both in arbitrary twisted phases , including helicoids and a - phase , and in the state unwound by magnetic field @xmath235 .
for example , figure [ fig : ferro ] shows the canting arrangement in the periodic structure , if the external magnetic field is along the [ 001 ] axis .
in the first approximation , all cantings lie in plane ( 001 ) , perpendicular to the field and magnetization .
it is obvious that the canting arrangement is symmetrical relative to the @xmath236 screw axes directed along the field .
the spin canting in the ferrimagnetic state of cu@xmath0oseo@xmath1 , induced by an external magnetic field applied in the [ 001 ] direction .
the centres of the grey circles coincide with projections of copper atoms in plane ( 001 ) .
the green and cyan arrows show directions and magnitudes of spin cantings of 4 cu - i and 12 cu - ii atoms , correspondingly.,width=264 ] the absolute atomic configuration of cu@xmath0oseo@xmath1 was determined in @xcite .
we find that it is very close to @xmath237 cubic symmetry @xcite , see table [ tabstruc ] where the idealized structure with the @xmath237 symmetry is compared with the observed structure .
thus , similar to the case of the @xmath8 structures @xcite , we can say that this atomic structure is right - handed because the space group @xmath237 contains only right - handed screw axes @xmath238 .
and in this right - handed structure the magnetic helicoids are expected to be right - handed according to above calculations . for another enantiomorph with inverse values of all atomic coordinates ,
the idealized atomic structure is left - handed ( @xmath239 space group ) .
the corresponding inverted real structure ( it has exactly the same energy as the structure determined in @xcite ) can be also considered as left - handed and its magnetic helicoid is expected to be left - handed , but its space group remains @xmath7 .
@lllll & & + & type & @xmath240 & type & @xmath240 + cu - i & @xmath91 & 0.8860,0.8860,0.8860 & @xmath241 & @xmath242 + cu - ii & @xmath92 & 0.1335,0.1211,-0.1281 & @xmath243 & @xmath244 + se - i & @xmath91 & 0.4590,0.4590,0.4590 & & + se - ii & @xmath91 & 0.2113,0.2113,0.2113 & & + o - i & @xmath91 & 0.0105,0.0105,0.0105 & & + o - ii & @xmath91 & 0.7621,0.7621,0.7621 & & + o - iii & @xmath92 & 0.2699,0.4834,0.4706 & @xmath243 & @xmath245 + o - iv & @xmath92 & 0.2710,0.1892,0.0313 & @xmath243 & @xmath246 + another geometrical approach is to reduce the magnetic structure of cu@xmath0oseo@xmath1 to that of the mnsi - type crystals @xcite .
for that , the structure is divided into `` strong '' tetrahedra of copper atoms , with the summary magnetic moments being considered as individual classical spins . indeed , as is seen from table [ 5bonds ] , the bonds 2 and 3 have the maximal absolute values of the exchange interaction constants .
the approximation of strong tetrahedra corresponds to the infinite constants : @xmath247 , @xmath248 .
then , the corresponding bonds with the exchange coordinates become zero , @xmath249 , and the tetrahedra transform to `` atoms '' in positions @xmath91 with coordinate @xmath250 besides , the bonds @xmath251 of cu@xmath0oseo@xmath1 transform into the bonds @xmath251 of mnsi , whereas the bonds @xmath143 and @xmath252 of cu@xmath0oseo@xmath1 become the bonds @xmath253 of mnsi .
the coordinate @xmath146 can be found from the solution of the system ( [ eq : system - coord ] ) , when passing to the limit @xmath254 , @xmath255 .
but , it is easier to use expression ( [ eq : xideal - mnsi ] ) for mnsi with the parameters @xmath256 , @xmath257 , and @xmath258 ( the sign minus is due to bonds @xmath143 and @xmath252 connecting atoms cu - i and cu - ii having opposite spins ) .
therefore , @xmath259 then , from ( [ eq : calj])([eq : k ] ) can be calculated @xmath260 mev , @xmath261 mev , and @xmath262 .
as we see , the approximation of `` strong '' tetrahedra with the data from @xcite gives value of the propagation number @xmath226 , which is two times less than one obtained with true energetic parameters of bonds @xmath263 and @xmath253 .
the calculated value of the propagation number , @xmath264 , is almost three times greater than one found in @xcite without taking the canting into account , and approximately two times bigger than the experimental value @xmath265 @xcite .
let us explain the cause of it .
in order to calculate @xmath226 in the one - helix model ( without canting ) , the formulae ( [ eq : calj])([eq : k ] ) can be used with real coordinates instead of exchange ones . indeed , in this case , using the data from @xcite , we find @xmath266 , which is very close to the value calculated in @xcite without taking the canting into account . in order to understand this essential difference , notice that , accordingly to ( [ eq : cald ] ) , the phenomenological parameter @xmath267 determining spiralling of the magnetic structure is composed of the scalar products @xmath268 .
it is well known from the superexchange theory , that the dm vectors are approximately perpendicular to the bonds ( see also third column in table [ bexch ] ) . were the perpendicularity of the dm vectors to the bonds exact ,
the spiralling would be absent on condition that expression ( [ eq : cald ] ) contained vectors @xmath78 with real coordinates .
however , the expression depends on the exchange ones , and , consequently , @xmath5 is a linear combination of differences @xmath269 with coefficients from the dm vector components .
therefore , the exact numerical values of @xmath146 can strongly affect the spiralling , determining both pitch and sense of the magnetic helix . in table
[ bexch ] , the calculated with the use of ( [ eq : coord ] ) lengths of the `` exchange '' bonds @xmath270 are listed .
as one can see , the lengths @xmath271 of the 2nd and the 3rd bonds are considerably less than the real ones .
it rather justifies the approximation of strong tetrahedra @xcite , in which @xmath272 .
another important factor is the angle between the dm vector and the bond @xmath270 in exchange coordinates . from table
[ bexch ] it is seen that , in accordance with @xcite , the angles @xmath273 are really close to @xmath274 .
but , in exchange coordinates , the differences of @xmath275 from the right angle is more considerable , particularly for the 2nd and the 3rd bonds .
@lccccc & @xmath276 & @xmath273 & @xmath271 & @xmath275 & @xmath277 + @xmath251 & 0.3422 & 94.3@xmath278 & 0.6340 & 99.5@xmath278 & 7.8@xmath278 + @xmath263 & 0.3417 & 85.3@xmath278 & 0.1053 & 71.3@xmath278 & 25.5@xmath278 + @xmath253 & 0.3615 & 99.5@xmath278 & 0.1599 & 110.2@xmath278 & 13.2@xmath278 + @xmath143 & 0.3702 & 86.3@xmath278 & 0.5443 & 83.8@xmath278 & 11.0@xmath278 + @xmath252 & 0.7114 & 85.8@xmath278 & 0.6325 & 86.9@xmath278 & 5.0@xmath278 + the significant difference between experimental and calculated values of propagation number can result from ( i ) an inaccuracy of _ ab initio _ calculations of the bond parameters @xmath2 and @xmath3 , ( ii ) the neglect of other bonds from the second and next magnetic shells , ( iii ) the presence of a non - isotropic interaction different from dzyaloshinskii
moriya one .
it is important here to attend to the calculation accuracy of dm vector projections on the bond directions ( in exchange coordinates ) , because only these projections affect the twist .
if vectors @xmath3 are almost perpendicular to bonds , then their projections on the bond directions are relatively small in size , and even a slight change of their values can have a strong influence on the phenomenological constant @xmath5 and consequently on the propagation number @xmath226 . on the other hand , the isotropic parameters @xmath2 can provide their influence both through the phenomenological constant @xmath4 and the exchange coordinates .
it is noteworthy , that although @xmath2 and @xmath3 decrease with distance , their contributions into the macroscopic parameters @xmath4 and @xmath5 , accordingly to ( [ eq : calj ] ) and ( [ eq : cald ] ) , are proportional to @xmath279 and @xmath280 , correspondingly .
moreover , the number of neighbours also increases with distance .
thus , for example , there are eight non - equivalent bonds in the second magnetic environment ranging from 5.35 to 5.54 . as can be seen from ( [ eq : coord ] ) , the maximal deviation of the exchange coordinate from the real one is for @xmath281 , @xmath282 . in @xcite
it is shown that for itinerant magnetics of the mnsi - type in the frame of rkky model , @xmath146 can have practically any value ( we would remind that the ideal coordinates are nothing but some functions of the exchange parameters @xmath2 ) .
nevertheless , some limitation can result from the condition of validity of the theory . in
order to study this we should first understand the origin of the exchange coordinates . when a transition is performed from discrete spins to continuous magnetic moment
, some frustrations appear from the fact that the spins correspond poorly to the values of smooth function ( e.g. helicoid ) in their physical positions .
it is just the cause why , accordingly to equation ( [ eq : u1 ] ) , two kinds of canting can be distinguished : the first one determined by the dm interaction , and the second one connected with spatial derivatives .
the latter is due to existence of several magnetic helices , with small phase shifts from the average single helix .
the appropriate displacement of the magnetic atoms into the fictitious positions ( which does not influence the magnetic energy ) makes all the individual helices confluent , and the canting due to spatial derivatives disappears ( figure [ fig : exch ] ) .
in fact , the small deviation @xmath269 of the exchange coordinate from the real one for cu@xmath0oseo@xmath1 means that the phase shifts of the individual helices are also small , which correlates with the assumption about small cantings between neighbouring spins .
greater phase shifts would mean that we could not already consider the spins as weak non - collinear , and the simplifications of section [ sec : collinear ] would be impossible .
( a ) in the real spin structure there are several magnetic helices , corresponding to different magnetic atoms in the unit cell , with phase shifts between them .
( b ) when the atoms `` move '' ( arrows ) from the real positions to the fictitious ( exchange ) ones , the phase shifts disappear and the helices become confluent.,width=264 ] another important property is the sense of the magnetic chirality . in the case when a twist arises from a spontaneous break of symmetry , the helix sense can be plus or minus with equal probability and in each case it is casual . in contrast , for the magnetics without centre of inversion , exemplified by mnsi , cu@xmath0oseo@xmath1 and other crystals , the sense of the magnetic chirality correlates with the structural one .
thus , in pure mnsi , the left - handed atomic structure results in the left - handed magnetic helix @xcite . besides , the chiral interlink in mnsi - type helimagnetics is found to be strongly dependent on their atomic composition @xcite . in the present work , using the magnetic data from @xcite , we obtain @xmath283 , i.e. the helicoid is right for the structure described in @xcite and used in simulations of @xcite . recently , this important result has been experimentally proved by dyadkin _
_ @xcite . in the first approximation ,
the cantings , being only small corrections to spins , are perpendicular to the latters . consequently , the cantings of all @xmath284 magnetic atoms in the unit cell are determined by @xmath285 variables .
thus , for cu@xmath0oseo@xmath1 the number of canting components amounts to 32 per unit cell .
surprisingly , that , owing the ( tetrahedral ) symmetry of the crystal , we can make with a considerably lower number of parameters in order to describe the local magnetic structure : four exchange coordinates @xmath286 , @xmath287 , @xmath288 , @xmath289 , and four tilt vector components @xmath290 , @xmath291 , @xmath292 , @xmath293 . for mnsi - type helimagnetics
there are only two parameters : @xmath146 and @xmath294 . in ( [ eq : rho2 ] ) , the values of tilt vectors @xmath295 are calculated with the use of energetic parameters @xmath163 , @xmath164 from @xcite .
in addition to the propagation number @xmath226 , given by ( [ eq : k ] ) , these , characterizing the canting , vectors are the experimentally observed parameters of the magnetic structure . in @xcite
the conditions were proposed of a diffraction experiment to find the canting in the mnsi - type helimagnets . because the magnetic atoms in these crystals are situated at the special positions @xmath91 , the tilt vectors are defined by only one parameter @xmath294 , which is proposed to be found in the measurement of the `` forbidden '' bragg reflection @xmath296 , @xmath297 in the unwound by magnetic field structure . in the general case of cubic helimagnets with almost collinear spins ,
the antiferromagnetic ( canting ) part of the structure factor is determined by the vector @xmath298 where @xmath299 is a reflection vector , and summation is taken over all magnetic atoms in the unit cell .
in particular , for the copper atoms in the general positions @xmath92 and the pure magnetic reflection @xmath296 , @xmath297 , this sum has the view @xmath300 it is obvious that , in order to find all components of tilt vectors @xmath295 for magnetic atoms in all non - equivalent positions , several forbidden reflections with different @xmath301 should be measured .
notice that , as far as the measurable parameters @xmath226 , @xmath295 are determined by greater number of constants @xmath2 , @xmath3 , the problem of finding of the latter from experimental data is unsolvable in general . at the best
, only some combinations of @xmath163 and @xmath164 can be calculated by inversion of equations ( [ eq : k ] ) and ( [ eq : system - rho2 ] ) .
nevertheless , the comparison of theoretical predictions with measurements can confirm ( or refute ) reliability of _ ab initio _ calculations of the energetic parameters .
we are grateful to o. janson , a. a. tsirlin and s. s. doudoukine for useful discussions and encouragement and to j. h. yang , h. j. xiang and x. g. gong for providing us with details of their dft calculations of the exchange parameters and the dm vectors in cu@xmath0oseo@xmath1 .
the reported study was partially supported by rfbr , research project no .
14 - 02 - 00268 a , and by two projects of the presidium of the russian academy of sciences : `` matter at high energy densities ; substance under high static compression '' and `` diffraction of synchrotron radiation in multiferroics and chiral magnetics '' .
99 grigoriev s v , maleyev s v , okorokov a i , chetverikov yu o , bni p , georgii r , lamago d , eckerlebe h and pranzas k 2006 magnetic structure of mnsi under an applied field probed by polarized small - angle neutron scattering _ phys .
_ b * 74 * 214414 adams t , chacon a , wagner m , bauer a , brandl g , pedersen b , berger h , lemmens p and pfleiderer c 2012 long - wavelength helimagnetic order and skyrmion lattice phase in cu@xmath0oseo@xmath1 _ phys .
lett . _ * 108 * 237204 seki s , kim j - h , inosov d s , georgii r , keimer b , ishiwata s and tokura y 2012 formation and rotation of skyrmion crystal in the chiral - lattice insulator cu@xmath0oseo@xmath1 _ phys . rev .
_ b * 85 * 220406(r ) maisuradze a , shengelaya a , berger h , djoki d m and keller h 2012 magnetoelectric coupling in single crystal cu@xmath0oseo@xmath1 studied by a novel electron spin resonance technique _ phys . rev
_ * 108 * 247211 belesi m , rousochatzakis i , abid m , rler u k , berger h and ansermet j - ph 2012 magnetoelectric effects in single crystals of the cubic ferrimagnetic helimagnet cu@xmath0oseo@xmath1 _ phys .
_ b * 85 * 224413 yang j h , li z l , lu x z , whangbo m - h , su - huai wei , gong x g and xiang h j 2012 strong dzyaloshinskii - moriya interaction and origin of ferroelectricity in cu@xmath0oseo@xmath1 _ phys . rev . lett . _ * 109 * 107203 effenberger h and pertlik f 1986 die kristallstrukturen der kupfer(ii)-oxo - selenite cu@xmath0o(seo@xmath1 ) ( kubisch und monoklin ) und cu@xmath305o(seo@xmath1)@xmath1 ( monoklin und triklin ) _ monatsh .
chem . _ * 117 * 88796 ishida m , endoh ya , mitsuda s , ishikawa yo and tanaka m 1985 crystal chirality and helicity of the helical spin density wave in mnsi .
ii . polarized neutron diffraction
_ j. phys .
. japan _ * 54 * 297582 shibata k , yu x z , hara t , morikawa d , kanazawa n , kimoto k , ishiwata s , matsui y and tokura y 2013 towards control of the size and helicity of skyrmions in helimagnetic alloys by spin orbit coupling _ nature nanotechnol . _
* 8 * 7238 janson o , rousochatzakis i , tsirlin a a , rler u k , van den brink j and rosner h 2013 microscopic modeling of the s=1/2 heisenberg ferrimagnet cu@xmath0oseo@xmath1 _ int .
workshop `` dzyaloshinskii moriya interaction and exotic spin structures '' ( veliky novgorod , russia ) _
( gatchina : petersburg nuclear physics institute ) p 67 dyadkin v , grigoriev s v , white j s , prsa k , huang p , rnnow h , magrez a , dewhurst c d and chernyshov d 2014 chirality of structure and magnetism in the magnetoelectric cu@xmath0oseo@xmath1 ( in press ) | twisted structures of chiral cubic ferromagnetics mnsi and cu@xmath0oseo@xmath1 can be described both in the frame of the phenomenological ginzburg landau theory and using the microscopical heisenberg formalism with a chirality brought in by the dzyaloshinskii
moriya ( dm ) interaction .
recent progress in quantum first - principal methods allows to calculate interatomic bond parameters of the heisenberg model , namely , isotropic exchange constants @xmath2 and dm vectors @xmath3 , which can be used for simulations of observed magnetic textures and comparison of their calculated characteristics , such as magnetic helix sense and pitch , with the experimental data . in the present work
, it is found that unaveraged microscopical details of the spin structures ( the local canting ) have a strong impact on the global twist and can notably change the helix propagation number .
coefficients @xmath4 and @xmath5 of the phenomenological theory and helix propagation number @xmath6 are derived from interatomic parameters @xmath2 and @xmath3 of individual bonds for mnsi and cu@xmath0oseo@xmath1 crystals and similar cubic magnetics with almost collinear spins . |
variational principles and symmetries are fundamental tools for the construction of interactions and the study of possible extensions of well - known classical field theories like gravity .
imposing global and local symmetries , together with unitarity and locality , provides us with a guide to explore extensions ( or modifications ) of the action of a given consistent theory .
what about duality symmetries ?
particularly , what about the electric - magnetic duality invariance ?
this symmetry shows up in diverse physical systems , from zero to higher spin theories , and carries many interesting features that make it depart somehow from the general character of global and local symmetries . in the case of linearized gravity and
free higher spin gauge fields in four dimensions , for example , duality invariance can be manifestly implemented at the level of the action and not just in the equations of motion .
such implementation breaks , however , manifest spacetime covariance @xcite .
this feature bookmarks an interesting link between duality and spacetime symmetries , something that acquires a concrete form in the case of an abelian gauge field for which it can be shown that duality invariance implies poincar invariance @xcite . moreover , evidence on the relation between duality and supersymmetry has been provided recently , showing that self - duality may arise in some models with @xmath1 supersymmetry plus a particular hidden @xmath1 supersymmetry ( see @xcite ) . on the other hand , an actual breaking of gauge invariance in maxwell theory
results if one tries to make duality invariance a local symmetry @xcite .
another interesting feature of duality invariant models is that making it manifest at the level of the action requires new variables , `` electric '' and `` magnetic '' potentials , and doubling the number of gauge symmetries ( see @xcite for the case of electromagnetism and gravity with electric and magnetic sources ) . despite these sorts of frictions between duality invariance and other symmetries in field theory ,
the interest in it comes from the belief that understanding an eventual relation between the two of them might enrich our view of hidden symmetries in supergravity and m theory @xcite . in @xmath0 dimensions ,
one has for gravity that the electric - magnetic duality relates the graviton , a rank-2 symmetric tensor , with fields of higher rank .
if we account for duality , then there are two equivalent descriptions of linearized einstein theory : one based on the metric @xmath2 and the other in terms of its dual field @xmath3 , a mixed young tableaux of @xmath4 symmetry type whose covariant action was given by curtright @xcite . in four dimensions
the dual graviton is a rank-2 symmetric tensor , as is the graviton itself . at nonlinear level , however , there are obstructions for the implementation of duality in the einstein - hilbert action .
this is because the curtright action does not admit any non - abelian deformation under the assumptions of spacetime covariance and locality @xcite .
the search for modifications and/or deformations of general relativity ( gr ) is actually a very active trend of current research .
the aim there is to find a gravity theory that bypasses the basic problems associated with the quantization of the classical formulation of gravity as a standard field theory . among the many directions in this trend one can find noncommutative gravity , stringy corrections to the gr action , new massive gravity , double field theory and , recently , hoava - lifshitz gravity @xcite .
the latter is proposed as a uv completion of gr using a `` heretic '' proposal @xcite : breaking diffeomorphism invariance to a subgroup that is still manifest in adm formalism . as a consequence of this breaking , different treatments for the kinetic and potential terms are necessary .
the kinetic term is still quadratic in time derivatives ( but deformed by introducing a new parameter @xmath5 ) , and the potential has terms that can be higher order in space derivatives of the metric .
the hoava - lifshitz ( hl ) action @xmath6 where @xmath7 and @xmath8 denote , respectively , the components of the extrinsic curvature and the ricci tensor of the spatial manifold , is built to be power counting renormalizable . at large distances
, higher derivative terms do not contribute and the theory is supposed to run to gr . the parameter @xmath5 is the parameter which controls the contribution of the trace of the extrinsic curvature and of some of the higher derivative terms . ; for example , see @xcite . ] for @xmath9 the theory has a fixed point in the ir , recovering gr .
the aim of the present note is to implement the electric - magnetic duality in the linearized part of the hl action .
the interesting point is that the breaking of full diffeomorphism invariance for generic values of @xmath5 is not an obstacle for the implementation of duality .
then , this can be viewed as an exercise to test how duality is affected when we put spacetime covariance under stress .
the implementation of duality as a symmetry of the action of linearized gravity was done in @xcite .
the approach of the authors in that work is to solve the hamiltonian and momentum constraints to obtain the prepotentials ( analogous to the magnetic and the electric vector potentials in electromagnetism the latter arising from solving the gauss constraint ) , and then to show the manifest invariance of the action under @xmath10 rotations of those variables .
here we do not follow this method but instead we parallel the work of @xcite where the authors connect fierz - pauli ( fp ) theory with its dual through a parental action ; still , we do make some comments on the former approach in sec . [
sec : diracanalisis ] , where we fix some ideas related to the model at hand .
the parent action approach is defined with a first order lagrangian from which both fp theory for the graviton in the framelike formulation and the corresponding theory for the dual graviton can be extracted ( see fig . [
fig : parentalscheme ] ) .
duality in the framelike formulation has been worked out also in @xcite .
having this scheme in mind , we proceed in sec .
[ sec : dualhl ] in two steps : first we look for an action in the _ vielbein _ basis for the linearized part of the hl gravity in sec .
[ sec : linhlvierbein ] .
this action is enhanced with respect to the latter in the sense that it enjoys diffeomorphism invariance , but still is not local lorentz invariant ; linearized hl gravity arises from it by imposing a gauge fixing condition . then we find a parent action in sec .
[ sec : hlparental ] that delivers this theory and its dual by the same mechanism used in @xcite .
@xmath11}\arrow{se , t}{e_{ab}~\mbox{\small{multiplier}}}\arrow{sw , t}{y^{ab\mid}_{}{c}~\mbox{\small{auxiliary}}}\node { } \\ \node{\boxed{{\mathcal{l}}[e_{ab}]}}\arrow{s , l}{\mbox{\small{gf}}~e_{[ab]}}\node{}\node{\boxed{{\mathcal{l}}[w^{abc\mid}{}_{d}]}}\arrow{s , l}{\mbox{\small{gf}}~w^{abc\mid}{}_{c}}\\ \node{{\mathcal{l}}_{\text{{\scriptsize fp}}}[h_{ab}]}\node{}\node{\mbox{\small{dual of fp:}}~{\mathcal{l}}_{\text{{\scriptsize c}}}[t_{a_1\dots a_{d-3}\mid b } ] } \end{diagram}\ ] ] the output of the procedure described above in the case of @xmath12 is that the dual of linearized hl gravity is the same functional deviation of the fp lagrangian for the dual graviton , but controlled by a different parameter @xmath13 . in sec . [ sec : curtright ] , and before some concluding remarks , we speculate in a sort of hoava - like deviation of the curtright action , the dual of fp theory in @xmath14 .
in this section we review some properties of the model under consideration , its gauge symmetries and degrees of freedom ( d.o.f . ) . particularly we want to comment on the constraint structure of linearized hl gravity , and in its regards to the electric - magnetic duality under the light of the analysis performed in @xcite .
hl gravity comes with a preferred foliation of spacetime arising from an assumed inherent anisotropy scaling between space and time .
in contrast to gr , the gauge symmetries of this theory are given by the _ foliation preserving diffeomorphisms _
, generated by @xmath15 and @xmath16 .
time reparametrizations @xmath16 are time dependent only and it has been shown that they are trivial gauge symmetries @xcite .
we will come back to this feature in the next section . as a consequence of the diffeomorphism invariance breaking , hl gravity contains a problematic extra mode @xcite , which in the linear approximation is manifest only for perturbations around spatially inhomogeneous and time - dependent backgrounds @xcite .
the theory generically exhibits also a dynamical inconsistency : in asymptotically flat spacetimes , the lapse function @xmath17 is generically constrained to vanish , preventing any interesting dynamics except for some particular cases ( like spacetimes with vanishing extrinsic curvature ) @xcite . despite these unappealing features , hl theory provides us with a concrete model for gravity that departs from the relativistic paradigm and is tractable for testing the electric - magnetic duality in it . for the present analysis we assume the theory is describing some solutions that are blind to the above mentioned problems , and we take the linearized part of the minimal deviation from gr of gr .
these are the dimensionless @xmath5 in the kinetic term and @xmath18 with mass dimensions @xmath19=1 $ ] for the potential energy .
we choose here @xmath20 just to deal with the deviation due to @xmath5 .
this action reproduces the eh action of gr in the @xmath21 limit ( see @xcite ) . ]
represented by the action @xmath22 in @xmath12 dimensions and adm separation .
the perturbation around flat spacetime @xmath23 , @xmath24 , and @xmath25 , to quadratic order gives the lagrangian @xmath26\end{gathered}\ ] ] ( @xmath27 and @xmath28 ) , with @xmath29 from now on we will focus our attention on the structural properties of lagrangian ( [ eq : lagadmlin ] ) .
this lagrangian deviates from fp theory in those terms proportional to @xmath5 , to which this theory goes in the @xmath30 limit .
despite that such modifications due to a value of @xmath31 are subtle , the structural consequences are big .
foliation preserving diffeomorphisms act on the field variables as @xmath32 and they are a subset of the full group of diffeomorphisms : @xmath33 . having less gauge symmetry than the fp lagrangian , we would expect linearized hl gravity to have more than 2 d.o.f .. however , this is not seen by the constraint structure of the model .
we briefly summarize its analysis here because it shows a detail that we use in the argument of the next section .
the highlights of the constraint analysis go as follows : we get from the start the set of primary constraints @xmath34 and @xmath35 , the variables @xmath36 being the conjugate momenta of @xmath37 , respectively . the canonical hamiltonian is given by @xmath38 with @xmath39 the variables @xmath40 being the conjugate momenta of @xmath41 .
the trace @xmath42 is proportional to the linearized trace of the extrinsic curvature : @xmath43 for @xmath44 there is one more primary constraint : @xmath45 .
this case has been analyzed by the authors of @xcite , where they show that the same is equivalent to fp theory ( and manifestly in the transverse - traceless gauge ) . in the following
we exclude from our analysis this particular value of @xmath5 .
stabilization of the primary constraints results in the set of secondary constraints , @xmath46 ( the same hamiltonian and momentum constraints as in fp theory ) , and there are more secondary constraints : @xmath47 @xmath48 results from stabilization of @xmath49 , placing a constraint on the trace of the extrinsic curvature , and @xmath50 from stabilization of @xmath51 . at the end ,
stabilization of @xmath52 results in conditions for a lagrange multiplier ( @xmath53 gives no further constraints nor conditions on lagrange multipliers ) . following the authors of @xcite , the constraint equation @xmath54 in
asymptotically flat spacetime determines the lapse function and in fact forces it to vanish , freezing in this way the evolution in time of the theory .
the authors also showed ( see @xcite ) that time reparametrizations are trivial gauge transformations and so there is no first class constraint associated to it .
indeed , by inspection of the poisson brackets between the constraints we see that they vanish except for @xmath55 and as a consequence the hamiltonian constraint , the one we would have expected to generate the evolution in @xmath56 , is second class .
thus we have a set of four second class ( @xmath57 ) and a set of six first class ( @xmath58 ) constraints , and thereby we might conclude that there are 2 d.o.f .. the fact that the extra mode is not manifest here is because we are taking the linearized theory around flat spacetime @xcite . given that @xmath53 are first class , the only are just arbitrary shifts of @xmath59 . ]
nontrivial gauge transformations of the lagrangian ( [ eq : lagadmlin ] ) are those associated to @xmath60 .
now , there is one detail of the constraint structure that is relevant for the electric - magnetic duality on lagrangian ( [ eq : lagadmlin ] ) . following the approach of @xcite for the case of fp
, we can try to analyze duality rotations in terms of the prepotentials @xmath61 and @xmath62 defined by solving the constraints @xmath63 and @xmath64 [ see eqs .
( 2.7 ) and ( 2.11 ) of @xcite ] .
however , in the present case we face an extra constraint placing a condition on @xmath62 , namely , @xmath65 with no counterpart for the @xmath61 variables . besides , there is no exchange between the kinetic and potential energies after these rotations as it happens in fp theory , and the condition above does not help to make the match .
this leaves us with the idea that linearized hl gravity is not self - dual .
nevertheless , we can analyze duality from another point of view that allows us to draw a map to the magnetic dual of linearized hl gravity we call the lagrangian ( [ eq : lagadmlin ] ) the electric one .
the path goes through a parent action , analogous to the one constructed in @xcite for fp theory .
this procedure is reformulated in @xcite in a way that suits the purpose of the present study . in the following section , we use this approach to find the magnetic dual of linearized hl gravity .
schematically , the road map presented in @xcite starts from the second order action in _ vielbein _
formalism for fp , taking it up to a first order parent action and then down from it to the second order dual theory given by the curtright lagrangian ( see fig . [
fig : parentalscheme ] ) . in the present case , and to keep parallelism with fp theory , we need an action in the _ vielbein _ basis as a starting point , the same that should be obtained by a parental one ( see left branch of fig .
[ fig : approach ] ) .
we proceed then in two steps .
first we look for an action corresponding to the lagrangian ( [ eq : lagadmlin ] ) in the _ vielbein _ basis in sec .
[ sec : linhlvierbein ] and next we construct a parent action from which we can obtain the latter and its dual in sec .
[ sec : hlparental ] .
the lagrangian in the _ vielbein _ basis we write is in fact invariant under full diffeomorphisms , but not local lorentz invariant though .
it has more fields and symmetries than linearized hl gravity and delivers this theory through gauge fixing conditions .
finally , on the dual side , we define the theory , also imposing analogous gauge fixing conditions ( see the right branch of fig .
[ fig : approach ] ) .
@xmath66 + + @xmath66 the lagrangian density @xmath67=-\frac{e}{2}\bigl ( \omega^{abc}\omega_{abc}+2\omega^{abc}\omega_{acb}-4\omega^{a}\omega_{a } \bigr)\ ] ] represents einstein - hilbert ( eh ) theory in the _ vielbein _
@xmath68 basis .
the inverses of these fields are written ) whereas greek letters refer to indices in spacetime .
symmetrization and antisymmetrization , denoted by brackets @xmath69 and @xmath70 $ ] , respectively , of any set of indices are defined with strength 1 . ]
@xmath71 , the functions @xmath72}{}^{c}$ ] are the anholonomy coefficients , @xmath73 and @xmath74 .
this lagrangian depends on @xmath75 fields in @xmath0 dimensions , certainly more variables than the independent components of the metric .
however , local lorentz invariance takes care of the excess .
linearization of the above lagrangian corresponds to fp theory , written as @xmath76=-\frac{1}{2}\bigl ( \omega^{abc}\omega_{abc}+2\omega^{abc}\omega_{acb}-4\omega^{a}\omega_{a } \bigr),\ ] ] where @xmath77c}$ ] .
( we apologize for the abuse of notation when using the same for the _ vielbein _ and its perturbation to linear order . )
two things are important to notice here : tangent and world space indices are identified , and the fields @xmath78 do not have definite symmetry . nevertheless , @xmath79 $ ] depends on the symmetric part @xmath80 and on the antisymmetric part @xmath81}$ ] in very different ways .
the above lagrangian splits as @xmath82={\mathcal{l}}_{\text{{\scriptsize fp}}}[h_{ab } ] + { \mathcal{m}}[h , f],\ ] ] where @xmath83=-\frac12\partial^ah^{bc}\partial_ah_{bc}+\partial_ah^{ac}\partial^bh_{bc}\\ -\partial^bh^{a}{}_{a}\partial^c h_{bc}+\frac12\partial_ah^{b}{}_{b}\partial^a h^{c}{}_{c}\end{gathered}\ ] ] is the fp lagrangian for @xmath84 , and @xmath85 is a total derivative , namely , @xmath86 so , while @xmath84 is a dynamical field , are the components in tangent space of the perturbation to the metric around flat spacetime : @xmath87 . ]
@xmath88 enters the action only through a total derivative term , having relevance , if any , only for boundary conditions . because of that , the lagrangian density ( [ eq : fplag ] ) is invariant under the transformations @xmath89 given by general coordinate transformations @xmath90 and arbitrary shifts of the components @xmath88 in terms of parameters @xmath91 . in other words ,
the lagrangian for fp theory in this language is invariant under the field variations @xmath92 using the gauge transformations for the components @xmath88 we can fix them on the boundary to be zero , eliminating them from the theory ( both from the action and boundary data ) .
the lagrangian ( [ eq : fplag ] ) is then a theory only for the symmetric components @xmath84 .
we are looking for a lagrangian density for linearized hl gravity just as a deviation from it .
let us think for the moment in @xmath12 and take the following prescription : @xmath93={\mathcal{l}}_{1}[e_{ab}]+2(1-\lambda)(\omega_0)^2,\ ] ] where the parameter @xmath5 is meant to be the same appearing in the lagrangian ( [ eq : lagadmlin ] ) .
again , the above lagrangian density depends on both the symmetric @xmath84 and antisymmetric @xmath88 parts of the _ vierbein_. one can anticipate that the presence of the last term for @xmath31 will break lorentz invariance somehow and thus prevent allocating @xmath88 inside a total derivative .
these variables will be eliminated by slightly different means to get a theory just for the symmetric components @xmath84 , the variables which linearized hl gravity depends on . to see that ,
let us inspect the symmetries of the lagrangian ( [ eq : lagrhorlin ] ) , where @xmath94,\quad h\equiv\eta^{ij}h_{ij},\ ] ] and which we do not expect to be more than those of eq .
( [ eq : fpgaugesymm ] ) .
here we assumed some choice of reference frame where @xmath95 is the time coordinate . both @xmath96 and @xmath97
depend on @xmath84 and @xmath88 , but in the latter @xmath88 enters only the total derivative @xmath85 given in ( [ eq : m ] ) .
for the following , it is convenient to make use of the adm separation of variables , for which @xmath98 , and we define the notation @xmath99 .
the way in which the lagrangian ( [ eq : lagrhorlin ] ) is written reflects that it does not share all the symmetries of @xmath97 given in ( [ eq : fpgaugesymm ] ) .
let us concentrate on how the @xmath100 term changes under those transformations . despite the manifest fact that it is not a lorentz covariant quantity ,
it is invariant under the global variations @xmath101 with parameters @xmath102 being constant.}+\partial_{[a}\omega_{b]c}$ ] , so for @xmath103 linear in @xmath104 and @xmath102 a constant , we have @xmath105 . ]
the same can be cast as invariant under general coordinate transformations by breaking local lorentz invariance . to see how ,
let us apply the transformations ( [ eq : fpgaugesymm ] ) on it .
that is , @xmath106 in order to have a symmetry , we have to require @xmath107 .
we can take this as an equation for @xmath108 restricting the arbitrariness of this parameter .
the general solution for it is the sum of the solution to the homogeneous equation ( any transverse @xmath109 ) and a particular solution to the inhomogeneous equation : @xmath110 we observe that shifts of @xmath111 with parameters @xmath112 survive the deviation due to @xmath31 as gauge symmetries , and also the gauge transformations of the transverse part of @xmath113 with parameters @xmath114 .
the transformation of the longitudinal part of @xmath113 on the other hand gets linked to the transformation of the longitudinal part of @xmath115 , and hence their shifts are not arbitrary anymore . to sum up , the gauge symmetries of lagrangian ( [ eq : lagrhorlin ] ) are given by [ eq : gaugsymm ] @xmath116 the transformations ( [ eq : gaugsymmh ] ) , which are just @xmath117 projected on adm variables , show that @xmath118 enjoys full diffeomorphism invariance , and the ( [ eq : gaugsymmf ] ) part that the variables @xmath111 and the transverse part of @xmath113 are pure gauge ( but not the longitudinal part of @xmath113 , which is the one appearing in @xmath96 ) .
let us now take a closer look into @xmath119 $ ] .
by direct inspection we see that the set of equations of motion ( eom ) of this lagrangian for @xmath120 is not equivalent to the one arising from linearized hl gravity . for instance , the eom for @xmath121 , @xmath122 has no parallel in the set of eom coming from lagrangian ( [ eq : lagadmlin ] ) .
nevertheless , it is not difficult to see that if we enforce that @xmath123 in the first set , the two systems are equivalent . in the above equation , for example , what happens is that this condition is bringing the constraint on the linearized trace of the extrinsic curvature to a lagrangian equation [ _ cf . _
( [ eq : eome ] ] after this projection with constraint @xmath48 in ( [ eq : k ] ) ) .
the equivalence of these two systems can also be seen directly at the level of the lagrangians . using the shorthand @xmath124
direct computation gives @xmath125.\end{aligned}\ ] ] it is straightforward to see that if we impose @xmath126 in the above expression , we have @xmath127 , and then @xmath128 hence , by imposing the condition ( [ eq : hlgaugefixcond ] ) on @xmath129 $ ] , we get the lagrangian of linearized hl theory modulo the total derivative @xmath130 , in which the longitudinal part of @xmath113 is fixed in terms of @xmath115 .
( that is all we can do using @xmath131 , since this condition says nothing about the transverse part of this vector . )
now we can go further and get rid of @xmath111 and the transverse part of @xmath113 inside @xmath130 by setting them to zero using the gauge symmetries given in ( [ eq : gaugsymmf ] ) .
in conclusion , taking @xmath118 simultaneously with @xmath131 we have a theory for the symmetric components @xmath84 of the _ _
vierbein__s , whose dynamics is given precisely by the eom of @xmath132 and for which we can provide boundary data that confidently allow us to forget all about @xmath85 .
moreover , the condition ( [ eq : hlgaugefixcond ] ) is preserved by the gauge symmetries of linearized hl gravity .
its variation under transformations ( [ eq : gaugsymm ] ) is @xmath133 hence , @xmath131 is preserved by transformations with parameters @xmath134 completely arbitrary and @xmath135 subject to equation @xmath136 .
arbitrary functions of time @xmath137 are solutions of this equation . actually , it is possible to see that @xmath131 is a partial gauge fixing condition that eliminates time reparametrizations as nontrivial gauge transformations . in the appendix , [ app : newconstr ]
we perform the constraint analysis of the lagrangian ( [ eq : lagrhorlin ] ) and see that there are first class constraints generating , in particular , the gauge transformations @xmath138 _ i.e. _ , eq .
( [ eq : gaugsymm ] ) for @xmath139 with @xmath140 arbitrary . the condition ( [ eq : hlgaugefixcond ] )
can be taken as a gauge fixing condition of one of those first class constraints .
now that we have a lagrangian in the _ _ vierbein__s basis for linearized hl gravity ( in the sense explained above and depicted in the left branch of fig .
[ fig : approach ] we proceed next to find the parent action from which @xmath129 $ ] and its dual theory can be obtained .
we follow here the authors of @xcite , where they define a parent action for fp theory , to find the corresponding parent action for the deviation of this theory defined with lagrangian ( [ eq : lagrhorlin ] ) . roughly speaking , we look for a first order theory depending on the _ vielbein _ and some auxiliary fields from which we extract both the second order electric lagrangian @xmath141 $ ] and its magnetic dual .
the former arises from solving the auxiliary fields into the parent action .
the dual theory is obtained by taking @xmath78 in it as lagrange multipliers .
in addition to the procedure to obtain the dual of @xmath141 $ ] , to actually get the dual of linearized hl gravity for the symmetric components of the _ _ vierbein__s in @xmath12 , we have to impose a magnetic analogue of the electric condition ( [ eq : hlgaugefixcond ] ) .
we start by considering the lagrangian ( [ eq : lagrhorlin ] ) , @xmath142={\mathcal{l}}_{1}[e_{ab}]+2(1-\lambda)(\omega_0)^2 , \ ] ] in @xmath0 dimensions .
a parent action for it can be defined with the lagrangian density @xmath143=- 2\bigl ( y^{ab\mid c}\omega_{abc}- y^{ab\mid c}y_{ac\mid b } + \frac{1}{(d-2)}y^{a } y_{a } \bigr)\\ + 2\frac{(\theta -1)}{(d-2)}(y_{0})^{2},\end{gathered}\ ] ] with @xmath77c}$ ] , and @xmath144 a parameter that we expect to be related with @xmath5 ( see below ) . here , besides the _ vielbein _ we have some tensor variables @xmath145 antisymmetric in @xmath146 , but with no definite symmetry in @xmath147 . as before , the notation @xmath148 means the trace @xmath149 .
we see from the eom of the lagrangian ( [ eq : lagrhorpo ] ) that @xmath145 are auxiliary fields , solved in terms of @xmath78 and their derivatives as @xmath150}+2\eta_{c[a}\delta_{b]}{}^{d}\bigl(\omega_{d}-f(\theta)\eta_{d0}\omega^{0}\bigr),\ ] ] with @xmath151 plugging eq .
( [ eq : yvarauxhor ] ) into the first order lagrangian ( [ eq : lagrhorpo ] ) we get the second order electric @xmath141 $ ] if @xmath152 the parameter @xmath153 plays the role of @xmath5 in the dual magnetic theory , a coefficient in front of the deviation from the dual of @xmath79 $ ] . hence , the electric and magnetic theories are deviations from fp and its dual by terms proportional to parameters belonging to complementary regimes , since when @xmath154 and _ vice versa_. before going on to the dual theory , let us check the symmetry transformations of the parent lagrangian
. transformations ( [ eq : gaugsymm ] ) of @xmath141 $ ] can be extended to gauge invariances of @xmath155 $ ] by the corresponding transformations of the auxiliary fields .
in fact , the restriction given in eq .
( [ eq : omegarestr ] ) arises here also from general assumptions . using that @xmath156}+\partial_{[a}\omega_{b]c}$ ]
, we get from eq . ( [ eq : yvarauxhor ] ) @xmath157}-\frac12\partial_{c}\omega_{ab}\\ + \eta_{c[a}\left(\eta_{b]d}-(1-\lambda)\eta_{b]0}\delta^{0}{}_{d}\right)(\partial^{d}\partial^{e}\xi_{e}-\partial^{e}\partial_{e}\xi^{d}+\partial_{e}\omega^{ed}).\end{gathered}\ ] ] the lagrangian is invariant , @xmath158 where @xmath159 means modulo total derivatives .
the first term in the last expression involves the @xmath78 variables , and it is the only one that does . requiring its invariance results in the differential equation that is solved with eq .
( [ eq : omegarestr ] ) , restriction on the parameters that implies @xmath160 , making the last term invariant on its own .
the rest of the terms cancel out on account of the antisymmetry property of @xmath145 in indices @xmath146 .
hence , the gauge symmetries of lagrangian ( [ eq : lagrhorlin ] ) are properly captured by symmetries of the parent lagrangian .
we now compute the dual theory . taking in the lagrangian ( [ eq : lagrhorpo ] ) the @xmath78 as lagrange multipliers , the constraints associated read @xmath161 .
these are solved by potentials @xmath162 completely antisymmetric in indices @xmath163 such that @xmath164 .
reduction of lagrangian ( [ eq : lagrhorpo ] ) gives the dual theory : @xmath165=2 \bigl(y^{ab\mid c}y_{ac\mid b}-\frac{1}{(d-2)}y^{a}y_{a}\bigr)\\+2\frac{(\theta-1)}{(d-2)}(y_{0})^{2},\end{gathered}\ ] ] where @xmath145 are functions of the potentials @xmath166 .
it is not difficult to see that the action defined with this lagrangian is invariant under transformations ( [ eq : gaugtry ] ) for @xmath145 , the restriction ( [ eq : omegarestr ] ) coming here from the last term . )
it can be seen that @xmath167 , meaning that the constraint defining the dual potentials @xmath168 is invariant for those gauge transformations .
] however , these do not exhaust the gauge symmetries of the dual theory .
there are more of these coming from an ambiguity in the potentials solving @xmath161 .
the redundancy is given by the gauge transformations @xmath169 with @xmath170\mid}{}_{d}$ ] arbitrary .
these transformations leave the field strengths @xmath171 invariant and so the lagrangian ( [ eq : hordual1 ] )
. a closer look into the symmetries of the dual magnetic theory and its structure comes from the decomposition @xmath172},\ ] ] consisting of extracting from @xmath173 its trace ( the fields @xmath174 are completely antisymmetric in @xmath175 indices and traceless ) . plugging the above decomposition into lagrangian ( [ eq : hordual1 ] )
, we end up with the dual theory written as @xmath176={\mathcal{l}}_{1}[x , z]+\frac{2(d-2)}{9}(\theta -1)(\partial_{i}z^{i0})^{2}.\ ] ] the @xmath177 $ ] part is the dual of the fp theory in @xmath0 dimensions : @xmath178=2~\partial_{d}x^{abd\mid c}\partial^{e}x_{ace\mid b}+\mathcal{m}[x , z],\ ] ] the last term being the total derivative @xmath179}\bigr)\ ] ] and the only place where @xmath177 $ ] depends on @xmath180 .
we identify lagrangian ( [ eq : hordual ] ) as the dual of ( [ eq : lagrhorlin ] ) .
it is manifest here that @xmath97 is invariant under arbitrary shifts of the traces : @xmath181 .
these include the projection of the gauge symmetries given in eq .
( [ eq : phiymm ] ) for those variables , which split as @xmath182e } \end{aligned } \right . .\end{gathered}\ ] ]
the @xmath183 part leaves @xmath97 invariant modulo a total derivative .
moreover , it is easy to see that @xmath184 , and so it is straightforward that these transformations leave also invariant the last term in lagrangian ( [ eq : hordual ] ) .
the same is not invariant , however , under arbitrary shifts of @xmath180 . instead
, demanding invariance of this term will restrict the arbitrariness of @xmath185 .
let us define @xmath186 and separate it into transverse and longitudinal parts , @xmath187 .
we can do the same with the gauge parameters @xmath188 , the transverse ( or divergenceless ) @xmath189 corresponding to the shift of the component @xmath190 .
invariance of the last term in lagrangian ( [ eq : hordual ] ) requires @xmath191 an equation that puts no restriction on the gauge transformations of @xmath190 , which stand as pure gauge ( as well as the components @xmath192 ) .
the parallel components @xmath193 on the other hand are restricted to be constant in the spatial hypersurface , that is , @xmath194 .
we have here a very similar , if not the same , breaking of local lorentz invariance as in the electric theory . to sum up
, the gauge symmetries for the dual theory are given by [ eq : dualgaugsymm ] @xmath195e } , \label{eq : dgaugsymmh } \\ & \delta_{\tilde{\omega}}z^{ij}=\tilde{\omega}^{ij } , \quad \delta_{\tilde{\omega } } z^{i}=\tilde{\omega}^{i}_\perp+\tilde{\omega}^{i}_\parallel(t),~\text{such that}~\partial_{i}\tilde{\omega}^{i}_\perp=0 .
\label{eq : dgaugsymmf } \end{aligned}\ ] ] we have constructed then the dual of lagrangian ( [ eq : lagrhorlin ] ) in @xmath0 dimensions , and identified the gauge symmetries of the same .
we did it by giving a parent action defined with lagrangian ( [ eq : lagrhorpo ] ) , from which we get both theories . in the case of the electric theory , we obtained linearized hl gravity from lagrangian ( [ eq : lagrhorlin ] ) by imposing the gauge fixing condition ( [ eq : hlgaugefixcond ] ) .
we wonder now if there is an analogous magnetic gauge fixing condition for lagrangian ( [ eq : hordual ] ) in @xmath12 dimensions such that the projected dual theory in reduced space is linearized hl gravity .
let us introduce the field redefinition @xmath196 such that @xmath197}=0,\ ] ] which makes manifest that the dual @xmath198 $ ] is the fp lagrangian for the dual graviton , in @xmath12 [ see eq .
( 2.16 ) in @xcite ] . in these variables ,
the action defined with lagrangian ( [ eq : hordual ] ) for the dual theory is written as @xmath199= \int\mathrm{d}^4x \bigl ( { \mathcal{l}}_{1}[t , z]+\frac{4}{9}(\theta -1)(\partial_{i}z^{i})^{2 } \bigr)\\ = 2 \int\mathrm{d}^4x\bigl(\partial^at^{bc}\partial_at_{bc}-2\partial_at^{ac}\partial^bt_{bc}+2\partial^bt^{a}{}_{a}\partial^c t_{bc}\\ -\partial_at^{b}{}_{b}\partial^a t^{c}{}_{c } + { \mathcal{m}}[t , z ] + \frac{2}{9}(\theta -1)(\partial_{i } z^{i})^{2}\bigr).\end{gathered}\ ] ] in @xmath12 the symmetries ( [ eq : dgaugsymmh ] ) can be written as @xmath200 with @xmath201 arbitrary , which implies @xmath202 . at this stage
we can get rid of @xmath192 and @xmath190 by fixing them to zero using the gauge invariance under arbitrary shifts .
the @xmath203 component is , however , still part of the picture . in order to have a theory for @xmath204 only , we need to impose a condition analogue to ( [ eq : hlgaugefixcond ] ) , with which we could write @xmath203 ( or more precisely , its divergence ) in terms of some components of @xmath204 .
let us separate first the lagrangian defining the action above in space and time components .
that is , @xmath205=-2\bigl ( \dot{t}_{ij}\dot{t}^{ij}-\dot{t}^{2}-4\partial_{i}t_{0j}(\dot{t}^{ij}-\eta^{ij}\dot{t } ) \\-2\partial^{i}t_{0i}\partial^{j}t_{0j}-2\partial_{i}t_{0j}\partial^{i}t^{0j}+2r_{t } \bigr ) \\ + { \mathcal{m}}[t , z]+\frac{4}{9}(\theta -1)(\partial_{i } z^{i})^{2},\end{gathered}\ ] ] where we have defined @xmath206 [ which is nothing but the potential ( [ eq : grpot ] ) as a function of the component @xmath207 of the dual graviton ] and @xmath208 . with the following _ ansatz _ for the magnetic projection condition ,
@xmath209 some arithmetics give @xmath210\mid_{\rho^{*}=0 } ~\simeq~ -2\bigl(\dot{t}_{ij}\dot{t}^{ij}-\textcolor{blue}{\gamma } \dot{t}^2 -4\partial_it_{0j}(\dot{t}^{ij } -\textcolor{blue}{\gamma}\eta^{ij}\dot{t})\\+2(1 - 2\textcolor{blue}{\gamma})\partial^it_{0i}\partial^jt_{0j}-2\partial_it_{0j}\partial^it^{0j } + 2r_{t}\bigr),\end{gathered}\ ] ] with @xmath211 here we were able to get rid of @xmath85 because now this term is a total derivative that can be eliminated from the picture with boundary conditions on @xmath204 ( this is the only field which the above lagrangian depends on ) .
so , we see that it is possible to supplement @xmath212 $ ] with the condition ( [ eq : magneticcond ] ) such that the reduced theory has the same structure as linearized hl gravity .
the suggested _ ansatz _ is motivated of course by recovering the linearized hl gravity lagrangian , and @xmath213 in it , which are the objects involved in the terms proportional to @xmath5 in the lagrangian for linearized hl gravity , it can be seen that those terms should be in a ratio of @xmath214 . ] but there is another justification for this particular form of the gauge fixing condition . in the electric theory
, we mentioned that the gauge fixing ( [ eq : hlgaugefixcond ] ) projects the eom for @xmath121 onto @xmath215 with @xmath216 the trace of the linearized extrinsic curvature .
so , @xmath131 is implementing the constant mean curvature ( cmc ) gauge , in which @xmath217 is a function of time only . in the case of the dual theory ,
the eom for @xmath218 is given by @xmath219 from which we conclude that @xmath220 is a function of time only .
so , the projection @xmath221 given by eq .
( [ eq : magneticcond ] ) is fixing this function to be the trace of the extrinsic curvature for the dual graviton , @xmath222 which is the cmc gauge again .
as in the electric theory , we want to take the dual lagrangian ( [ eq : hordual ] ) together with condition ( [ eq : magneticcond ] ) as the dual theory of linearized hl gravity . then , we would like to preserve such condition by the gauge symmetries of the theory .
applying transformations ( [ eq : dualgaugsymm ] ) to @xmath221 , where @xmath202 , we get @xmath223 this means that gauge symmetries such that @xmath224 preserve the gauge fixing condition , while the same is invariant with @xmath225 completely arbitrary .
these are the gauge symmetries of linearized hl gravity . in conclusion , in @xmath12 the magnetic dual of linearized hl gravity
( [ eq : maghld4 ] ) has the same structure of the former electric theory , but parametrized with @xmath226 .
the electric and magnetic linearized hl deviations come then proportional to parameters with values in ` complementary regimes ' .
for example , we see that @xmath227 .
the case @xmath228 corresponds of course to fp theory on both sides of the duality , and the two parameters are equal once again for @xmath229 ( see fig . [
fig : gl ] ) . for the latter , electric and magnetic linearized hl gravity are given by the same deviation from the fp lagrangian with the same parameter .
the two mentioned values of @xmath5 belong to two branches in parameter space separated by an asymptote in @xmath44 , where duality is not defined . so far so good for @xmath12 .
what happens in @xmath230 dimensions ? in the next section
we analyze the particular case of @xmath14 and explore a hl - like deviation of the curtright action for a mixed symmetry gauge field , the latter being the dual of the graviton in that case .
the field redefinition ( [ eq : fieldred ] ) in @xmath14 is @xmath231 . the dual action defined with lagrangian ( [ eq : hordual ] ) in these variables is written @xmath232=\int\mathrm{d}^5x{\mathcal{l}}_{\lambda}[t , z]=\\ \int\mathrm{d}^5x \bigl ( { \mathcal{l}}_{\text{{\scriptsize c}}}[t]+ { \mathcal{m}}[t , z]+\frac{2}{3}(\theta -1)(\partial_{i } z^{i})^{2 } \bigr),\end{gathered}\ ] ] where @xmath233 = \partial_{a}t_{bc|d}\partial^{a}t^{bc|d}-\partial^{a}t_{bc|a}\partial_{d}t^{bc|d } \\-2\partial_{a}t_{bc|}{}^{c}\partial^{a}t^{bd|}{}_{d } -4\partial^{a}t_{bc|a}\partial^{b}t^{cd|}{}_{d}\\ -2\partial_{a}t^{ac|d}\partial^{b}t_{bc|d } + 2\partial_{a}t^{ac|}{}_{c}\partial_{b}t^{bd|}{}_{d } \end{gathered}\ ] ] is the curtright lagrangian @xcite . when @xmath234 , we know this lagrangian is invariant under arbitrary shifts of the @xmath180 variables , and so the term @xmath235 $ ] can be cast out .
this is not the case for the action ( [ eq : curtrightallahl ] ) : because of the form of the last term , @xmath192 and @xmath236 components are still pure gauge , but not @xmath237 . in the following we assume we have gotten rid of the pure gauge components . to take out of the picture the @xmath238 variables and obtain a reduced theory just for the mixed symmetry @xmath239
, we impose a condition @xmath240=0 $ ] analogous to eq .
( [ eq : magneticcond ] ) .
so , by a hl - like deviation of the curtright action we mean the above lagrangian together with such a condition .
the function @xmath241 should be first order in derivatives ( we want a second order reduced theory when we enforce @xmath242 into @xmath243 $ ] ) and involve @xmath244 traces of the space and time components of @xmath239 . at the same time we expect that demanding preservation of such condition by the gauge transformations of the action ( [ eq : curtrightallahl ] ) will restrict some of those symmetries .
the gauge symmetries of @xmath245 $ ] , besides the shifts ( [ eq : dgaugsymmf ] ) , include the following transformations of the @xmath239 fields : @xmath246c}+\frac{2}{3}(\partial_{[a}a_{b]c}-\partial_{c}a_{ab}).\ ] ] transformations given in eq .
( [ eq : dgaugsymmh ] ) can be recast in this form by realizing that the parameters can be written as @xmath247 and using the separation @xmath248 , with @xmath249 and @xmath250 .
the identity @xmath251}=0 $ ] has been implemented , and hence the third piece of eq .
( [ eq : symmcu ] ) ( see @xcite ) . to build the function @xmath252 $ ] within @xmath253 , we first separate @xmath254 into space and time components , and
then we observe that there are just three terms that can be used to form the traces of the desired form : @xmath255 these are @xmath256 where @xmath257 ( analogous to @xmath258 , the trace of @xmath207 in @xmath12 ) .
finally , we define @xmath259 as the general linear combination @xmath260 the variation of @xmath242 gives @xmath261 here @xmath262 .
this is definitely more complicated than eq .
( [ eq : timerep ] ) for the @xmath12 case , but we immediately see that the choice @xmath263 makes @xmath241 invariant under the gauge transformations with parameters @xmath264 , and poses a differential equation on the parameters @xmath265 . this differential equation can be written @xmath266[j,0 ] } -\frac{\upsilon_{2}}{2}(\partial^{i}\partial^{j}s_{ij}-\delta s)=0\ ] ] ( the `` @xmath267 '' at the end of the first term means partial derivative ) . the combination given in eq .
( [ eq : condmagcu ] ) is not unfamiliar : for @xmath268 , @xmath269 , and @xmath270 , @xmath271 matches with a component of the gauge fixing condition that is imposed on the lagrangian eom coming from @xmath272 $ ] to get the wave equation @xmath273 [ see eq .
( 16 ) in @xcite ] .
the choice @xmath263 is a different gauge , and will deliver a nonrelativistic wave equation .
we will not go further in the analysis of this deviation of the curtright action in the present work .
a proper study of nonrelativistic deviations of this theory deserves its own space .
in this work the powerful tool of parental actions to construct duality relations in classical field theories has been implemented in a nonrelativistic theory of gravity .
we followed the scheme of the authors of @xcite , and hence we proceeded in two steps .
first we constructed a covariant action in the _ vielbein _ basis defined with the lagrangian @xmath129 $ ] , which entails a deviation from fp theory in the framelike formalism [ see lagrangian ( [ eq : lagrhorlin ] ) ] and that in an appropriate gauge reduces to linearized hl gravity .
the gauge fixing ( [ eq : hlgaugefixcond ] ) is such that it breaks the diffeomorphism invariance to a subgroup , precisely the group of foliation preserving diffeomorphisms of hl gravity ( in the linear approximation ) . in a second step we found a parent action defined with lagrangian ( [ eq : lagrhorpo ] ) , from which both @xmath129 $ ] and its dual partner eq .
( [ eq : hordual ] ) can be derived . the further gauge fixing ( [ eq : magneticcond ] ) on the dual branch left us with the magnetic dual of linearized hl gravity .
we observed that the gauge fixing in both sides of the duality was just realizations of the cmc gauge .
we found that the magnetic dual of linearized hl gravity in @xmath12 has the same structure as the electric theory , and moreover that for the value @xmath274 of the characteristic parameter , the theory is self - dual .
this example shows then that self - duality is not tied to lorentz invariance . in the case of @xmath14
, we suggested a nonrelativistic deviation of the curtright action by imposing a gauge fixing condition analogous to the one used in the case of @xmath12 .
breaking lorentz invariance at the linearized level in this way might unlock nonlinear extensions _
la _ hoava of the curtright action .
a different approach for such a nonrelativistic deformation could arise from studying a deviation similar to lagrangian ( [ eq : lagrhorpo ] ) but in the parent action proposed by the authors of @xcite to implement duality in nonlinear einstein theory .
finally , we want to make an observation based on the result reported here for the dual of linearized hl gravity with characteristic parameter @xmath5 . in @xmath12 the magnetic dual theory for the dual graviton was found to have a lagrangian with the same structure as the electric linearized hl gravity for the graviton .
its characteristic parameter @xmath275 given in ( [ eq : gamma ] ) , however , is in general different from @xmath5 . with the theory for the dual graviton having the same ( tensorial ) structure as the theory for the graviton
, we can take the magnetic theory as input in the bottom of the left branch of the diagram depicted in fig .
[ fig : approach ] , _ i.e. _ , as an electric theory , to perform a sort of iteration process .
the output of this iteration will be a theory with the same structure as linearized hl gravity again , but the characteristic parameter will be given by @xmath276 with @xmath277 .
this iteration process is of course only possible in @xmath12 , where the magnetic theory ( the output ) possesses the same tensorial structure as the original electric theory ( the input ) .
it is clear that in @xmath14 , for example , this iteration exercise can not be performed . the same kind of iteration exercise can be thought of in the relativistic @xmath9 case .
there , however , the output not only possesses the same structure but also the same value of the characteristic parameter .
the same happens with the nonrelativistic theory with @xmath274 .
these two values behave then like `` equilibrium points '' of the iteration . hence , in @xmath12 and after any number of iterations when @xmath278 , the output will be always linearized hl gravity with a different parameter each time .
the peculiarity that we want to mention is that the parameters resulting after each iteration come closer and closer to 1 . this value is a sort of ` stable equilibrium point ' , while @xmath274 is an ` unstable equilibrium point ' , because in contrast to the former , the values of the parameters resulting from successive iterations run away from it . in fig .
[ fig : g11 ] a picture of the generic behavior is presented .
the same results no matter what the initial value of @xmath5 is , with the exceptions of the equilibrium points , and of any value that after a finite number of dualizations gives @xmath279 , for which the duality map is ill defined .
we do not know what the content of this curious behavior is , the study of which we leave for future work .
the nonrelativistic model we used for this exploration was a linearized lagrangian in flat spacetime
. it would be interesting to perform the same analysis within ( a)ds or lifshitz backgrounds .
this could be of interest , perhaps , for recent discussions regarding the nonrelativistic holographic correspondence @xcite .
i.c . thanks riccardo argurio , glenn barnich , and gustavo lucena gmez for useful discussions during the elaboration of this work .
i.c . also thanks nicolas boulanger and dmitry ponomarev for discussions and comments on the results .
this work is partially supported by sep - conacyt cb-2008 - 104649 grant ( mexico ) _ aplicaciones de la teora de cuerdas al plasma de quarks y gluones_. the work of i.c . is partially supported by conacyt ( mexico ) through the _ estancias sabticas y posdoctorales al extranjero para la consolidacin de grupos de investigacin _ program , by iisn - belgium ( conventions 4.4511.06 and 4.4514.08 ) , by the _
communaut franaise de belgique _ through the arc program and by the erc through the `` sydugram '' advanced grant .
in this section we show that @xmath131 in eq .
( [ eq : hlgaugefixcond ] ) is a gauge fixing condition for a first class constraint of lagrangian @xmath280 $ ] .
we will just summarize the highlights of the constraint analysis for it .
we assume here we already eliminated @xmath111 and the transverse part of @xmath113 according to what has been described in sec .
[ sec : linhlvierbein ] ( @xmath113 denotes in what follows the longitudinal part of this field ) . since @xmath118 does not depend on time derivatives of @xmath281 , we have the primary constraints @xmath34 , @xmath282 and @xmath283 for the momenta conjugated to those variables .
the canonical hamiltonian is given by @xmath284,\ ] ] where @xmath144 , @xmath285 , and @xmath286 are the same as in sec .
[ sec : diracanalisis ] .
stabilization of the primary constraints delivers a set of secondary constraints given by @xmath287 and @xmath288 in contrast to the dirac analysis carried out for @xmath132 , there are no more secondary constraints coming from stabilization of these , but a condition on lagrange multipliers . from the poisson brackets between all these constraints we get at first sight a very odd result .
there is just one first class constraint , @xmath289 , and the rest ( 13 of them ) are second class .
this is odd because we do not see the first class constraints generating the gauge transformations ( [ eq : gaugsymm ] ) , particularly the diffeomorphisms in the spatial hypersurfaces .
besides , the amount of second class constraints being odd gives a noninteger number for the d.o.f .
we can , however , make a redefinition of the set of constraints that reveals the first class content that generates the gauge transformations of the theory .
this is @xmath290 ( notice that @xmath291 , with @xmath53 the momentum constraint of sec .
[ sec : diracanalisis ] ) , which at the end delivers the following classification : the gauge transformations generated by the set of first class constraints are given by @xmath292 for arbitrary @xmath293 , functions of @xmath294 and @xmath56 .
the gauge transformations ( [ eq : gaugsymm ] ) are reproduced by these for the parameters @xmath295 . as can be checked directly , the counting of d.o.f .
gives 2 .
now we can discern the role of the condition @xmath131 defined in sec .
[ sec : linhlvierbein ] .
this condition is a gauge fixing constraint for the first class @xmath296 , forming together a second class pair : @xmath297 this constraint is accessible : if we have @xmath298 , a gauge transformation @xmath299 and @xmath300 gives @xmath301 there exists @xmath302 such that @xmath303 , and it is @xmath304 as can be easily deduced noticing that @xmath305 is the divergence of the longitudinal component of @xmath306 : @xmath307 and @xmath308 $ ] . in conclusion
, @xmath131 can be part of a full set of gauge fixing constraints . | known as a symmetry of vacuum maxwell equations , the electric - magnetic duality can be lifted actually to a symmetry of an action .
the lagrangian of this action is written in terms of two vector potentials , one electric and one magnetic , and while it is manifestly invariant under duality rotations , it is not manifestly lorentz covariant .
this duality symmetry exists also in linearized gravity in four dimensions , and can be lifted off shell too . in @xmath0 dimensions , the link between linearized gravity and its dual can also be seen from the point of view of a parental action .
this is defined by a first order lagrangian ( with the help of some auxiliary variables ) that delivers both fierz - pauli theory and its dual . in this work
we use this formalism to implement the electric - magnetic duality in the nonrelativistic deviation of fierz - pauli theory arising from hoava - lifshitz gravity .
because this theory breaks diffeomorphism invariance , one finds that such implementation includes some peculiarities .
pacs numbers : : 04.20.fy , 04.50.kd |
the disk - instability model in a thin accretion disk is the most popular model for dwarf - novae ( dne ) outbursts and soft x - ray transients ( sxts ) ( for recent reviews see @xcite or @xcite ) . in this model
the initial fast rise of luminosity in an outburst is due to the propagation of a heating front at the local thermal speed , while the exponential decay follows from the subsequent slow propagation of a cooling wave .
the alternation of heating and cooling in these systems comes from the characteristic ` s ' shaped curve in the midplane temperature versus column density equilibrium curve ( see fig .
[ fig1 ] below ) . at all points along this curve
the heating due to the local dissipation of orbital energy is balanced by radiative losses and convective heat transport .
the unstable middle branch of the equilibrium curve , where column density falls with increasing midplane temperature , is produced by the hydrogen ionization transition and its effect on radiative opacities . the local heating rate is usually taken to follow some variant of the @xmath0 prescription ( @xcite ) .
vertical heat transport in accretion disks is modeled following the usual treatment for radial heat transport in stars , with radiation and convection defining the thermal energy balance . unlike main - sequence stars which have stable convective zones and internal veloicty fields ,
convection in accretion disks varies with time as they evolve along the thermal equilibrium curves .
typically , convection has a very strong influence on the structure of the thermally and viscously unstable middle branch of the thermal equilibrium curve since the optically depth of a partially ionized disk is very large .
convective mixing is not the only source of turbulent transport in stars and disks .
diffusive mixing induced by shear instabilities in rotating stars has been examined by several researchers ( @xcite , @xcite , @xcite ) for its effects on the luminosity , and evolution of stars .
similarly , it is appropriate to include shear - induced turbulent mixing in the study of accretion disk structure , especially since the luminosity of accretion disks is ultimately due to the dissipation of orbital energy through some sort of shear instability .
turbulent mixing is an unavoidable part of this process .
furthermore , a particular local instability has been identified ( velikhov 1959 , chandrasekhar 1960 , balbus and hawley 1991 , hawley and balbus 1992 ) which is uniquely suited to playing this role , and consequently gives us a theoretical basis for evaluating the efficiency of turbulent mixing . in accretion disks
the velikhov - chandrasekhar instability is commonly referred to as the balbus - hawley instability .
here we abbreviate this as the bh instability .
in contrast to convective transport , which only affects the convectively unstable regions and moves heat down temperature gradients , turbulent transport driven by the bh instability is ubiquitous within magnetized accretion disks , and will always act to reduce entropy gradients .
the effect of turbulent diffusion due to some sort of shearing instability , not necessarily the bh instability , has been studied in the context of radial transport in thin ( important for transition fronts , e.g. see @xcite ) and advective ( @xcite ) disks , and in the context of vertical transport in t tauri disks ( @xcite ) . in a real disk
the bh instability should generate turbulence which is only moderately anisotropic ( @xcite ) , an expectation which is consistent with the available numerical simulations ( see gammie 1997 and references contained therein ) .
consequently , we expect that the dimensionless form of all the transport coefficients should differ only by factors of order unity .
more precisely , we note that the numerical simulations consistently suggest that the angular momentum transport is dominated by magnetic tension , and that the relevant magnetic field moment @xmath4 is a few times @xmath5 .
the vertical diffusion coefficient is @xmath6 , where @xmath7 is the velocity correlation time scale , and @xmath8 is a factor of a few , where @xmath9 is the local orbital frequency . we conclude that the vertical diffusion coefficient is comparable to the radial diffusion coefficient . in other words , @xmath10 where @xmath11 is the usual notation for the turbulent kinematic viscosity . here
we will assume that equation ( [ eq : vert ] ) is an exact equality , although in reality we expect that this is only a rough approximation .
consequently , if we assume that the turbulent prandtl number is unity , the vertical thermal diffusion coefficient @xmath12 can be written as @xmath13 where @xmath9 is the keplerian angular velocity , and @xmath14 and @xmath15 are the disk pressure and mass density .
the turbulent mixing rate across the disk is @xmath16 where we have assumed a vertical wavelength @xmath17 comparable to the vertical global pressure scale height @xmath18 . since this is also the thermal equilibration rate for the disk , this implies that turbulent mixing can have a significant effect on disk structure .
the bh instability follows from the presence of a magnetic field in the disk , which requires some sort of magnetic dynamo .
the value of @xmath0 is a fraction of @xmath19 , the ratio of magnetic pressure to ambient pressure .
this ratio is determined by the balance between turbulent dissipation and flux generation via the ( uncertain ) dynamo process .
it has long been known that the observed luminosities and timescales for dne require a larger hot state viscosity , @xmath20 , than in quiescence , @xmath21 ( @xcite , @xcite ) .
this could indicate a dependence of @xmath0 on temperature , or more probably on the disk geometry . on the other hand
, it could simply reflect the fact that cold disks have a large neutral fraction and a high resistivity .
the latter consideration has led gammie and menou ( 1998 ) to propose that disks in quiescence are cold enough to shut down all dynamo activity .
evidence in favor of the former idea comes from the observed exponential decay of soft x - ray luminosity for soft x - ray transients which seems to require a scaling law for @xmath0 of the form ( @xcite , @xcite,@xcite ) @xmath22 where @xmath23 , @xmath24 is the isothermal sound speed at the mid - plane , and @xmath25 is the radial position of the annulus .
the value of @xmath26 is proportional to the assumed masses of the central black holes used to calibrate this relation .
cannizzo et al .
assumed a black hole mass of @xmath27 .
this scaling law has been criticized ( see e.g. @xcite and references therein ) on the grounds that irradiation should substantially modify the structure and stability of disks around black holes . certainly a naive calculation of the optical emission from the disks of sxts during the decline from outburst fails to match the observed behavior ( @xcite ) .
however , a recent study ( @xcite ) has shown that it is much more difficult for self - consistent models of irradiation to change the basic disk structure ( and by extension the x - ray light curve ) than to alter the optical emission from the disk .
equation ( [ eq : ascale ] ) is not compatible with the original suggestion by balbus and hawley ( 1991 ) that the dynamo process should saturate in equipartition between the magnetic field and the gas pressure .
this suggestion was motivated by the notion that the dynamo process arises from the random stretching of field lines in a turbulent medium ( @xcite , @xcite ) . on the other hand ,
this kind of dynamo is rigorously justifiable only in the limit where magnetic forces are negligible , which is obviously not the case here .
equation ( [ eq : ascale ] ) may be compatible with the notion that the instability gives rise to an incoherent dynamo effect ( @xcite ) , although this predicts a somewhat steeper dependence on geometry as @xmath28 .
however , the saturation of the instability as the magnetic field approaches equipartition may explain the discrepancy ( @xcite ) , and is , in any case , necessary to explain the low sensitivity to geometry seen in the numerical simulations ( whose parameters are generally equivalent to @xmath29 of order unity ) .
we stress that since the dependence is fundamentally one of geometry rather than mach number , the current generation of numerical simulations are compatible with either model . here
we will explore the consequences of equation ( [ eq : ascale ] ) for vertical structure , although we will also discuss the alternative , that @xmath0 is constant down to a limiting value of the magnetic reynolds number .
another major issue in vertical structure calculations is the role of convection .
although convection is unlikely to account for any significant level of radial angular momentum transport ( @xmath30 , where @xmath31 and @xmath32 are the fluctuating velocity fields ) in keplerian disks ( @xcite ) , it may still play an important role in vertical energy transport ( @xmath33 , where @xmath34 is the fluctuating temperature field ) .
the use of standard mixing length theory ( mlt ) is even more uncertain in disks than in stars , and previous structure calculations have included simple nonlinear closure schemes ( @xcite , @xcite , @xcite ) , detailed linear analyses of the convective modes ( @xcite ) , and a reduced effective mixing length ( @xcite ) . some of these calculations included the assumption that angular momentum transport is mediated by convective motions , so that the results can not be extended to ionized disks .
there is an additional problem with the use of nonlinear closure schemes since such schemes typically fail to reproduce experimental and numerical results in extremely strong shearing environments , although there has been some progress in the intervening years ( cf .
speziale et al .
1991 ; speziale and gatski 1997 ) .
a particular problem with such schemes is that they tend to underestimate the degree of anisotropy for low rossby numbers . in any case ,
none of the previous work included the effect of diffusive damping and mixing due to turbulent motions acting across the reduced radial length scales .
how can we determine the ` typical mode ' of convection ( i.e. in the spirit of mlt ) when differential rotation and bh turbulence are simultaneously present ? to account for nonlinear effects within the context of linear theory , we start with the linear perturbation theory of convection for keplerian disks and find the modes most resistant to the dissipative effects of shear and turbulent mixing .
we use the properties of these linear modes as a guide to the nonlinear state and the typical velocity and eddy size of convective cells .
we defer discussion of the empirical basis for this approach to the last section of this paper .
the purpose of the paper is to calculate the vertical structure and thermal equilibrium s - curves for a variety of astrophysical disks , including the effects of turbulent energy transport throughout the whole thickness of the disk , reasonable modifications of mlt for convection in disks , and the scaling formula for @xmath0 given in equation ( [ eq : ascale ] ) .
standard mlt and the associated vertical energy transport normally used in keplerian disk models are reviewed in 2 .
section 3 describes the turbulent energy flux driven by the bh instability .
this energy flux is particularly important if convection is suppressed .
we examine the conditions for the suppression of convection in 4 . in 5 , we propose a new model for vertical energy transport in disks .
we present and discuss our results for dne , sxts , and agns in 6 .
finally , we summarize our results and present our conclusions in 7 .
the basic equations for the vertical structure of a geometrically thin , keplerian , optically thick disk are derived from mass conservation , radiative and convective energy transport , and thermal and hydrostatic equilibrium . these give four differential equations for the column density @xmath35 , the total vertical energy flux @xmath36 , the total pressure @xmath14 , and the temperature @xmath37 as a function of distance from the disk midplane : ( @xcite , @xcite ) @xmath38 where @xmath39 is the vertical coordinate perpendicular to the disk plane and is zero at the mid - plane
, @xmath40 is the mean molecular weight , @xmath41 is the rosseland opacity , @xmath42 is the radiative flux , @xmath43 and @xmath44 are defined as @xmath45 associated with the radiative flux @xmath42 and the total flux @xmath36 respectively ( @xcite ) : @xmath46 the gas pressure @xmath47 , is the difference of the total pressure and radiation pressure ; i.e. , @xmath48 we note that @xmath44 , the gradient @xmath49 when the total energy transport is carried by radiation , does not necessarily correspond to any real gradient .
it is a mathematical convenience used to calculate the total flux @xmath36 . in our model the turbulent energy flux is always nonzero , although sometimes negative , and the total flux is never carried by radiation alone .
we will address the nature of the turbulent energy transport later .
these four differential equations have four boundary conditions .
the definition of the column density @xmath50 gives @xmath51 .
the reflection symmetry about the mid - plane gives @xmath52 .
the other two boundary conditions follow from the definition of the photosphere , @xmath53 and @xmath54 , where @xmath55 is the half - thickness ( from the mid - plane to the bottom of the photosphere ) of the disk .
we need one more differential equation for the variable @xmath0 .
since we lack a model for the vertical distribution of dissipation , we will assume that the scaling law for @xmath0 depends only on the mid - plane sound speed .
this gives us a trivial differential equation @xmath56 with the boundary condition @xmath57 where @xmath58
there are good reasons to think that equation ( [ df5 ] ) is not correct ( see , for example , figure 10 in @xcite ) , and will have to be modified once we understand the @xmath39 dependence of the dynamo process .
now we turn to the cooling processes which play important roles in disks in the relevant temperature range , i.e. thousands to tens of thousands of degrees k. the usual approximations for vertical energy transport in a keplerian , optically thick accretion disk include the rosseland approximation for the radiative energy flux @xmath42 and mlt for the convective energy flux @xmath59 .
hence , the total vertical flux in the convectively unstable region is given by @xmath60 where @xmath61 is the vertical speed of the convective bubble in the mixing - length theory , @xmath62 is the mixing length for convection , @xmath63 is the specific heat at constant pressure and @xmath64 is @xmath65 estimated inside the convective bubble . notice that using @xmath64 instead of the adiabatic gradient @xmath66 implies heat exchange between a convective bubble and its environment over the course of its lifetime .
this tends to erase entropy inhomogeneities and reduce the convective flux . in general , this process is described by the equation @xmath67 where @xmath68 , the convective efficiency , is given by @xmath69 here we have assumed isotropic convection so that the volume to surface ratio is @xmath70 . the convective speed in this expression is @xmath71 , the square of brunt - visl frequency with radiative damping @xmath72 , and @xmath73 is the ratio of the excess heat generated within the convective bubbles to the excess energy radiated by the convective bubbles during their lifetime , and @xmath74 is the coefficient of expansion . here
the energy exchange could be the radiative losses given by @xmath68 , or it could include any other diffusive effect given by @xmath73 .
we shall see that this is one of the crucial points in this paper .
equation ( [ frfc ] ) can be rewritten in terms of @xmath44 and the convective efficiency for convective bubbles @xmath68 .
@xmath75 where @xmath76 . equations ( [ gamma ] ) and ( [ conv ] ) were originally used for stellar models and their use for disks is somewhat questionable .
the important feature of traditional mlt is implicit in our expression for @xmath68 which describes nearly isotropic convection without any drag on the motion of convective bubbles .
we shall see below that equation ( [ gamma ] ) should be modified to take into account the unique properties of accretion disks : shear as a result of differential rotation and turbulent mixing , and turbulent drag , due to the bh instability .
as discussed above , it seems reasonable to assume that viscous heating and angular momentum transport in hot accretion disks are driven by the bh instability . just as convective mixing leads to vertical energy transport through mixing , turbulent eddies driven by bh instability
will also produce vertical energy transport at rate proportional to the local entropy gradient .
hereafter the term `` turbulent flux '' will be used to denote the energy flux induced by the bh instability . as noted before , in principle the turbulent thermal conductivity is not necessarily equal to the turbulent viscosity ( @xcite ) .
however , given the approximately isotropic nature of the bh instability , and for the reasons given in the discussion of equations ( [ eq : vert ] ) and ( [ conduct ] ) we will take @xmath77 . by analogy with the formula for the convective flux , we can obtain the turbulent flux by replacing the convective diffusion coefficient @xmath78 with @xmath11 .
in other words , the turbulent flux driven by bh stability is given by @xmath79 where @xmath80 is ( @xmath45 ) estimated inside a turbulent bubble driven by the bh instability .
introducing @xmath80 means that we also assume that there is an exchange of radiative energy between turbulent eddies driven by bh instability and their surroundings .
we can define a turbulent efficiency @xmath81 for turbulent eddies analogous to @xmath68 for isotropic convection .
in other words , replacing the term ( @xmath78 ) with @xmath11 in the expression for @xmath68 , we have @xmath82 and its relation with the entropy inhomogeneities associated with turbulent cells is analogous to that for convective cells : @xmath83 where @xmath84 is the ratio of the excess heat generated within the turbulent bubbles to the excess heat radiated from the bubbles during their lifetime . a similar set of equations for the radiative damping of the turbulent energy flux due to shear instabilities in a differentially rotating star has been studied by maeder ( 1995 ) . in physical terms , the radiative efficiency is the ratio of the radiative cooling time scale to the eddy turnover time .
the cooling time scale due to radiation in the optically thick limit is the square of the radiative diffusion length associated with the turbulent bubbles divided by the radiative conductivity @xmath85 .
there are several reasons to believe that turbulent energy transport could be very important .
first , convection will vanish when entropy per mass increases away from the midplane , while the turbulent flux driven by bh instability is always present .
moreover , the turbulent flux can carry energy toward hot mid - plane when a disk is convectively stable .
second , the turbulent flux inferred by equation ( [ turbflux ] ) has its maximal value when @xmath64 is equal to @xmath66 . in this particular case of weak radiative losses , equation ( [ turbflux ] ) is roughly equal to @xmath86 where @xmath87 is the entropy scale height .
the ratio of this flux to the radiative flux ( @xmath88 ) is @xmath89 , which will be of order unity unless a disk is nearly adiabatic ( @xcite ) .
in other words , the turbulent energy flux can be as large as the total energy flux when radiative losses are small .
finally , and most importantly , convection can suffer from strong shearing in a keplerian disk and from strong mixing in a turbulent disk .
we expect that turbulent mixing will be important whenever convective mixing is suppressed in spite of a unstable entropy gradient .
we address this point in the next section .
modeling convection is often considered one of the main sources of uncertainty in stellar models . despite this mlt
has provided a reasonable framework for achieving an approximate understanding of stellar interiors and has been broadly applied to accretion disks .
however , the question of its applicability to accretion disks has to be raised since strong shearing and mixing can be expected to have a dramatic effect on the disk environment , and to vitiate the usual assumption of nearly isotropic convection .
shearing will shorten the radial width of convective cells whenever convection is weak ( i.e. the brunt - visl frequency @xmath90 ) .
turbulent mixing may prevent the formation of the kind of organized convective cells assumed in mlt . here
we will show how to modify mlt to take these important effects into account .
we begin by considering the linear perturbation equations for nearly incompressible ( boussinesq ) , axisymmetric convection in an accretion disk some local viscosity .
they are @xmath91 where @xmath92 is the frequency measured by a local observer rotating with the disk , @xmath93 is the local fractional density perturbation , @xmath94 is the pressure perturbation divided by the density , @xmath31 , @xmath95 , and @xmath96 are velocity perturbations , @xmath97 and @xmath98 are the vertical and radial wavenumbers , and the vertical gravitational field is given by @xmath99 , where the second term is from the self - gravity of the disk .
self - gravity is not negligible for agn disks for small @xmath0 ( @xcite , @xcite , @xcite ) which , as we shall see later , can happen when we use equation ( [ eq : ascale ] ) .
this system of equations implicitly assumes that the vertical wavenumber is small compared to the pressure scale height , and other vertical scale lengths .
of course , this is wrong , but it makes only a small difference if we use these equations and identify @xmath97 with the inverse pressure scale height at the end .
a more serious issue is that we have simply ignored the inevitable loss of axisymmetry in the convective turbulence .
again , this is not as bad as it first appears .
the existence of strong shear will tend to suppress high angular wavenumber modes .
physically this follows from the fact that modes with short angular wavelengths will respond to strong shearing by deforming into a spiral wave with a short radial wavelength .
this will have an overwhelming effect on mode structure when the shearing time is shorter than the growth time of an unstable mode .
complete suppression of the instability follows when the radial wavelength becomes longer than the characteristic scale for the variation of the comoving frequency . in other words , we require @xmath100 where @xmath101 is the convective growth rate measured by a local observer rotating with the disk , and @xmath102 is the azimuthal wavenumber .
when @xmath102 satisfies this constraint , the effects of shearing will be modest .
in general , this will make @xmath102 no larger than @xmath97 .
when we calculate the volume to surface ratio for a typical convective cell shape below we will assume that @xmath103 . finally , we have left the effects of turbulent damping out of the entropy equation ( equation ( [ per5 ] ) .
we do this to avoid double counting , since we add this effect to the usual effect of radiative damping when calculating the efficiency of convection .
we can get a rough understanding of how the local shear modifies the shape of convective cells by considering the case @xmath104 . then equations ( [ per2 ] ) through ( [ per6 ] ) imply @xmath105 where @xmath106 , @xmath107 , and @xmath108 is the convective growth rate . if convection is weak in the sense that @xmath109 , @xmath110 and the radial length of convection is shortened by the strong shear .
in fact , we have @xmath111 if the dominant mode has a growth rate which is some fraction of @xmath112 .
the radial thinning of convective cells under the influence of the local shear has important consequences for the suppression of convection .
first , this radial distortion makes it much easier for turbulent diffusion , due to the bh instability , to damp out convective cells
. the damping rate due to radial mixing becomes @xmath113 which is stronger than the isotropic result , @xmath114 ( assuming that @xmath115 ) . using our full perturbation equations ( [ per2 ] ) - ( [ per6 ] ) we obtain the dispersion relation @xmath116 we see that a sharp increase in @xmath117 due to the effects of shear is associated with much smaller values of @xmath101 .
the second effect associated with the eddy distortion is that the average radiative diffusion length is now reduced .
this means that the expression of radiative efficiency equation ( [ isogamma ] ) needs to be modified for radially deformed convective cells .
the definition of the radiative efficiency gives ( @xcite ) @xmath118 where @xmath119 is the distance between the center of the convective element , where the temperature perturbation is a maximum , and its surface .
the ratio @xmath120 is the volume to surface ratio for a convective bubble . under normal circumstances
@xmath119 is nearly proportional to @xmath120 and is @xmath121 for the isotropic case .
we can make a plausible estimate for this length scale by comparison with @xmath70 in the isotropic case .
we use @xmath122 for an anisotropic parallelepiped and assume for simplicity that the convective cells have a parallelepiped shape .
then the expression for radiative efficiency becomes @xmath123 where we have defined the anisotropy factor @xmath124 ^ 2={9a\over(1 + 2a^{1/2})^2}.\ ] ] this has a maximum value of one when @xmath125 , that is , in the isotropic limit .
note that this reduction in convective efficiency has nothing to do with turbulent mixing , and is purely the result of shearing .
equation ( [ anigamma ] ) shows that the radiative efficiency is reduced when the eddies are slow , @xmath126 , due to the turbulent drag and radially thin , that is @xmath127 .
we have already incorporated turbulent drag into the dispersion relation ( cf . equation ( [ dispersion ] ) ) .
radiative damping can be added by replacing the brunt - visl frequency for the adiabatic case with one that uses @xmath64 , as discussed in section 2 .
the dimensionless gradients can be related to @xmath68 ( equation ( [ anigamma ] ) ) through equation ( [ gamma ] ) .
before we do this , we need to include one last damping effect , the role of turbulent mixing in smoothing entropy gradients between convective bubbles .
this is the topic of the next section .
the diffusion , and eventual loss , of the entropy inhomogeneities associated with convective bubbles due to turbulent mixing is similar to the effects of radiative loss ( @xcite , @xcite ) and can therefore be treated in an analogous fashion .
we do this by extending equation ( [ gamma ] ) which describes efficiency of convection in the face of radiation transfer between entropy inhomogeneities .
the diffusive term @xmath73 in equation ( [ gamma ] ) can have either sign , depending on the physical context .
for example , convection in extremely hot stellar interiors can involve a local competition between local thermonuclear energy generation and neutrino losses ( @xcite ) . in a similar way
, entropy transfer through turbulent mixing in accretion discs can be modeled by defining @xmath73 as the ratio of the turbulent to radiative heat transfer between convective bubbles .
the turbulent mixing loss under pressure equilibrium is equal to the turbulent conduction @xmath128 , where @xmath129 is the temperature gradient within the convective bubbles .
the radiative loss is @xmath130 in the optically thick limit .
the ratio between these two , using definition of @xmath81 given in equation ( [ dpdef ] ) , gives @xmath131 which gives a decrease in convective efficiency .
in summary , shearing and turbulence act together to reduce convective heat transport .
shear alone shortens the radiative diffusion length , turbulent mixing alone smooths away the entropy inhomogeneities associated with convective bubbles , and the interplay of these two gives a turbulent drag on convective eddies . we can account for these three factors by modifying equations ( [ gamma ] ) and ( [ mlt1 ] ) using equation ( [ anigamma ] ) to obtain @xmath132 when these corrections lead to a suppression of convection , the turbulent heat flux induced by bh instability can have a significant effect on the vertical structure of accretion disks . as we shall see below , weak convection , where the convective growth rate is less than @xmath9 ,
is typically suppressed in a wide variety of accretion disks .
however , before we can show this we need to determine ` typical ' convective eddy sizes and speeds allowing for shearing and turbulent mixing .
we do so in the next section .
a full treatment of vertical energy transport needs to include radiation , convection , and turbulent mixing driven by bh instability . as discussed above
, mlt needs to be corrected by incorporating the effects of shear and turbulent mixing . the starting point for our corrections
is the linear dispersion relation equation ( [ dispersion ] ) for convection .
however , it would be unreasonable to expect the use of purely linear perturbation equations to give us an accurate model of the fully nonlinear end state .
the linear dispersion relation gives a family of solutions , any one of which is an equally valid starting point , but whose properties can vary tremendously .
this means that we can give physically reasonable estimates for the properties of the nonlinear regime from the linear theory only if we are able to find a ` typical ' linear mode whose properties are the best guide to the nonlinear state ; i.e. , the mode which is most resistant to viscous damping and secondary instabilities .
we are not going to propose a model for the turbulent spectrum for convection . instead , our objective is to produce estimates for the typical vertical velocity @xmath96 and the typical radial extent of a convective cell ( i.e. still one - mode model ) , that reduce to the usual mixing length estimates when convection is strong and small scale turbulence is weak , but which incorporate physics neglected by mlt .
the growth of the convective instability is ultimately limited by dissipative turbulence generated from secondary instabilities and from the interactions between convective modes .
we can quantify this by considering the ratio of the mode shear to the growth rate squared , or @xmath133 where @xmath134 and @xmath101 is the mode growth rate .
when this is large the mode will quickly succumb to micro instabilities , so by minimizing @xmath135 we can choose the linear mode most relevant to the nonlinear limit .
this criterion is not , by itself , adequate because we are also dealing with the effects of turbulent mixing on small scales , which will introduce an effective viscosity into the system .
this suggests that we also need to minimize the ratio of viscous damping to the mode growth rate .
since this is an independent effect , we will add the two in quadrature by defining a second ratio @xmath136 and minimize @xmath137 the relative weighting of these two terms is not obvious , but an equal weighting seems the most reasonable choice .
note that when the system is heavily damped , @xmath138 will be large for all modes , and minimizing the sum will reduce to a minimization of @xmath138 .
this guarantees that this criterion will always pick the modes that can still grow as we approach the limit of marginal stability . for an axisymmetric , incompressible convective roll , we have @xmath139 we can combine this with the expression for @xmath32 from equation ( [ per3 ] ) , replacing @xmath140 with @xmath101 , to obtain @xmath141 the last factor of this equation should be approximately independent of mode shape ( @xmath142 ) .
the dispersion relation equation ( [ dispersion ] ) in the inviscid , non - shearing , and isotropic limit gives @xmath143 . since we need to reproduce mlt in this limit we have @xmath144 in mlt we take @xmath145 in mlt , that is , the perturbation wavelength is half the size of an isotropic eddy .
consequently , the last factor in equation ( [ p_s ] ) is equal to one . in the absence of rotation and viscosity
, minimizing @xmath135 implies @xmath146 or @xmath147 . in this limit @xmath135
does not depend on the viscosity , and the influence of viscosity on mode shape is felt entirely through @xmath138 .
we note that our treatment assumes that the vertical scale length @xmath148 remains the same as in standard mlt , @xmath121 .
this follows from our use of the minimization principle .
both @xmath135 and @xmath138 involve wavenumbers through the ratio of @xmath98 to @xmath97 , which does not affect the value of @xmath97 , and through the strength of viscous effects , which always favor a small @xmath97 .
therefore we assume that @xmath148 is as small as possible , about one pressure scale height . with this in mind , we need to minimize @xmath149 as a function of @xmath142 and @xmath101 , with @xmath150 this minimization is subject to constraints from the dispersion relation equation ( [ dispersion ] ) and to the total energy transport equations involving radiation , convection , and turbulent mixing driven by bh instability . the purpose of the paper is to solve the differential equations ( [ df1 ] ) , ( [ df2 ] ) , ( [ df3 ] ) , ( [ df4 ] ) , and ( [ df5 ] ) with the corresponding boundary conditions to investigate the vertical structures and , furthermore , to see how the disk structure changes when one uses the scaling law for @xmath0 .
to solve these equations , we need to express the variable @xmath43 ( or say @xmath42 ) in the equation ( [ df4 ] ) in terms of the thermodynamical variables we want to solve . since the formulae for @xmath43 depend upon the vertical energy transport mechanisms , the thermal gradient @xmath43 has different forms within and outside convection zones .
furthermore , the usual criterion for convective instability by comparing @xmath44 with @xmath66 needs to be modified since @xmath44 is never physically relevant , due to the constant presence of turbulent flux .
therefore , we adopt a convective criterion involving the turbulent flux , where the thermal gradient @xmath44 is replaced with @xmath151 which is @xmath152 of the medium when convection is absent ( @xcite ) . using this new criterion , we are ready to solve for @xmath43 inside and outside the convection zones .
we begin with the case @xmath153 .
then all three energy transport mechanisms coexist within the convectively unstable region .
hence the total vertical energy flux , @xmath36 , is @xmath154 combining this equation with the definition of @xmath44 gives @xmath155 with the definition of @xmath68 in equation ( [ anigamma ] ) and the definition of @xmath96 is equation ( [ v ] ) .
moreover , the convective efficiency @xmath68 and the turbulent efficiency @xmath81 can be related to the @xmath43s by using equations ( [ gamma_nabla ] ) and ( [ gamma ] ) .
since the heat generation by viscosity is proportional to @xmath156 , which is a constant inside and outside the convective bubbles , @xmath84 should be zero and @xmath157 as discussed above .
combining equations ( [ gamma ] ) , ( [ anigamma ] ) , ( [ gamma_nabla ] ) , ( [ v ] ) and ( [ ftot ] ) we can find the thermal gradients in terms of @xmath68 , @xmath81 , @xmath158 and therefore in terms of @xmath101 and @xmath142 .
we obtain @xmath159 where @xmath160 note that @xmath161 has lost its original meaning referring to convective efficiency ( @xcite ) and used here purely for convenience .
equation ( [ nabla ] ) is one constraint to the minimization problem arising from the total energy transport . using it
we can write @xmath112 as @xmath162 now we can combine all the constraints together by substituting this expression for @xmath163 into the dispersion relation , so that the result becomes the only constraint on the minimization problem in terms of the two variables @xmath101 and @xmath142 .
once the minimization is solved numerically , the value of @xmath161 is determined by @xmath101 and @xmath142 .
this in turn gives @xmath43 in terms of @xmath44 and @xmath66 using equation ( [ nabla ] ) .
we get the ratio of the radiative flux to the total flux ( @xmath164 ) in the differential equation ( [ df4 ] ) and then solve the four differential equations . outside the convection zone , @xmath165
, we have @xmath166 solving for @xmath43 , we have ( @xcite ) @xmath167 we can see the physical meaning of the turbulent efficiency in this case ( @xcite ) .
when @xmath81 goes to infinity , we have @xmath168 note that when turbulent transport is very efficient , it does not necessarily imply that the turbulent flux carries of the flux .
when @xmath81 goes to zero , we have @xmath169 finally , @xmath43 in equation ( [ frft ] ) gives the expression of @xmath151 : @xmath170 @xmath151 is equal to @xmath43 in equation ( [ delfrft ] ) outside the convection zones .
we note that using @xmath151 instead of @xmath44 does not matter much in the end .
when they differ , it will be because of either strong turbulent mixing or weak convection . in either case turbulent mixing will tend to suppress convection altogether .
the computer program used to solve the vertical structure equations is a modification of the one described in two previous papers ( @xcite , @xcite ) .
the adiabatic gradient @xmath66 , mean molecular weight @xmath40 , and coefficient of expansion @xmath171 were supplied by h. saio for these papers using the input physics given in @xcite .
we assumed a solar abundance of na , k , fe ( assuming that the overall ionization potential @xmath172 ev ) , and c. we use the rosseland opacities @xmath41 of @xcite for @xmath173 , those of @xcite for @xmath174 , and those of @xcite for @xmath175 .
we smoothed @xmath176 linearly over @xmath177 and @xmath178 to eliminate the slight discontinuity in those temperature ranges as done in the original code . in the calculation
, we set the mixing length @xmath62 for convection equal to max(@xmath18 , @xmath39 ) . the nonlinear minimization problem for the subjective function @xmath179 subject to the equality constraint equation ( [ dispersion ] ) with positive variables @xmath101 and @xmath142 is achieved by the fortran routines donpl2 ( @xcite ) and nlpql ( @xcite ) .
[ fig1 ] shows the thermal equilibrium s - curves for a dn at @xmath180 cm with a white dwarf mass @xmath181 .
every point in the thermal equilibrium curves is one solution to the differential equations ( [ df1 ] ) to ( [ df4 ] ) and each curve is associated with a different model for the vertical energy transport .
we do this to illustrate the effect of the different mechanisms and to show how our model differs from the previous standard model . here
we show the results of a model with heat carried by radiation and mlt convection ( @xmath182 ) , by radiation and turbulent flux driven by bh instability ( @xmath183 ) , by all three transport mechanisms but with mlt convection ( @xmath184 ) , or by all three transport mechanisms with modified ( damped ) mlt ( @xmath185 ) .
we have also added a fifth curve showing the sensitivity of our results to the choice of @xmath26 .
these curves differ mainly on the middle , unstable , branch because the disk is partially ionized at these temperatures and the resulting opacity peak lowers the efficiency of radiative transfer relative to convective or turbulent mixing .
the surface densities at the left and right turnovers of the unstable branches are usually denoted by @xmath186 and @xmath2 . in this case
, only @xmath2 shows any sensitivity to the input physics .
the short stable branch within the unstable branch , which appears in some of the curves , is an effect of very strong mixing .
it has been previously shown , for convective mixing , that the unstable branch moves up and to the right in the @xmath187 plane as the mixing length @xmath62 increases ( @xcite , @xcite , @xcite ) .
increasing the mixing length is equivalent to enhancing the strength of convection .
this suggests that regardless of the mixing mechanism at work , the upper - right shift and the intermediate branch of the unstable branch indicate the presence of stronger mixing .
this explains why both the standard mlt convection model and the addition of turbulent mixing without damping produce qualitatively similar curves , with the latter lying consistently to the right . on the other hand ,
the curve for radiation and turbulent mixing alone lies on top of the curve with damped convection .
this joint curve lies consistently to the left at lower temperatures , although the old standard model ( @xmath182 ) crosses it near the intermediate branch .
the fact that the addition of damped convection produces no significant changes indicates that turbulent damping is very efficient in this case .
the tendency of the joint curve to lie below , and to the left , of the othe curves shows that vertical heat transport via turbulent mixing is generally less effective than undamped convective transport .
the difference between the joint curve and the old standard model largely vanishes at high temperatures because the value of @xmath0 , and the consequent strength of turbulent mixing , increases in case .
when convection is very strong the convective growth rate becomes of order @xmath9 and @xmath188 so that a small @xmath0 implies a small role for turbulent mixing .
finally , we note that the results of our model differs from previous work in two interesting ways .
firstly , the intermediate stable branch is completely suppressed and this implies that dne accretion disks do not come to rest , even briefly , midway through the ionization transition .
this result would weaken the stagnation effect " ( @xcite ) which relied partly on the existence of this middle unstable branch to account for uv delay in the early rise of dn outbursts .
one would still have stagnation , however , due to the large specific heat at the temperatures between @xmath189 and @xmath190 k. secondly , the value of @xmath2 is reduced by about 20% and this is quantitatively interesting .
however , we see from this figure that this correction is less important than arriving at a definitive value of @xmath26 in the scaling law
. choosing @xmath191 , which is a conceivable , albeit extreme , choice , can almost double @xmath2 while producing only a minor change in @xmath186 .
[ fig2 ] shows the fractional heat fluxes as a function of height within a dn accretion disk .
the top panel shows a model with turbulent transport and damped convection while the bottom one is a model with undamped convection .
these two cases have the same total half - thickness @xmath192 and lie on the middle branch of the s - curve .
the undamped convection model is roughly at the upper end of the intermediate branch .
both models have @xmath193 and use the scaling law for @xmath0 .
there is no significant change in @xmath42 between the two models , indicating that the total mixing flux is roughly the same , but in the case of damped convection this flux is almost entirely carried by @xmath194 .
we note that formally the convectively unstable region ( @xmath195 ) is not dramatically smaller in the damped case . instead , while both models have broad , and roughly equal , convectively unstable zones ( @xmath196 ) , the shrunken convective zone shown in fig .
[ fig2 ] in the damped case is due to the lack of solutions to the minimization problem for weak convection .
_ this result for weak convection shows that the presence of an available source of free energy is not a guarantee of instability if the environment supplies strong damping_. in table [ tb - dn_conv ] we see that where convection is strongest in the undamped convection model , it just barely survives in the damped model , even though the eddies are almost isotropic and roughly adiabatic ( @xmath197 ) . at this level ,
even their complete elimination would not change the s - curve significantly since the curve for @xmath198 already coincides with the convectionless model .
the factor @xmath199 describing the turbulent losses is smaller as @xmath68 increases .
this means that convective and turbulent mixing becomes stronger ( or weaker ) than radiation simultaneously and this in turn explains why turbulent mixing is not important on high and low temperature stable branches of the s - curves where radiative transport dominates and the convectively unstable region is narrow .
although @xmath194 exists in the convectively stable region and carries energy toward the mid - plane , increasing the temperature gradient , @xmath194 behaves in other respects like @xmath200 with a shorter mixing length and is severely damped by radiative losses when radiative transport is important .
our calculations show that @xmath201 in the stable hot and cold states for the @xmath183 curve in fig .
[ fig1 ] , rendering mixing transport of any kind largely irrelevant .
this is why all the s - curves coalesce in the hot and cold states .
[ fig2 ] also shows that for both models turbulent mixing dominates near the midplane where the disk is more opaque and where convection is weak due to small @xmath202 . the numbers in parentheses in table [ tb - dn_conv ] ( or in table [ tb - lmxb_conv ] , and table [ tb - agn_conv ] ) are the corresponding results for the case @xmath203 for the same @xmath55 .
as expected , these numbers are larger than those for damped convection .
the growth rate , @xmath101 , for undamped ( mlt ) convection in these tables is @xmath204 as defined in the preceding section .
table [ tb - dn_alpha ] shows the values of @xmath0 along the s - curves in fig .
the outburst duration for dne seems to imply that @xmath205 in the hot state ( cf .
@xcite , @xcite ) .
evidently the scaling law leads to a considerable overestimate ( up to about 0.6 ) as we go up the hot state branch of the s - curve for dn disks .
in addition , the use of the scaling law for @xmath0 in the hot state in dn disks leads to a contradiction with the bailey relation which says that the decay time associated with the decline from outburst in dne increases with orbital period .
both of these points suggest that the scaling law should start to saturate as @xmath0 rises above @xmath206 .
we note that the scaling law is motivated by the behavior of sxts , which have no bailey relation and much stronger evidence for an exponential decline from outburst .
one simple compromise might be to let @xmath0 rise to a maximum value of @xmath207 and then abruptly go over to a constant .
this would have no effect on the middle and lower branches of the s - curve , where turbulent mixing and the suppression of convection can have some effect .
however , we note that a more physically motivated geometrical dependence for @xmath0 involves a prolonged transition from a steep scaling to essentially constant behavior ( @xcite ) .
an alternative possibility is to choose a smaller value of @xmath26 , equivalent to asserting that the central black hole mass in sxt s is smaller .
table [ tb - dn_alpha ] also displays the values of @xmath0 based on the formula @xmath208 along the s - curve indicated by the dash - dotted line in fig .
[ fig1 ] , equivalent to asserting that the central mass in sxt systems is typically @xmath209 , which is a very low value .
the s - curves for sxt disks are illustrated in fig .
[ fig3 ] at @xmath180 cm with the central object of @xmath210 .
the much larger @xmath186 and @xmath2 along with the more pronounced hysteresis curve for the total flux ( @xmath182 ) reflect on stronger convection in sxts than that in dne at @xmath211 .
this results from the larger @xmath35 for sxts .
the energy flux radiated from the surface is given by @xmath212 since s - curves for different systems operate at the similar @xmath213 , @xmath41 ( i.e. the condition for partially ionized hydrogen ) and @xmath214 , the above equation implies @xmath215 moreover , the scaling law for @xmath0 gives @xmath216 which actually does not depend strongly on @xmath35 .
together with equation ( [ sigma1 ] ) this gives @xmath217 in accordance with the usual result that the column densities in the ionization transition region drop as a cooling wave propagates inward in a disk ( for recent calculations see @xcite , @xcite ) .
similarly , we can obtain a crude estimate of the strength of convection on the unstable branch from equation ( [ isogamma ] ) .
since @xmath41 is roughly the same for disks with similar ionization fractions and temperatures , and since we have already seen that the midplane temperature differences will be slight , we can rewrite equation ( [ isogamma ] ) as @xmath218 where we have ignored the change in @xmath219 .
this implies that convection is relatively more important for larger central masses and at smaller radii .
in fact , sxt disks have larger column densities than dn disks for ionization transitions at the same @xmath25 and consequently stronger mixing .
[ fig3 ] shows that the curves ( @xmath182 ) and ( @xmath183 ) are more different than in fig .
conversely , the smaller difference between the curves @xmath203 and ( @xmath220 ) also illustrates the increased importance of convection ( and reduced importance of turbulent mixing ) .
adding to this effect is the smaller @xmath0 which reduces the importance of turbulent mixing and damping .
the end result is that the damped @xmath200 in sxts is stronger than in dne .
although the @xmath198 curve is still close to the curve with no convection , it consistently lies to the left and includes a very small intermediate branch .
as before the value of @xmath2 is reduced , in this case by about 30% , relative to the old standard model . following table [ tb - lmxb9_alpha ] )
we note that the values of @xmath0 used for these models never reach the unrealistically high values seen on the hot branch in dne .
[ fig4 ] is corresponds to fig .
[ fig2 ] , but for sxts instead of dne . in this case
the total height for both models is @xmath221 . the damped convection model in the top panel lies near the upper end of the short intermediate branch .
as expected , convection is much more significant in sxts in terms of both its intensity and its vertical range .
this is also seen in table [ tb - lmxb_conv ] which shows that @xmath179 is of the order of unity in the confined convective region , suggesting that damping is weak for the strongest convective modes .
in addition , the difference between that of damped and undamped convective speeds is much smaller than for the dne cases shown in table [ tb - dn_conv ] .
table [ tb - dn_conv ] and table [ tb - lmxb_conv ] ( including table [ tb - agn_conv ] which we will discuss later in agn cases ) also show that for sufficiently strong convection the aspect ratio @xmath142 usually increases with height since radiative damping ( @xmath222 ) and turbulent drag ( @xmath223 , see equation ( [ drag ] ) ) are more severe farther from the midplane .
in fact , for other values of @xmath55 we find that slightly radially distended convective eddies ( @xmath142 slightly larger than 1 ) can actually be favored at high altitudes .
the survival of strong convection in the middle of convectively unstable zones and the complete suppression of weak convection once the convective growth rate drops raises the question of convective overshooting .
the error that results from ignoring overshooting should be small in our case for two reasons .
first , the total mixing flux does not change if damping is added .
second , overshooting is important in cases where no other mixing mechanism exists . in accretion disks
there is always turbulent mixing .
finally , we note that in our case convective overshooting still carries energy away from the mid - plane in the overshoot region since @xmath195 .
this is in contrast to the stellar case , where @xmath200 is negative in overshoot region , which is convectively stable , and this in turn increases the temperature gradient .
[ fig3.9 ] shows the s - curves for sxt disks for @xmath224 and @xmath225 . although the typical @xmath35 is smaller than at a radius of @xmath211 , as expected from equation ( [ sigma ] ) , the larger value of @xmath226 ( cf . equation ( [ conv / rad ] ) ) and the smaller value of @xmath0 ( cf . equation ( [ alpha ] ) and table [ tb - lmxb9_alpha ] ) make convection slightly stronger and turbulent mixing slightly weaker .
consequently , the two curves ( @xmath182 ) and @xmath227 coalesce , while we see a larger difference between the curves for ( @xmath182 ) and ( @xmath183 ) .
however , when we compare @xmath112 to @xmath9 , we see that the strength of convection drops at smaller @xmath213 and that this effect is stronger at smaller radii .
the end result is that the final curve , @xmath198 follows the convective curves at high temperatures and switches abruptly to the turbulent mixing curve at low temperatures when the aspect ratio of the convective cells drops to the point where radiative losses and turbulent mixing turn off convection .
this leads to a pronounced intermediate branch at small radii . for @xmath228
the reduction in @xmath2 is less than 20% .
[ fig5 ] displays the s - curves for the usual agn parameters : @xmath229 and @xmath230 .
agn are extreme cases relative to dne because of their much smaller @xmath0 , which implies weaker turbulent damping and larger @xmath231 , and much larger @xmath35 , hence larger @xmath232 .
consequently the curve @xmath203 lies close to the curve ( @xmath182 ) and both are far from the curve ( @xmath183 ) .
there is almost no difference between the curve @xmath198 and the curve @xmath203 .
all of this is as expected for a disk where convection is strong and turbulent mixing is weak . the weakness of turbulent mixing is also revealed in the relatively large difference between @xmath233 and @xmath213 for the @xmath183 curves compared to the convection s - curves ; that is , the temperature difference is increased in the middle branch as a result of inefficient turbulent mixing .
we also ran the code for ( @xmath182 ) with a reduced mixing - length @xmath234 with the usual scaling law for @xmath0 .
the s - curve in this case , depicted by the line composed crosses in fig .
[ fig5 ] , is more or less similar to the cchp model , which attempts to account for shear through a general reduction in the mixing length , at least for @xmath235 ( @xcite ) .
we see that this approach exaggerates the suppression of convection for agn
. in our model convection remains relatively efficient for agn despite shearing and turbulent damping .
the relative importance of the vertical heat transport mechanisms for agn is illustrated in fig .
we choose models with @xmath236 lying on the unstable branch .
we see that the overall structure of the disk is affected only slightly by including turbulent damping of convective cells . in both cases turbulent
mixing dominates near the midplane and decreases rapidly in importance away from it .
radiative transport plays a major role only in the region just below the photosphere . the vertical structure of the damped convection zone
is also shown in table [ tb - agn_conv ] .
we see larger convective efficiency factors associated with radiative losses @xmath68
, small turbulent drag @xmath223 , and very small turbulent loss factors @xmath237 compared to those in table [ tb - dn_conv ] and table [ tb - lmxb_conv ] . as a result
, @xmath179 is close to unity and convection is well approximated , at least in terms of geometry , by isotropic convection without turbulent mixing . as in the sxt cases
, we also find that @xmath142 is occasionally bigger than 1 in the most robust convective modes . the larger role for turbulent losses
@xmath237 for agns is really a consequence of very small radiative losses , and does not reflect an overwhelming suppression of convection by turbulent mixing .
we can see this by considering the total efficiency factors obtained by dividing the @xmath68 without parentheses by the corresponding @xmath238 .
in fact , tables [ tb - dn_conv ] , [ tb - lmxb_conv ] , and [ tb - agn_conv ] show that the total convective efficiency @xmath239 for these three systems are of the same order .
this is because the small values of @xmath0 and huge values of @xmath35 for agns assure that radial turbulent mixing and radiative losses are almost negligible for them while the turbulent flux is extremely weak .
the very large values of @xmath35 implied by the scaling law for @xmath0 applied to agn disks ( see table [ tb - agn_alpha ] ) leads to the question of disk self - gravity .
if the self - gravity of the disk is as important as the gravity imposed by the central object , then local gravitational instabilities may be important and our description of agn disks is not self - consistent . in fig .
[ gself_agn.eps ] we see the ratio between self - gravity , @xmath240 , and @xmath241 for complete disk models of agn with turbulent damping . we see that the two become comparable when the agn disk @xmath213 approaches @xmath242 k. this should raise @xmath0 and turn the curves sharply to the left .
[ conv.eps ] and fig .
[ anisotropy.eps ] illustrate the overall behavior of convective flux along the unstable branches of different accretion disk systems .
the ranges of effective temperature displayed in these figures are associated with those for the thick solid lines ( the curves @xmath198 ) in fig .
[ fig1 ] , fig .
[ fig3 ] , and fig .
maximum values of fractional convective fluxes ( both mlt and damped ) and the vertical extent of the convective zones , @xmath243 ( scaled with half - thickness @xmath55 ) , are depicted for damped convection as a function of @xmath213 in fig .
[ conv.eps ] .
mlt convection ( dashed lines ) becomes stronger and stronger as one goes from dne ( top panel ) , to sxts ( middle panel ) , to agn ( bottom panel ) and the difference between damped and mlt convection diminishes over the same range .
the importance of damped convective flux can also be measured by @xmath243 ( dotted lines ) , where @xmath243 is the vertical range for which solutions to the minimization problem exist ( for which convection can actually occur ) . on average , @xmath243 increases from @xmath244 for dne to @xmath245 for agn .
[ anisotropy.eps ] displays the maximal and minimal values of @xmath246 at a given @xmath213 associated with the s - curves for @xmath198 ( denoted by thick solid lines in figures [ fig1 ] , [ fig3 ] , and [ fig5 ] ) for different systems . as mentioned above , @xmath142 tends to increase with height and so a larger separation between maximal and minimal @xmath142 goes with a larger vertical range for damped convection .
values of @xmath246 as small as @xmath247 can exist for agn but not for dne in accord with the result that convection is strong in agn .
in fact , table [ tb - dn_conv ] and table [ tb - agn_conv ] show that the convective growth rate @xmath248 for dn at @xmath249 is comparable to that for agn at @xmath250 .
no solutions exist on the left side of dotted lines for weak convection as a result of the severe damping of narrower convective cells . on the other hand
, ` fatter ' ( @xmath251 ) convective eddies can exist for strong convection , since these solutions reduce radiative losses and radial mixing and extract only a small penalty in the form of increased vulnerability to secondary instabilities .
metals are the main supply of free electrons in the cold , neutral accretion disk state . in our code the ratio of electron gas pressure to the total gas @xmath252 is @xmath253 around @xmath2 , indicating that the disks are only weakly ionized .
we calculated the microscopic resistivities due to the electron - neutral collisions using the equation @xmath254 ( @xcite ) and used the standard formula for electron - ion collisions , @xmath255 .
we found magnetic reynolds numbers for the isotropic turbulence driven by bh instability is @xmath256 . ]
@xmath257 in quiescent dn and sxt disks are less than @xmath258 , in accord with previous work ( @xcite ) , although this neglects the possibility of ionizing particles from the disk coronae .
[ metal.fig ] illustrates how the s - curves for disks with solar metal abundances differ from ones without any metals around @xmath2 for different systems . in order to distinguish the effects of mlt convection from turbulent mixing
, we use the curves for ( @xmath182 ) and ( @xmath183 ) . for curves with mlt convection , the discrepancy between metal and no metal models near cases around @xmath2 increases as one goes from dne , to sxts , to agns .
this is because convective mixing becomes stronger in agns . on the other hand
, the opposite trend happens to the curves with turbulent mixing which follows from the fact that turbulent mixing becomes weaker in agn . finally , in fig .
( [ s_curve_dnconst.eps ] ) we test the consequences of using a constant value of @xmath0 instead of the scaling law .
this is motivated by gammie and menou s suggestion ( @xcite , see also @xcite ) that the timing of dn outbursts can be explained by the relatively high resistivity of quiescent dn accretion disks .
if the bh instability shuts off below @xmath259 , then the low values of @xmath0 inferred from the intervals between outbursts reflect simply the operation of some different angular momentum transport mechanism in quiescent disks , and not a general scaling law .
( [ s_curve_dnconst.eps ] ) shows the s - curve down to relatively low temperatures , and the associated range of @xmath260 .
the important feature here is that a cutoff of @xmath261 or less results in s - curves with a very small range in column density .
while @xmath2 may still be very large if @xmath0 drops sharply at still smaller temperatures , the propagation of a cooling front in an accretion disk depends to a large extent on the nature of the s - curve near @xmath186 and detailed models of cooling fronts indicate that the cooling sections of the disk trace a path just below the unstable equilibrium curve ( @xcite ) .
a cooling wave with a nearly vertical drop , as seen here , can not be described by the rapid cooling wave theory ( @xcite,@xcite ) but should instead propagate at a fraction of the viscous accretion speed , just as expected for an s - curve with a small value of @xmath262 ( @xcite ) .
in fact , models with constant @xmath0 not only show slow cooling front speeds but also frequent reversals ( @xcite , @xcite ) .
such models lack a sharp drop in @xmath0 at very low temperatures and may not represent realistic calculations of the results of gammie and menou s model .
however , it is not clear that even the existence of a second unstable branch at very low temperatures would affect the cycling of a dn disk .
the s - curve shown here is sufficient to produce a complete , albeit weak , thermal cycle without ever moving to lower temperatures .
we caution however that this conclusions depends on using a particular value of @xmath260 for a cutoff .
if the cutoff value of @xmath260 lies in the range @xmath263 to @xmath264 then a robust thermal limit cycle is still be possible for a constant @xmath0 model .
we have proposed a new model for vertical energy transport in optically thick , keplerian accretion disks which takes into account the effects of turbulent mixing due to the bh instability .
this includes both an evaluation of the turbulent flux @xmath194 driven by approximately isotropic turbulent cells with an effective vertical thermal conductivity @xmath265 , and the increased damping of the sheared convective cells .
the latter effect includes not only radiative losses , which have been broadly discussed in the literature , but also turbulent mixing which enhances heat transport and retards convective motion .
simple physical arguments , and the linear dispersion relation , indicate that weak ( @xmath266 ) convective eddies are radially thin in the shearing flow of an accretion disks .
this enhances the effects of turbulent drag , radial mixing , and radiative losses .
we account for these effects through an extension of linear perturbation theory using a one mode model .
we choose the linear mode most resistant to secondary instabilities and to turbulent damping and by calibrate the effects of this mode by requiring agreement with mlt if turbulent viscosity and shearing are removed . following this procedure ,
our model for convection becomes a minimization problem for the function @xmath179 subject to the linear dispersion relation involving several damping mechanisms .
our use of quasilinear theory , calculating quadratic transport quantities and typical eddy shapes using linear modes , is reasonable , but not rigorously justifiable .
generally this procedure works only when energy flows through a system in ways that already appear in a linear analysis .
the appearance of strong finite amplitude instabilities would destroy the rationale behind this approach .
fortunately thin accretion disks do not appear to possess such instabilities ( @xcite ) .
one of the other major components of our approach , the suppression of one family of unstable modes ( convection , in this case ) by another is unusual , but not unprecedented . in laboratory plasmas
we have the example of radial convection being disrupted by the spontaneous appearance of shear ( cf .
burrell 1997 ) . in the context of accretion
disks the suppression of the parker instability by bh induced turbulence ( @xcite ) , and previously predicted in vishniac and diamond ( 1992 ) , is very closely analogous to suppression of convection .
the most problematic aspect of our treatment is our method for choosing a ` typical ' linear mode ( cf . equation ( [ criterion ] ) ) . as far as we know this method has never been proposed before and there are no definitive tests of its usefulness .
however , it does capture the correct limits , i.e. settling on the remaining unstable mode in the case of marginal stability and choosing the modes least susceptible to secondary kelvin - helmholtz instabilities when damping is unimportant .
we conclude that it is at least qualitatively correct .
the thermal equilibrium s - curves calculated in our model differ from previous models in several ways .
our models tend to reproduce previous work along the hot stable branch of the s - curve , and give the same , or nearly the same @xmath186
. however , the unstable branch tends to lie somewhat below , and to the left , of previous work in the @xmath3 plane .
in addition , the short , intermediate stable branch seen in previous work often disappears , and is usually more nearly vertical when it does appear .
the value of @xmath2 is smaller in our models , though only by 20 - 30% . crudely speaking ,
these effects are stronger for thicker disks , that is , for dne and sxts , rather than agn . however , in some respects these effects are more dramatic for sxts than for dne , largely because the different transport mechanisms give rise to more divergent s - curves in the former .
agn disks are still reasonably well described by standard mlt theory without turbulent mixing .
there are two reasons for the way turbulent effects scale from dne to agn .
first , in an average sense , convection becomes stronger as we go to thinner disks around more massive objects .
this leads to faster convective growth rates , which means that convective cells become more resistant to disruption by any reasonable level of local turbulence .
second , we have adopted a scaling law which reinforces this trend by lowering the value of @xmath0 , and consequently the strength of turbulent dissipation , in thinner disks .
since @xmath0 decreases and @xmath35 increases as one progresses from dne , to sxts , to agns , we find an almost complete suppression of convection in dne .
models with no convection are reasonably good approximations to our models with damped convection .
the specific calculations presented here rely to large extent on rather uncertain approximations in extending mlt to strongly sheared environments and in estimating the level turbulent dissipation in disks . while the procedure followed here
is reasonable enough , it would clearly be better to have means of calibrating our theory for anisotropic convection .
nevertheless , the qualitative nature of our results are not sensitive to the specific choices we have made . on the whole ,
the results are less sensitive to the exact level at which weak convection is suppressed than to the qualitative point that weak convection is vulnerable to turbulent mixing in a sheared environment . the work presented here is based on the assumption that the scaling law equation ( [ eq : ascale ] ) for @xmath0 is a reasonable approximation .
this tends to make our corrections to mlt less important for thin disks , especially agn disks .
in fact , the status of this scaling law is somewhat uncertain , for reasons given above .
however , we note that a constant @xmath0 , down to some limiting value of the magnetic reynolds number , seems unable to reproduce the basic phenomenology of dne outbursts . at the same time
a very low @xmath0 seems required from observations of quiescent dne ( for example @xcite ) .
unless the cooling wave theory is discarded altogther it seems necessary to postulate some sort of geometric dependence for @xmath0 . on the other hand ,
the bailey relation for dne suggests that the scaling law saturates for values of @xmath0 typical of dne in outburst .
finally , we note that the role of irradiation ( @xcite ) and possible advection flows ( @xcite ) in sxts remain unclear , and prevent a clear understanding of the form of @xmath0 required to explain observations of thin disks .
we thank jaw - luen tang and wei - jr wu for their assistance in the code .
we thank the anonymous referee for valuable comments .
we are grateful to craig wheeler and stefano migliuolo for informative discussions .
we also appreciate the prompt help of paola dalessio in obtaining his phd thesis on the internet
. part of this work was completed when one of us ( pg ) was a visitor at the high energy physics laboratory in university of texas at austin , and he would like to thank roy schwitters for his generous hospitality .
this work was supported in part by nasa grant nag5 - 2773 .
crrrrrrrr 0.67 & 2.01(5.01 ) & 0.07(0.51 ) & 0.81 & 117 & 1.10(2.07 ) & 0.03(0.21 ) & 0.76 & 0.06 0.70 & 1.72(3.29 ) & 0.12(0.56 ) & 0.88 & 45 & 1.16(1.98 ) & 0.05(0.22 ) & 0.77 & 0.11 0.72 & 1.02(1.57 ) & 0.16(0.61 ) & 0.95 & 26 & 1.18(1.70 ) & 0.06(0.23 ) & 0.81 & 0.21 0.74 & 0.12(1.57 ) & 0.06(0.58 ) & 0.91 & 249 & 1.04(1.44 ) & 0.02(0.21 ) & 0.99 & 0.42 crrrrrrrr 0.47 & 34.5(83.5 ) & 0.07(0.21 ) & 0.55 & 23 & 1.25(1.51 ) & 0.03(0.09 ) & 0.30 & 0.01 0.66 & 14.4(15.7 ) & 0.18(0.36 ) & 0.64 & 3.2 & 1.49(1.50 ) & 0.07(0.14 ) & 0.21 & 0.09 0.72 & 6.63(6.95 ) & 0.25(0.44 ) & 0.72 & 2.4 & 1.53(1.52 ) & 0.09(0.15 ) & 0.22 & 0.20 0.79 & 1.11(1.35 ) & 0.33(0.52 ) & 0.81 & 2.0 & 1.35(1.36 ) & 0.10(0.15 ) & 0.28 & 0.61 crrrrrrrr 0.28 & 7.73@xmath280(7.40@xmath281 ) & 0.05(0.11 ) & 0.28 & 10.8 & 1.06(1.03 ) & 0.01(0.03 ) & 9.92@xmath282 & 1.70@xmath283 0.46 & 4.91@xmath280(2.54@xmath281 ) & 0.11(0.16 ) & 0.35 & 4.6 & 1.07(1.05 ) & 0.03(0.04 ) & 9.45@xmath282 & 5.61@xmath283 0.66 & 9.33@xmath284(2.46@xmath280 ) & 0.28(0.41 ) & 0.56 & 2.1 & 1.15(1.12 ) & 0.05(0.06 ) & 8.81@xmath282 & 5.94@xmath285 0.85 & 1.66@xmath286(3.05@xmath286 ) & 0.93(1.12 ) & 0.90 & 1.13 & 1.60(1.56 ) & 0.11(0.12 ) & 9.66@xmath282 & 7.90@xmath287 0.93 & 1.63(2.84 ) & 2.99(3.52 ) & 0.99 & 1.02 & 1.80(2.25 ) & 0.24(0.27 ) & 1.68@xmath288 & 9.36@xmath288 alexander , d. r. , johnson , h. r. , and rypma , r. l. 1983 , , 272 , 773 armitage , p. j. , livio , m. , and pringle , j. e. 1996 , , 457 , 332 balbus , s. , and hawley , j.f .
1991 , , 376 , 214 balbus , s. , and hawley , j. f. 1998 , rev .
phys . , 70 , 1 batchelor , g.k .
1950 , proc .
lond . , a201 , 405 blaes , o. m. , and balbus , s. a. 1994 , , 421 , 163 burrell , k.h .
1997 , phys .
plasmas , 4 , 1499 cabot , w. , canuto , v. m. , hubickyj , o. , & pollack , j. b. 1987 , icarus , 69 , 423 cannizzo , j. k. 1992 , , 385 , 94 cannizzo , j. k. 1993 , in accretion disks in compact stellar systems , ed .
j. c. wheeler
( singapore : world scientific ) , p. 6
cannizzo , j. k. 1994 , , 435 , 389 cannizzo , j. k. , and cameron , a. g. w. 1988 , , 330 , 327 cannizzo , j. k. , chen , w. , and livio , m. 1995 , , 454 , 880 cannizzo , j. k. , and wheeler , j. c. 1984 , , 55 , 367 canuto v. m. , and hartke , g. j. 1986 , , 168 , 89 chandrasekhar , s. 1960 , proc , nat .
, 46 , 253 chen , w. , shrader , c. r. , and livio , m. 1997 , , 491 , 312 cox , j. p. , and giuli , r. t. 1968 , principles of stellar structure ( new york : gordon and breach ) cox , a. n. , and stewart , j. n. 1969 , sci . inf .
council , acad .
ussr , no .
15 , p. 1 .
dalessio , p. , cant , j. , calvet , n. , and lizano , s. 1998 , , 500 , 411 dubus , g. , la - sota , j.p . ,
hameury , j .-
m . , and charles , p.a . 1999 , , 302 , 731 gammie , c.f .
1997 , astro - ph/9712233 gammie , c.f . , and menou , k. 1998 , , 492 , l75 hameury , j .-
m . , menou , k. , dubus , g. , lasota , j .-
p . , and hure , j .-
1998 , , 298 , 1048 hawley , j. , and balbus , s. 1992 , , 400 , 595 hawley , j.f . ,
balbus , s.a . , and winters , w.f . 1999 , , 518 , 394 honma , f. 1996 , , 48 , 77 iben , jr . , i. 1995 , , 138 , 452 kazantsev , a.p . 1967 , jetp , 53 , 1806 king , a.r . , and ritter , h. 1998 , , 293 , 42 lin , d. n. c. , and shields , g. a. 1986 , , 305 , 28 livio , m. , and spruit , h. c. 1991 , , 252 , 189 ludwig , k. , meyer - hofmeister , e. , and ritter , h. 1994 , , 290 , 473 maeder , a. 1995 , , 299 , 84 maeder , a. , and meynet g. 1995 , , 313 , 140 menou , k. , hameury , j .- m . , and stehle , r. 1999 , , in press meyer , f. , and meyer - hofmeister e. 1982 , , 106 , 34 mineshige , s. 1988 , , 190 , 72 mineshige , s. , and osaki , y. 1983 , , 35 , 377 narayan , r. , mcclintock , j. e. , and yi , i. 1996 , , 457 , 821 osaki , y. 1996 , , 108 , 39 pojmaski , g. 1986 , acta astr . , 36 , 69 pollack , j. b. , mckay , c. p. , and christofferson , b. m. 1985 , icarus , 64 , 471 ruden , s. p. , papaloizou , j. c. b. , & lin , d. n. c. 1988 , , 329 , 739 rdiger , c. , elstner , d. , and tschpe , r. , 1988 , acta astr . , 38 , 299 sakimoto , p. j. , and coroniti , f. v. , 1981 , , 247 , 19 schittkowski , k. 1985/86 , annals of operations research , 5 , 4850 shakura , n. i. , and sunyaev , r. a. 1973 , , 24 , 337 smak , j. 1984 , acta astr . , 34 , 161 spellucci , p. 1998 , url : http://plato.la.asu.edu/guide.html speziale , c.g . ,
sarkar , s. , and gatski , t.b .
1991 , j. fluid mech . , 227 , 245 speziale , c.g . ,
sarkar , s. , and gatski , t.b .
1997 , j. fluid mech .
, 344 , 155 stone , j. , hawley , j.f . ,
gammie , c.f . , and balbus , s.a .
1996 , , 463 , 656 talon , s. , and zahn , j .- p .
1997 , , 317 , 749 velikhov , e. p. 1959
, soviet phys .
- jetp , 36 , 1398 vishniac , e. t. 1993 , in accretion disks in compact stellar systems , ed . j. c. wheeler ( singapore : world scientific ) , p. 41
vishniac , e. t. 1997 , , 482 , 414 vishniac , e.t .
1998 , proceedings of the 1997 maryland meeting , in press ( available as astro - ph/9802232 ) vishniac , e.t . 1999 ,
proceedings of the 1998 chapman conference on magnetic helicity , ed .
a. pevtsov ( agu : new york ) vishniac , e. t. , and diamond , p. h. 1992 , , 398 , 561 vishniac , e.t . , and brandenburg , a. 1997 , , 475 , 263 vishniac , e. t. , and wheeler , j. c. 1996 , , 471 , 921 wood , j. h. , horne , k. , berriman , g. , & wade , r. a. 1989 , , 341 , 974 wood , j. h. , horne , k. , berriman , g. , wade , r. a. , & odonoghue , d. 1986 , , 219 , 629 zahn , j .- p . , 1992 , , 265 , 115 | we consider vertical heat transport in keplerian accretion disks , including the effects of radiation , convection , and turbulent mixing driven by the balbus - hawley instability , in astronomical systems ranging from dwarf novae ( dne ) , and soft x - ray transients ( sxts ) , to active galactic nuclei ( agn ) . in order to account for the interplay between convective and turbulent energy transport in a shearing environment
, we propose a modified , anisotropic form of mixing - length theory , which includes radiative and turbulent damping .
we also include turbulent heat transport , which acts everywhere within disks , regardless of whether or not they are stably stratified , and can move entropy in either direction .
we have generated a series of vertical structure models and thermal equilibrium curves using the scaling law for the viscosity parameter @xmath0 suggested by the exponential decay of the x - ray luminosity in sxts .
we have also included equilibrium curves for dne using an @xmath0 which is constant down to a small magnetic reynolds number ( @xmath1 ) .
our models indicate that weak convection is usually eliminated by turbulent radial mixing , even when mixing length estimates of convective vertical heat transport are much larger than turbulent heat transport .
the substitution of turbulent heat transport for convection is more important on the unstable branches of thermal equilibrium s - curves when @xmath0 is larger .
the low temperature turnover points @xmath2 on the equilibrium s - curves are significantly reduced by turbulent mixing in dne and sxt disks . however , in agn disks the standard mixing - length theory for convection is still a useful approximation when we use the scaling law for @xmath0 , since these disks are very thin at the relevant radii . in accordance with previous work , we find that constant @xmath0 models give almost vertical s - curves in the @xmath3 plane and consequently imply very slow , possibly oscillating , cooling waves . |
the design of reliable and efficient a posteriori error indicators for guiding local refinement when solving numerically partial differential equations is essential , both for defining a robust and automatic adaptive procedure and for ensuring to find suitable approximations of the desired solution without exceeding the limits of available softwares .
the main idea behind a posteriori error estimation is to build a properly locally refined mesh in order to equidistribute the approximation error .
whereas for standard finite element methods several intuitive ways of refining locally a mesh are clear and broadly analysed , for isogeometric methods @xcite , the development of efficient and robust strategies to get suitably locally refined meshes constitutes a challenging problem because the tensor product structure of b - splines @xcite is broken .
different alternatives have been proposed in order to tackle this situation , such as hierarchical splines , t - splines , lr - splines or pht - splines . among them , hierarchical splines based on the construction presented in @xcite ( see also the previous work @xcite ) are probably the easiest to define and to implement for their use in the context of isogeometric methods . regarding a posteriori error estimation using hierachical spline spaces for discretizations in isogeometric methods , up to this moment we can only mention the results from @xcite , where residual based error indicators has been proposed . in that case , the authors considered an approach following the standard techniques in classical finite elements for deriving element - based a posteriori error estimators using the truncated basis for hierachical splines introduced in @xcite .
the presented proof for the reliability of such estimators needs to assume some restrictions over the hierachical meshes , with the purpose of controlling the overlap of the truncated basis function supports . we notice that although truncation is indeed a possible strategy to recover partition of unity , this procedure requires a specific construction that entails complicated basis function supports
, that may be non convex and/or not connected , and their use may produce a non negligible overhead with an adaptive strategy .
thus , in this article , we consider the hierachical basis without truncation recovering the simplicity of basis function supports , that in this case are boxes .
moreover , taking into account the hierachical space defined in @xcite , we can also recover the partition of unity .
the main goal of this article is to obtain simple residual type a posteriori error estimators for linear second order elliptic problems using discretizations in hierarchical spline spaces .
as pointed out in @xcite , where the hierarchical basis can be obtained simply through parent - children relations , the design of estimators associated to basis functions ( instead of elements ) seems to be more suitable or natural for guiding adaptive refinements .
we derive reliable function - based a posteriori error indicators without any restrictions over the hierachical mesh configurations , i.e. , we are able to bound the error in energy norm by our global a posteriori error indicator . in order to prove such upper bound for the error , it is key the use of some poincar type inequalities where the b - splines are weight functions and also that their supports are , obviously , convex sets . in this point
it is important to remark that a posteriori error estimations in the context of classical finite element methods have been widely studied in the available literature , see for example , @xcite and @xcite . on a hand ,
our approach can be considered as a generalization to high order splines of some existent poincar type inequalities ; and on the other hand , in order to do that , we follow closely the lines from @xcite , where specific poincar type inequalities are proved and the classical barycentric coordinate functions appear as weight functions .
this article is organized as follows . in section [
s : setting ] we briefly introduce the variational formulation of the elliptic problem that we consider , and in section [ s : discretization ] we describe precisely the hierachical spline spaces that we use for its galerkin discretization .
next , we state and prove some poincar type inequalities that have b - splines as weight functions in section [ s : poincare ] , which are used in section [ s : estimators ] to derive function - based a posteriori error estimators and to prove that such estimators constitute an upper bound for the energy error . in section [ s : refinement ]
we analyse a reduction property of our estimators after refinement of the hierachical mesh . finally , in section [ s : aigm ] we propose an adaptive algorithm guided by our estimators and illustrate its behaviour through several numerical tests , showing that the global estimator is efficient and the algorithm experimentally converges with the optimal rate .
for simplicity , we consider the following linear elliptic problem on the parametric domain @xmath0^d\subset { \mathbb r}^d$ ] , @xmath1 , @xmath2 where @xmath3 is uniformly symmetric positive definite over @xmath4 , i.e. , there exist constants @xmath5 such that @xmath6 @xmath7 , @xmath8 .
we assume that @xmath9 .
we say that @xmath10 is a weak solution of if @xmath11=f(v),\quad\forall\,v\in h^1_0(\omega),\ ] ] where @xmath12 is the bounded bilinear form given by @xmath13:=\int_\omega { \mathcal a}\nabla u\cdot \nabla v + \bm{b } \cdot \nabla u \,v + c\,u\,v,\ ] ] and @xmath14 is the lineal functional defined by @xmath15 taking into account and that @xmath9 , it is easy to check that @xmath16 is coercive , that is , @xmath17,\qquad\forall\,v\in h^1_0(\omega).\ ] ] thus , as a consequence of the lax - milgram theorem , we have that problem is well posed .
in this section we revise briefly the definitions of univariate and tensor product b - splines that we use to build a basis for a hierachical spline space like those from @xcite .
then , we state the discrete formulation of problem when considering such spaces . [ [ univariate - b - spline - bases ] ] univariate b - spline bases + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath18 be a @xmath19-open knot vector , i.e. , a _ sequence _ such that @xmath20 where the two positive integers @xmath19 and @xmath21 denote a given polynomial degree , and the corresponding number of b - splines defined over the subdivision @xmath22 , respectively . here , @xmath23 .
we also introduce the _ set _
@xmath24 of breakpoints ( i.e. , knots without repetitions ) , and denote by @xmath25 the multiplicity of the breakpoint @xmath26 , such that @xmath27 with @xmath28 .
note that the two extreme knots are repeated @xmath29 times , i.e. , @xmath30 .
we assume that an internal knot can be repeated at most @xmath29 times , that is , @xmath31 , for @xmath32 .
let @xmath33 be the b - spline basis @xcite associated to the knot vector @xmath22 .
in particular , we remark that the _ local knot vector _ of @xmath34 is given by @xmath35 , which is a subsequence of @xmath36 consecutive knots of @xmath22 ; and that the support of @xmath34 , denoted by @xmath37 , is the closed interval @xmath38 $ ]
. additionally , the b - spline basis @xmath39 is in fact a basis for the space @xmath40 of the piecewise polynomials of degree @xmath19 over the mesh @xmath41\,|\,j=1,\dots,{\tilde n}-1\}$ ] , that have @xmath42 continuous derivatives at the breakpoint @xmath26 , for @xmath43 . if @xmath44 for some @xmath45 , the splines in @xmath40 can be discontinuous at @xmath26 . [ [ tensor - product - b - spline - bases ] ] tensor product b - spline bases + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath46 . in order to define a tensor product @xmath47-variate spline function space on the parametric domain @xmath48^d\subset { \mathbb r}^d$ ]
, we consider @xmath49 the vector of polynomial degrees with respect to each coordinate direction and @xmath50 , where @xmath51 . for @xmath52 ,
let @xmath53 be a @xmath54-open knot vector , i.e. , @xmath55 where the two extreme knots are repeated @xmath56 times and any internal knot can be repeated at most @xmath56 times .
we denote by @xmath57 the tensor product spline space spanned by the b - spline basis @xmath58 defined as the tensor product of the univariate b - spline bases @xmath59 .
more precisely , @xmath60 if and only if @xmath61 where @xmath62 , for @xmath63 , and @xmath64 denotes the @xmath45-th component of @xmath65 .
we notice that the support of @xmath66 , denoted by @xmath67 , is a box in @xmath68 given by @xmath69 finally , the associated cartesian grid @xmath70 consists of the cells @xmath71 , where @xmath72 is an element ( closed interval ) of the @xmath73-th univariate mesh @xmath74 , for @xmath75 .
[ [ sequence - of - tensor - product - spline - spaces ] ] sequence of tensor product spline spaces + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in order to define a hierachical structure , we assume that there exists an underlying sequence of tensor product @xmath47-variate spline spaces @xmath76 , where @xmath77 is called _ the space of level @xmath78 _ , such that @xmath79 each of these spaces are indeed obtained from a tensorization of univariate spline spaces as we explain now . let @xmath49 denote the chosen vector of polynomial degrees for the univariate splines in each coordinate direction . for @xmath80 , @xmath81 is the tensor product spline space and @xmath82 is the corresponding b - spline basis , that we call the set of _ b - splines of level @xmath78 _ , for some @xmath83 . in order to guarantee , we assume that if @xmath84 is a knot in @xmath85 with multiplicity @xmath86 , then @xmath84 is also a knot in @xmath87 with multiplicity at least @xmath86 , for @xmath75 and @xmath80 .
furthermore , we denote by @xmath88 the corresponding cartesian mesh , and we say that @xmath89 is a _ cell of level @xmath78_. we note that we assume that the cells are closed sets .
b - splines possess several important properties , such as non - negativity , partition of unity , local linear independence and local support , that make them suitable for design and analysis , see @xcite for details .
moreover , we have that b - splines of level @xmath78 can be written as linear combinations of b - splines of level @xmath90 with non - negative coefficients , which is known as _ two - scale relation_. more specifically , if @xmath91 , where @xmath92 is the dimension of the space @xmath77 , for @xmath80 ; this property can be stated as follows : @xmath93 with @xmath94 .
we notice that , due to the local linear independence of b - splines , only a limited number of the coefficients @xmath95 are different from zero . taking into account
we define the set of children of @xmath96 , denoted by @xmath97 , consisting of the functions @xmath98 such that @xmath99 , that is , @xmath100 as we will see later on , in cases of interest such as subsequent levels obtained by dyadic refinement , the number of children is bounded and the bound depends solely on the degree @xmath101 .
[ [ hierarchical - spline - space ] ] hierarchical spline space + + + + + + + + + + + + + + + + + + + + + + + + + in @xcite the authors considered a particular subspace of the hierarchical space presented in @xcite , which still enjoys of good local approximation properties and leads to simple refinement schemes , because it is defined in a way that focus on the relation between functions . in order to define precisely a basis for the hierachical space that we consider ,
we first need to fix a hierarchy of subdomains of @xmath102^d$ ] as in the next definition , which in turn provides the different levels in the multilevel structure .
[ def : hierarchy of subdomains][hierarchy of subdomains ] let @xmath103 be arbitrary .
we say that the set @xmath104 is a hierarchy of subdomains of @xmath4 of depth @xmath21 if @xmath105 and each subdomain @xmath106 is the union of cells of level @xmath107 , for @xmath108 .
we now are in position of introducing the basis for the hierachical space .
[ d : hierachical basis][hierarchical basis ] let @xmath76 be a sequence of spaces like with the corresponding b - spline bases @xmath109 , and @xmath104 a hierarchy of subdomains of depth @xmath21 .
we define the _ hierarchical basis _
@xmath110 computed with the following recursive algorithm : @xmath111 an interesting property of the hierachical basis @xmath112 is that the coefficients for writing the unity are strictly positive .
that is , we have @xmath113 with @xmath114 ( see ( * ? ? ?
* theorem 5.2 ) ) . in the following
we say that @xmath66 is an _ active function _
if @xmath115 , it is an _ active function of level @xmath78 _ if @xmath116 , and it is a _ deactivated function of level @xmath78 _ if @xmath117 .
moreover , @xmath118 is the union of active and deactivated functions of level @xmath78 .
we remark that , unlike in the definition given in @xcite where a b - spline of level @xmath90 is added to @xmath119 if its support is contained in @xmath120 , in definition [ d : hierachical basis ] b - splines of level @xmath90 are added only if they are children of a deactivated function of level @xmath78 . the hierarchical spline basis @xmath112 is associated to an underlying _ hierarchical mesh _
@xmath121 , given by @xmath122 analogously , we say that @xmath123 is an _ active cell _
if @xmath124 , and it is an _ active cell of level @xmath78 _ if @xmath125 .
we will also say that @xmath123 is a _ deactivated cell of level @xmath78 _ if @xmath126 and @xmath127 .
finally , we notice that a b - spline of level @xmath78 is active if all the active cells within its support are of level @xmath78 or higher , and at least one of such cells is actually of level @xmath78 .
a b - spline is deactivated when all the cells of its level within the support are deactivated .
[ [ discretization - of - the - variational - problem ] ] discretization of the variational problem + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in order to discretize problem we consider a hierarchy of subdomains @xmath128 of @xmath4 and the corresponding spline space @xmath129 with the hierarchical basis @xmath112 and the mesh @xmath130 as defined above .
now , we define the discrete space @xmath131 by @xmath132 thus , the discrete counterpart of consists in finding @xmath133 such that @xmath134=f(v),\quad\forall\,v\in{\mathcal s}_0.\ ] ]
in this section we briefly revise some basic notions about weighted sobolev spaces and then we state weighted poincar type inequalities ( see theorems [ t : poincare inequality ] and [ t : friedrichs ] below ) that will be needed for proving the reliability of the function - based a posteriori error estimators to be presented in the next section .
let @xmath135 be a bounded domain with lipschitz boundary .
if @xmath136 is a nonnegative locally integrable function , we denote by @xmath137 the space of measurable functions @xmath138 such that @xmath139 notice that @xmath137 is a hilbert space equipped with the scalar product @xmath140 we also define the weighted sobolev space @xmath141 of weakly differentiable functions @xmath138 such that @xmath142 , where @xmath143 finally , @xmath144 is the closure of @xmath145 in @xmath141 . before stating the main result of this section , we recall the definition of concave functions . [
d : concave function ] a function @xmath146 defined on a convex set @xmath135 is _ concave on @xmath147 _ if for any @xmath148 , @xmath149 , there holds @xmath150 the weighted poincar inequality stated in @xcite holds for a weight @xmath136 such that @xmath151 is a concave function on its support , for some @xmath152 .
thus , in view of theorem [ t : b - spline concave ] below , which states that this is the case when @xmath136 is a multivariate b - spline basis function , the following result is an immediate consequence of ( * ? ? ?
1.1 and thm . 1.2 ) ( see also ( * ? ? ?
* lemma 5.2 ) ) .
[ t : poincare inequality ] if @xmath66 is a tensor product b - spline basis function , then @xmath153 where @xmath154 .
in order to prove that @xmath155 is a concave function on its support , for some @xmath152 , when @xmath66 is a tensor product b - spline basis function , we first notice that the result holds for univariate b - splines thanks to the brunn minkowski inequality , as explained in ( * ? ? ?
* section 2 ) .
more precisely , the following result holds .
[ l : concave ] let @xmath66 be a univariate b - spline basis function of degree @xmath19 . then , @xmath156 is concave on its support
. on the other hand , the following _ generalized cauchy - schwarz inequality _ can be proved by mathematical induction : @xmath157 for all nonnegative numbers @xmath158 .
now , as a consequence of we have following result .
[ l : product of concave functions ] if @xmath159 are nonnegative concave functions on a convex set @xmath160 then @xmath161 is concave on @xmath147 .
the assertion of this lemma follows from taking @xmath162 and @xmath163 , for @xmath75 , and regarding definition [ d : concave function ] . finally , using lemma [ l : concave ] and the last lemma we can prove the following result . [
t : b - spline concave ] let @xmath66 be a tensor product @xmath47-variate b - spline basis function as in .
if the univariate b - splines that define @xmath66 are of degree @xmath19 , then @xmath164 is concave on its support .
from we have that @xmath165 , where @xmath166 are univariate b - splines for @xmath63 , and @xmath64 denotes the @xmath45-th component of @xmath65 .
now , by lemma [ l : concave ] we have that @xmath167 is concave , for @xmath63 . finally , applying lemma [ l : product of concave functions ] we obtain that @xmath168 is concave on its support . when considering dirichlet boundary conditions as in problem , it is useful to have a suitable weighted poincar - friedrichs inequality in the case that the weight function is a b - spline which does not vanish on a part of the boundary of @xmath4 .
more precisely , we have the following result , which generalizes ( * ? ? ?
* lemma 5.1 ) to high order splines .
[ t : friedrichs ] let @xmath66 be a tensor product b - spline basis function such that @xmath169 .
then , there exists a constant @xmath170 , independent of @xmath66 , such that @xmath171 for all @xmath172 satisfying @xmath173 , where @xmath174 is a set with positive @xmath175-dimensional lebesgue measure .
the friedrichs inequality stated in the last theorem can be proved following the same steps from @xcite . in particular ,
such inequality can be obtained as a consequence of the weighted poincar inequality given in theorem [ t : poincare inequality ] and a suitable weighted trace inequality .
the following result is a trace theorem where b - splines are used as weight functions and can be seen as an extension of ( * ? ? ?
* proposition 4.3 ) ; its proof follows exactly the same lines , but we include it here for completeness .
[ p : weighted trace theorem ] let @xmath66 be a tensor product b - spline basis function given by @xmath176 where @xmath166 are univariate b - splines of a fixed degree @xmath19 , for @xmath63 , and @xmath64 denotes the @xmath45-th component of @xmath65 .
assume that @xmath169 , and let @xmath177 be a cell of the associated cartesian grid that has a side @xmath178 such that @xmath179 .
then , @xmath180 where @xmath181 , for @xmath182 . here , @xmath183 is any vertex of @xmath123 that does not belong to @xmath184 and @xmath73 denotes the coordinate direction given by the unit vector @xmath185 that is orthogonal to the side @xmath184 . in this proof
we use the symbol @xmath186 to denote the average of a function @xmath146 over a set @xmath147 , that is , @xmath187 . if @xmath188 , as an immediate consequence of gauss divergence theorem ( see also ( * ? ? ?
* proposition 4.2 ) ) to the vector field @xmath189 , it follows that @xmath190 now , let @xmath191 .
applying with @xmath192 we obtain @xmath193 let @xmath73 be the coordinate direction given by the unit vector @xmath185 that is orthogonal to the side @xmath184
. taking into account that @xmath179 and that @xmath178 , we have that @xmath194 , where @xmath183 is any vertex of @xmath123 that does not belong to @xmath184 , which in turn implies that @xmath195 thus , from we obtain that @xmath196 finally , follows from the last equation dividing both sides by @xmath197 , and taking into account that @xmath198 .
we finish this section using the last proposition for proving the weighted friedrichs inequality stated in theorem [ t : friedrichs ] .
let @xmath199 and let @xmath172 such that @xmath173 . then , @xmath200 let @xmath201 and @xmath202 be a cell that has @xmath184 as a side . applying proposition [ p : weighted trace theorem ] with @xmath203 , taking into account that @xmath204 and using hlder inequality we have @xmath205 now we use the last inequality to bound the second term in the right hand side of , @xmath206 where we have used the cauchy - schwarz inequality .
finally , follows taking into account , theorem [ t : poincare inequality ] and the last inequality .
in this section we use the poincar type inequalities proved in the previous section in order to get an a posteriori upper bound for the energy error when computing the discrete solution of problem using the hierarchical schemes proposed in section [ s : discretization ] .
once such key inequalities are available , the procedure to derive reliable residual type a posteriori error estimators follows the standard steps used when considering classical finite elements .
let @xmath207 be the solution of problem and @xmath208 be the galerkin approximation of @xmath138 satisfying .
specifically , the main goal of this section is to define some computable quantities @xmath209 , for @xmath115 , so that @xmath210 for some constant @xmath211 .
notice that the coercivity of the bilinear form implies that @xmath212 = \frac{1}{\gamma_1 } \langle{\mathbf r}(u),u - u\rangle,\ ] ] where the _ residual _ @xmath213 of a function @xmath214 is given by @xmath215=b[u - v , v ] , \qquad \text{for all $ v\in h^1_0(\omega)$}.\ ] ] thus , we have that @xmath216 and therefore , we have to bound @xmath217 . assuming that @xmath218 , since @xmath208 , integration by parts yields @xmath219 now , for each @xmath220 we associate a discrete function @xmath221 given by @xmath222 here , the coefficients @xmath223 are those given by and @xmath224 ( cf . ) .
taking into account that @xmath225 on @xmath4 and using that @xmath226 satisfies , we have that @xmath227 by hlder s inequality and the weighted poincar type inequalities given in theorems [ t : poincare inequality ] and [ t : friedrichs ] , it follows that @xmath228 where @xmath229 ( cf . ) . in consequence , @xmath230 regarding
we finally obtain that @xmath231 let @xmath232 be the sequence of underlying cartesian meshes associated to the different levels as explained in section [ s : discretization ] . without losing generality ,
we assume that @xmath233 where @xmath234 denotes the diameter of the cell @xmath123 . now
, we define the _ meshsize @xmath235 at level @xmath78 _ ( corresponding to the cartesian grid @xmath236 ) by @xmath237 we also define @xmath238 where @xmath78 is such that @xmath239 ; and notice that @xmath240 is equivalent to @xmath241 .
finally , for @xmath242 and @xmath115 , we define the _ local error indicator _ @xmath243 by @xmath244 and summarize we have just proved in the following result .
let @xmath207 be the solution of problem and @xmath208 be the galerkin approximation of @xmath138 satisfying .
then , there exists a constant @xmath211 such that @xmath245
in this section we explain precisely how to perform the refinement of a hierachical mesh and we show that the error estimator defined in the previous section is reduced after mesh refinement in the sense of ( * ? ? ?
* corollary 3.4 ) , which is an important property in order to get results about the optimality of adaptive algorithms .
we start with the following basic definition .
let @xmath104 and @xmath246 be hierarchies of subdomains of @xmath4 of depth ( at most ) @xmath21 and @xmath247 , respectively .
we say that @xmath248 is an _
enlargement _ of @xmath249 if @xmath250 in order to enlarge the current subdomains we have to select the regions in @xmath4 where more ability of approximation is required .
such a choice can be done by selecting to _ refine _ some active functions , as we explain in section [ s : aigm ] below where we will consider a precise way of enlarging the hierarchy @xmath249 . if @xmath248 is an enlargement of @xmath249 , we denote by @xmath251 the hierarchical basis ( associated to the hierarchy @xmath248 ) as in definition [ d : hierachical basis ] . from ( * ? ? ?
* theorem 5.4 ) ) it follows that @xmath252 and thus , we say that @xmath251 is a _ refinement _ of @xmath112 .
finally , we denote by @xmath253 the _ refined _ hierarchical mesh given by @xmath254 let @xmath232 be the sequence of underlying cartesian meshes associated to the different levels as explained in section [ s : discretization ] and let @xmath255 be the sequence of meshsizes defined by . in order to analyse the behaviour of the error estimator under refinement
, we assume that the successive levels are obtained performing @xmath256-adic refinement in the tensor product meshes .
more precisely , we state the following assumption .
[ a : q adic refinement ] there exists @xmath257 , @xmath258 such that @xmath259 let @xmath112 be the hierarchical basis associated to a hierarchy of subdomains of depth @xmath21 , @xmath104 .
let @xmath260 be the sequence of positive numbers as in such that @xmath261 , for @xmath262 .
for @xmath242 and @xmath115 , the local error indicator @xmath243 defined in is given by @xmath263 where @xmath240 is defined by .
additionally , for @xmath264 we define @xmath265 by @xmath266 we remark that the _ global indicator _ @xmath267 can be computed as @xmath268 where the function @xmath269 is given by @xmath270 the main result of this section is the following proposition . [
p : estimator reduction 1 ] let @xmath255 be the meshsizes defined in and let assumption [ a : q adic refinement ] be valid .
let @xmath112 be a hierarchical basis and let @xmath251 be a refinement of @xmath112 .
if @xmath271 denotes the set of _ refined basis functions _
, then @xmath272 where @xmath273 .
this result is a consequence of lemma [ l : aux estimator reduction ] stated below . in the proof of that lemma
we will use the following result which is a consequence of the fact that deactivated functions of level @xmath78 ( in @xmath251 ) can be written in terms of active functions of higher levels ( in @xmath251 ) .
more precisely , from ( * ? ? ?
* lemma 5.4 ) we have that @xmath274 [ c : decomposition ] let @xmath112 be a hierarchical basis and let @xmath251 be a refinement of @xmath112 . if @xmath271 denotes the set of refined functions , then @xmath275 let @xmath276 for some @xmath277 . since @xmath278 , @xmath279
thus , we have that @xmath280 due to @xmath281 .
finally , implies that @xmath282 .
[ l : aux estimator reduction ] let @xmath255 be the meshsizes defined in and let assumption [ a : q adic refinement ] be valid .
let @xmath112 be a hierarchical basis and let @xmath251 be a refinement of @xmath112 .
if @xmath271 denotes the set of refined functions , then @xmath283 where @xmath284 and @xmath285 are defined as in , and @xmath286 is the constant appearing in assumption [ a : q adic refinement ] . notice that @xmath287 by corollary [ c : decomposition ] , we have that @xmath288 and therefore , @xmath289 for some constants @xmath290
then , @xmath291 notice that the last equation gives a partition of unity with functions @xmath251 .
let @xmath292 be the sequence of positive numbers such that @xmath293 notice that the difference between and is that in the former some basis functions appear more than once .
finally , taking into account assumption [ a : q adic refinement ] we have that @xmath294 which concludes the proof .
we finish this section with the proof of the estimator reduction property .
let @xmath242 and let @xmath273 . by lemma [ l : aux estimator reduction ] , we have that @xmath295
in this section we propose an adaptive algorithm guided by the a posteriori error estimators defined in section [ s : estimators ] .
additionally , we show the performance of such adaptive procedure in practice through several numerical examples . the adaptive loop that we consider is quite standard and consists of the following modules : @xmath296 we start with a tensor product mesh and the corresponding spline space , regarded as the current hierarchical mesh @xmath130 and hierarchical space @xmath297 , respectively .
we perform the steps in in order to get an adaptively refined mesh @xmath253 and its corresponding hierachical space @xmath298 .
next , we consider @xmath253 and @xmath251 as the current hierachical mesh and basis , respectively , and perform the steps in , and so on .
we now briefly describe the modules of the adaptive loop .
* solve : compute the solution @xmath226 of the discrete problem in the current hierachical space @xmath297 .
* estimate : use the current discrete solution @xmath226 to compute the _ a posteriori error estimators _
@xmath299 defined in , for each @xmath115 .
* mark : use the a posteriori error estimators @xmath300 to compute the set of _ marked functions _ @xmath301 using the _ maximum strategy _ with parameter @xmath302 , that is , @xmath303 consists of the basis functions @xmath304 such that @xmath305 * refine : use the set of marked functions @xmath306 to enlarge the current hierarchy of subdomains @xmath307 as follows : + let @xmath308 , for @xmath277 .
now , we define the hierarchy of subdomains @xmath309 of depth at most @xmath247 , by @xmath310 if @xmath251 is the hierarchical basis associated to @xmath311 , we notice that @xmath312 ; in other words , at least the functions in @xmath306 have been refined and removed from the hierarchical basis @xmath112 .
now , we present some numerical tests to show the performance of the proposed algorithm . in particular
, we study the decay of the energy error in terms of degrees of freedom ( dofs ) in each example , and analyse the rates of convergence .
the implementation of the adaptive procedure was done using the data structure and algorithms introduced in @xcite .
we consider the problem @xmath313 giving in each particular example the definition of the domain @xmath4 and the problem data @xmath146 and @xmath314 .
[ ex : regular solution ] we consider @xmath102\times [ 0,1]$ ] and the problem data @xmath146 and @xmath314 in are chosen such that the exact solution @xmath138 is given by @xmath315 . in figure [ f : example regular solution meshes ] we plot the exact solution , some hierarchical meshes and the decay of the energy error vs. degrees of freddom for different spline degrees .
as expected , both tensor product meshes and hierarchical meshes reach optimal orders of convergence , but notice that in all cases , the curves corresponding to the adaptive strategy are meaningfully by below .
for example , for attaining an energy error of around @xmath316 using biquadratics , the global refinement requires @xmath317 dofs whereas the adaptive strategy only needs @xmath318 dofs .
some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) . notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) . notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : regular solution ] ; for biquadratics with @xmath319 elements and @xmath320 dofs ( top left ) , bicubics with @xmath321 elements and @xmath322 dofs ( top middle ) and biquartics with @xmath323 elements and @xmath324 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] additionally , in figure [ f : example regular solution meshes 2 ] we present the adaptive meshes obtained for different polynomial degrees , starting with an initial tensor product mesh of @xmath326 elements , reaching in all cases an enegy error @xmath327 .
it is interesting to remark that although it is well known @xcite that the element - by - element assembly is very costly for higher degree , the use of adaptivity changes this picture .
indeed , due to the reduced number of elements for bicubics and biquartics , the time - to - solution for bicubics is @xmath328 and for biquartics is @xmath329 of the time - to - solution for biquadratics .
comparison of adaptive meshes obtained for example [ ex : regular solution ] using different polynomial degrees , obtaining in all cases that @xmath330 .
the mesh for biquadratics has @xmath319 elements and @xmath320 dofs ( left ) , the mesh bicubics has @xmath331 elements and @xmath332 dofs ( middle ) and the mesh for biquartics has @xmath333 elements and @xmath334 dofs ( right).,title="fig : " ] comparison of adaptive meshes obtained for example [ ex : regular solution ] using different polynomial degrees , obtaining in all cases that @xmath330 .
the mesh for biquadratics has @xmath319 elements and @xmath320 dofs ( left ) , the mesh bicubics has @xmath331 elements and @xmath332 dofs ( middle ) and the mesh for biquartics has @xmath333 elements and @xmath334 dofs ( right).,title="fig : " ] comparison of adaptive meshes obtained for example [ ex : regular solution ] using different polynomial degrees , obtaining in all cases that @xmath330 .
the mesh for biquadratics has @xmath319 elements and @xmath320 dofs ( left ) , the mesh bicubics has @xmath331 elements and @xmath332 dofs ( middle ) and the mesh for biquartics has @xmath333 elements and @xmath334 dofs ( right).,title="fig : " ] [ ex : diagonal ] we take @xmath102\times[0,1]$ ] and choose @xmath146 and @xmath314 such that the exact solution @xmath138 of is given by @xmath335 . in figure [ f : example diagonal ] we plot the exact solution , some hierarchical meshes and the decay of the energy error vs. degrees of freddom for different spline degrees . as in the previous example , both tensor product meshes and hierarchical meshes reach optimal orders of convergence , but notice that in all cases , the curves corresponding to the adaptive strategy are again meaningfully by below . for example , for attaining an energy error of around @xmath336 using biquadratics
, the global refinement requires @xmath337 dofs whereas the adaptive strategy only needs @xmath338 dofs .
some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) . notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) . notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : diagonal ] ; for biquadratics with @xmath339 elements and @xmath340 dofs ( top left ) , bicubics with @xmath341 elements and @xmath342 dofs ( top middle ) and biquartics with @xmath343 elements and @xmath338 dofs ( top right ) .
notice that although all meshes have nearly the same amount of elements , the refinement is more spread for high order splines due to the sizes of the basis function supports .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] [ ex : lshaped ] we consider the l - shaped domain @xmath344 ^ 2\setminus((0,1)\times(-1,0))$ ] and choose @xmath146 and @xmath314 such that the exact solution @xmath138 of is given in polar coordinates by @xmath345 . in figure
[ f : example l - shaped ] we plot the exact solution , some hierarchical meshes and the decay of the energy error vs. degrees of freddom for different spline degrees .
we notice that the global refinement associated to tensor product spaces does not reach the optimal order of convergence due to the singularity . on the other hand
, the adaptive strategy recovers the optimal decay for the energy error given by @xmath346 .
some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) .
in addition , we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : lshaped ] obtaining in all cases @xmath347 ; for biquadratics with @xmath348 elements and @xmath349 dofs ( top left ) , bicubics with @xmath350 elements and @xmath351 dofs ( top middle ) and biquartics with @xmath352 elements and @xmath353 dofs ( top right ) . in addition
, we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] [ ex : singular solution ] we consider a problem whose solution is not too smooth .
specifically , we take @xmath354\times[0,1]$ ] , and choose @xmath355 and @xmath146 such that the exact solution @xmath138 of is given by @xmath356 .
notice that in this case , @xmath357 and there are singularities along the sides @xmath358 and @xmath359 , being a bit stronger the singularity along @xmath358 ; see figure [ f : example singular solution ] .
some hierarchical meshes and the error decay in terms of degrees of freddom for different polynomial degrees are presented in figure [ f : example singular solution meshes ] .
we notice that both global refinement and the adaptive refinement reach the optimal order of convergence when using biquadratics ( bottom left ) , but only the adaptive refinement converges with optimal rates when using bicubics ( bottom middle ) and biquartics ( bottom right ) , due to the singularity of the solution .
the exact solution @xmath356 ( left ) corresponding to example [ ex : singular solution ] and its derivatives @xmath360 ( middle ) and @xmath361 ( right).,title="fig : " ] the exact solution @xmath356 ( left ) corresponding to example [ ex : singular solution ] and its derivatives @xmath360 ( middle ) and @xmath361 ( right).,title="fig : " ] the exact solution @xmath356 ( left ) corresponding to example [ ex : singular solution ] and its derivatives @xmath360 ( middle ) and @xmath361 ( right).,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] comparison of adaptive meshes for the solution of example [ ex : singular solution ] for different polynomial degrees , obtaining in all cases , @xmath362 .
the mesh for biquadratics has @xmath363 elements and @xmath364 dofs ( top left ) , the mesh for bicubics has @xmath365 elements and @xmath366 dofs ( top middle ) and the mesh for biquartics has @xmath367 elements and @xmath368 dofs ( top right ) . on the other hand , we plot the energy error @xmath325 vs. degrees of freedom for different polynomials degrees.,title="fig : " ] [ ex : ring ] in this case , we consider the domain @xmath4 given in polar coordinates by @xmath369 and we choose the problem data @xmath146 and @xmath314 in such that the exact solution @xmath138 is given by @xmath315 . despite optimal rates of convergence
are reached using both tensor product meshes and hierarchical meshes ( see figure [ f : example ring ] ) , we emphasize that in this case the adaptive strategy is still convenient . as an example
, we notice that to get an energy error of @xmath370 using bicubics , the adaptive strategy requieres less than the @xmath371 of the degrees of freedom utilised by the global refinement , because the former procedure requires @xmath372 dofs whereas the latter needs @xmath373 dofs .
some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] some hierarchical meshes for the solution of example [ ex : ring ] obtaining @xmath374 in all cases ; for biquadratics with @xmath375 elements and @xmath376 dofs ( top left ) , bicubics with @xmath377 elements and @xmath378 dofs ( top middle ) and biquartics with @xmath379 elements and @xmath380 dofs ( top right ) .
we plot the energy error @xmath325 vs. degrees of freedom ; for biquadratics ( bottom left ) , bicubics ( bottom middle ) and biquartics ( bottom right).,title="fig : " ] [ ex : cube ] we consider the cube @xmath102 ^ 3 $ ] and choose @xmath146 and @xmath314 such that the exact solution @xmath138 of is given by @xmath381 . since the solution is smooth enough , both strategies reach optimal orders of convergence , as showed in figure [ f : example cube errors ] .
however , we notice that in all cases , the curves corresponding to the adaptive strategy are meaningfully by below , which in practice is equivalent to achieve a given accuracy with considerably fewer degrees of freddom .
energy error decay in terms of degrees of freedom for the solution of the example [ ex : cube ] ; using biquadratics ( left ) , bicubics ( middle ) and biquartics ( right).,title="fig : " ] energy error decay in terms of degrees of freedom for the solution of the example [ ex : cube ] ; using biquadratics ( left ) , bicubics ( middle ) and biquartics ( right).,title="fig : " ] energy error decay in terms of degrees of freedom for the solution of the example [ ex : cube ] ; using biquadratics ( left ) , bicubics ( middle ) and biquartics ( right).,title="fig : " ] [ [ on - the - efficiency - of - the - error - estimators ] ] on the efficiency of the error estimators + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + finally , we analyse the behaviour of the efficiency index @xmath382 . in figure
[ f : efficiency indices ] , we plot this index at each iteration step for all the examples previously presented . we see that the energy error @xmath383 and the global a posteriori error estimator @xmath384 are equivalent quantities , that is , there exists constants @xmath385 and @xmath386 such that @xmath387 at each iteration step .
thus , we conclude that our estimators are not only reliable but also experimentally efficient .
efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ] efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ] efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ] efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ] efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ] efficiency indices @xmath388 ; for example [ ex : regular solution ] ( top left ) , example [ ex : diagonal ] ( top middle ) , example [ ex : lshaped ] ( top right ) , example [ ex : singular solution ] ( bottom left ) , example [ ex : ring ] ( bottom middle ) , example [ ex : cube ] ( bottom right).,title="fig : " ]
a. buffa was partially supported by erc adg project change n. 694515 , by miur prin project `` metodologie innovative nella modellistica differenziale numerica '' , and by istituto nazionale di alta matematica ( indam ) .
garau was partially supported by conicet through grant pip 112 - 2011 - 0100742 , by universidad nacional del litoral through grants cai+d 500 201101 00029 li , 501 201101 00476 li , by agencia nacional de promocin cientfica y tecnolgica , through grants pict-2012 - 2590 and pict-2014 - 2522 ( argentina ) . this support is gratefully acknowledged .
kraft , r. ( 1997 ) .
adaptive and linearly independent multilevel @xmath16-splines . in _ surface fitting and multiresolution methods ( chamonix mont - blanc , 1996 ) _ , pages 209218 .
vanderbilt univ . press , nashville , | in this article we develop function - based a posteriori error estimators for the solution of linear second order elliptic problems considering hierarchical spline spaces for the galerkin discretization .
we prove a global upper bound for the energy error .
the theory hinges on some weighted poincar type inequalities , where the b - spline basis functions are the weights appearing in the norms .
such inequalities are derived following the lines in @xcite , where the case of standard finite elements is considered . additionally , we present numerical experiments that show the efficiency of the error estimators independently of the degree of the splines used for discretization , together with an adaptive algorithm guided by these local estimators that yields optimal meshes and rates of convergence , exhibiting an excellent performance .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * keywords : * a posteriori error estimators , adaptivity , hierarchical splines _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
fast radio bursts ( frbs ) are bright millisecond flashes mostly found at high galactic latitudes @xcite .
since discovery , they have received a large amount of theoretical study on possible progenitors , burst mechanisms ( e.g. * ? ? ?
* ; * ? ? ?
* ) and potential usage to probe the intergalactic medium ( igm , * ? ? ?
* ; * ? ? ?
* ) and cosmology @xcite . the strongest argument for their cosmological origin is that the large dispersion measures ( dms ) , the line - of - sight electron column density @xmath8 , are inconsistent with being from the average interstellar medium ( ism ) , stellar corona , hii regions and supernova remnants ( snrs ) in the galaxy ( or even the local group , * ? ? ?
* ; * ? ? ?
the fact that frb 121102 has been observed to repeat @xcite rules out catastrophic events at least for this burst .
frb models are mostly based on neutron stars ( nss ) , which could naturally accommodate the short durations @xmath9 , bright coherent emission , repetitivity on a range of intervals ( @xmath10 to @xmath11 or longer , depending on the flux level ) .
in addition , many frbs show pulse widths of @xmath12 consistent with multi - path propagation spreading and are much longer than what intergalactic or galactic scattering could account for @xcite .
this means the local plasma surrounding the progenitors must be more strongly scattering than the average ism .
the rotation measure given by the linearly polarized frb 110523 is much larger than intergalactic or galactic contributions , meaning the progenitor is located in a dense magnetized nebula @xcite .
the dm from a snr decreases with time , but the repeating frb 121102 has a constant @xmath13 for 3 yrs .
this means a possible ns must be older than @xmath14 and the dm from the snr is smaller than @xmath15 @xcite .
therefore , frbs are likely from nss of 100 yr to 1 myr old embedded in snrs or star - forming regions . due to insufficient monitoring time , the other ( so - far ) `` non - repeating '' frbs could also be repeating . in this _ letter _ , we assume that frbs are from nss at cosmological distances and that they repeat with a universal energy distribution function ( edf ) .
below , we first summarize the properties of frb 121102 in section [ sec:121102 ] , and then explore the answers to the following questions : ( i ) _ do frb statistics so far support a universal edf ?
if so , what constraints can we put on it ? _
( ii ) _ is frb 121102 representative of the ensemble ? _ ( iii ) _ what is the spacial density of frb progenitors ? _
the isotropic equivalent energy of each burst is @xmath16 where @xmath17 is the fluence , @xmath7 is the radio emission efficiency , @xmath18 is the luminosity distancecdm cosmology with @xmath19 , @xmath20 , and @xmath21 . ] and @xmath22 is the bandwidth of the frb spectrum .
the dm from igm is given by @xcite @xmath23 and @xmath24 is the number density of free electrons .
frb 121102 has @xmath25 which corresponds to redshift @xmath26 and luminosity distance @xmath27 .
the repeating rate above fluence @xmath28 at 1.4 ghz is @xmath29 .
the progenitor s time - averaged isotropic equivalent luminosity is @xmath30 if we add up the total burst fluence of @xmath31 during arecibo . ]
telescopes on - source time @xmath32 . the energy reservoir required to supply the bursting activity for a time @xmath33 is @xmath34 where @xmath35 is the total sky coverage of all bursts from one progenitor .
if frbs are concentrated only in a narrow range of stellar latitude ( e.g. poles ) , we have @xmath36 .
the fluence distribution function of frb 121102 is a power - law @xmath37 @xcite , so @xmath38 should mainly come from the most energetic bursts ( which we missed due to insufficient monitoring time ) , and this is why eq.([eq:3 ] ) is a lower limit .
if the energy reservoir is the magnetosphere of a ns , it requires a magnetic field strength of @xmath39 .
in this section , we assume frb rate tracks the core - collapse supernova rate and that all frbs repeat with a universal edf . we calculate the detection rate of frbs on the earth as a function of source redshift , and then by comparing it with observations , we constrain the parameters of the edf
. core - collapse supernova rate tracks the cosmic star - formation rate and is given by @xmath40^{5.6 } } \rm\ gpc^{-3}\ d^{-1}\ ] ] we assume that a fraction @xmath41 of nss are able to produce observable frbs and stay in the active phase for a local - frame time @xmath42 . during the active phase
, a ns undergoes multiple ( @xmath43 ) bursts intermittently and we call it a `` bns '' , short for `` bursting neutron star '' .
the edf is assumed to be a power - law with a high - end cutoff have been observed from frb 121102 , a possible low - end cutoff could be at @xmath44 . as can be seen later from eq.([eq:11 ] ) , such a low @xmath45 will be observationally noticed only for very nearby sources at distances @xmath46 , which corresponds to a @xmath47 .
current observations have found no frbs below @xmath48 and hence put no constraint on the possible low - end cutoff . ]
@xmath49 where @xmath50 is the reference burst energy corresponding to a fluence @xmath51 when the source is at redshift @xmath52 and @xmath5 is the integrated rate above @xmath50 . without losing generality , we base the reference point ( @xmath50 , @xmath52 and @xmath51 ) on frb 121102 @xmath53 where @xmath52 has an uncertainty of @xmath54 ( due to unknown @xmath55 ) , the uncertainty of @xmath50 mostly comes from @xmath52 ( through eq .
[ eq:4 ] ) , and @xmath7 is the radio emission efficiency .
the statistics of frb 121102 give @xmath56 , but we keep the normalization constant @xmath5 as a free parameter , since whether frb 121102 is representative is to be determined .
the other two free parameters ( @xmath57 , @xmath58 ) are restricted in the ranges @xmath59 and @xmath60 , because weaker bursts are more frequent and we have detected bursts with @xmath61 . integrating over the cosmic volume
, we get the all - sky detection rate above a fluence threshold @xmath62 @xmath63 where @xmath64 is the differential comoving volume , @xmath65 is the threshold energy above which bursts from a given redshift @xmath66 are detectable @xmath67 and @xmath68 is the given by @xmath69 . combining eq.([eq:8])([eq:11 ] ) , we calculate the differential detection rate @xmath70 as a function of redshift .
we assume that the dimensionless quantity @xmath71 does not depend on redshift , so the normalized cumulative distribution of the all - sky event rate only depends on the two free parameters ( @xmath57 , @xmath72 ) .
we use @xmath73 formalism to fit it with the observational normalized cumulative distribution of dm , i.e. @xmath74 where @xmath75 is only the igm component ( eq .
[ eq:14 ] ) and we use a constant host galaxy contribution and the difference is small compared to the other uncertainties . ]
@xmath76 . the allowed parameter space in the ( @xmath57 , @xmath72 )
is determined by the significance probability @xmath77 . on the other hand , by matching the normalization at a given @xmath57 and @xmath72 with the all - sky detection rate from observations
, we calculate the product @xmath78 , so the allowed parameter space in the @xmath57-@xmath72 plane constrains @xmath78 . in the frb search strategy
( de - dispersion @xmath79 single - pulse search ) , the signal - to - noise ratio and hence detectability depends on both the fluence and de - dispersed pulse width as @xmath80 @xcite .
if there is a maximum intrinsic pulse width @xmath81 , the threshold fluence for a given @xmath82 is @xmath83 . @xcite
derive a completeness threshold of @xmath84 for the parkes frbs .
therefore , we test our model only on parkes frbs with @xmath85 , which converts to a sample of 8 bursts @xcite
. some other factors could introduce biases : ( i ) scattering by the galactic ism may bias against detection of frbs at low galactic latitudes @xcite ; ( ii ) in de - dispersion trials , typically , dms of @xmath86 and @xmath87 are not considered , resulting in effectively a low- ( @xmath88 ) and high - redshift ( @xmath89 ) blindness ; ( iii ) the source position within a single beam ( @xmath90 ) is unknown and most authors use the on - axis assumption and report a lower limit .
due to ( i ) , we have discarded frbs that occurred below galactic latitude @xmath91 .
it turns out ( ii ) does not introduce significant biases because the observed frb are all between @xmath92 and the true rate should cut off sharply below and above this range .
based on ( iii ) , we also try the sample selected by @xmath93 ( with 11 frbs ) for comparison . in fig.([fig : comp ] ) , we compare the theoretical event rate above threshold @xmath94 with observations , for three different power - law slopes ( @xmath95 ) and various energy cutoffs ( @xmath58 from 10 to @xmath96 ) .
we can see that , for a larger @xmath57 and smaller @xmath72 , more frbs are expected to have small @xmath97 .
we quantify the goodness of the fit by calculating the reduced @xmath73 , as shown in fig.([fig : chisq ] ) .
the upper panel is for @xmath94 ( 8 bursts ) and the lower for @xmath98 ( 11 bursts ) .
we can see that the two samples generally agree with each other and the allowed parameter space is @xmath99 the bottom - right corner with \{@xmath100 , @xmath101 } may be in tension with observations but the confidence level is not high and more data is needed .
we also mark the power - law index for frb 121102 @xmath102 , with 1- and 2-@xmath103 errors between dashed and dotted lines .
the agreement with the constraints from the whole sample means that frbs may indeed share a universal edf .
frb 121102 may be representative in that its power - law index lies within the range allowed by the statistics from the whole population .
as for the normalization , we draw in fig.([fig : ndet ] ) the contours of @xmath104\ ] ] for the two samples with @xmath105 ( upper ) and @xmath106 ( lower panel ) . dashed curves are for negative values . for the parameter space allowed by parkes data ( red contours , eq . [ eq:16 ] )
, we obtain @xmath107 .
the observational all - sky frb rate above @xmath108 is estimated to be @xmath109 @xcite .
considering all the uncertainties , we obtain @xmath110 however , the normalization parameter @xmath5 is still unconstrained so far and will be discussed in next section .
@xcite conducted a follow - up survey of the fields of 8 known frbs as in section [ sec : obs ] ( only 3 of them overlap ) . ] and found none repeating .
other follow - up surveys , e.g. @xcite , are less constraining . for a given burst @xmath111 at redshift @xmath112 ,
the average number of repeating events above the threshold @xmath62 within a monitoring time @xmath113 is @xmath114.\ ] ] the probability of observing none of the 8 repeating is @xmath115 we take the monitoring time @xmath116 from @xcite and assume that the survey is sensitive to any repeating bursts above the threshold @xmath94 .
we calculate @xmath117 as a function of @xmath57 and @xmath72 from eq.([eq:18 ] ) and ( [ eq:19 ] ) and show the result in fig.([fig : p0rep ] ) for @xmath118 ( upper ) and @xmath119 ( lower pannel ) .
we find that , if all frbs repeat at the same rate as frb 121102 ( @xmath120 ) , the probability of observing none repeating is @xmath121 @xmath122 .
this , again , supports frb 121102 as being representative of the whole population .
note that @xcite used a relatively short cutoff of de - dispersed width @xmath123 , which could lead to incompleteness above @xmath84 and further increases @xmath117 .
on the other hand , @xcite detected a new frb at redshift @xmath124 in the survey .
the average number of bnss at redshift @xmath125 in a sky area of @xmath126 is @xmath127 the probability of having at least one bns with @xmath128 in the observed area is @xmath129 the product @xmath78 is constrained by eq.([eq:17 ] ) , so we have @xmath130 if every frb repeats similarly as frb 121102 ( @xmath131 ) , the probability of having @xmath132 frbs at @xmath128 in a 4 @xmath133 area is @xmath134 , with lower @xmath5 giving larger @xmath135 . putting @xmath117 and @xmath135 together ,
we find the observations by @xcite favor a normalization constant @xmath5 smaller than the value of frb 121102 , but @xmath120 is not ruled out at a high confidence . a relatively steep power - law with @xmath136
is also favored , as can be seen in fig.([fig : p0rep ] ) .
therefore , we conclude that frb 121102 may be slightly more active than average ( with a relatively large @xmath5 ) but is so far consistent with being representative of the ensemble .
this _ letter _ is based on three hypotheses : ( i ) frbs are produced by nss at cosmological distances with their dms mostly due to free electrons in the igm ; ( ii ) frb rate tracks core - collapse supernova rate ; ( iii ) frbs repeat with a universal edf @xmath137 with a high - end cutoff at @xmath1 . based on these assumptions , we find that observations so far are consistent with a universal edf with power - law index @xmath2 , high - end cutoff @xmath138 and normalization @xmath4 , where @xmath5 is defined as the integrated rate above @xmath139 ( @xmath7 being the radio emission efficiency ) .
we also put constraints on the dimensionless product @xmath140 , where @xmath41 is the fraction of nss that are able to produce observable frbs and @xmath42 is the length of the bursting phase .
better statistics in the future could narrow down the uncertainties , if our hypotheses are correct .
some of the implications of our results on the frb progenitors are as follows : * the edf of frbs is shallower than the power - law tail of crab giant pulses ( @xmath141 , * ? ? ?
* ) and consistent with magnetar bursts ( @xmath142 , * ? ? ?
* ) and other avalanche events explained by self - organized criticality @xcite . * burst energy can reach @xmath143 , which is supported by the `` lorimer '' burst with @xmath144 @xcite . *
the spacial density of frb progenitors is @xmath145 where we have used @xmath146 to get the numerical values .
although @xmath5 still has large uncertainties , the fact the frb 121102 is consistent with being representative means that @xmath5 is likely not much smaller than @xmath147 .
if we want to use bnss to probe the density fluctuations of igm @xcite , the spacial resolution may be @xmath148 .
* the fraction of nss that are able to produce observable frbs is @xmath149 . if we consider the possibility that @xmath150 , the fraction could be as low as @xmath151 !
although the true fraction is a factor of @xmath152 higher ( @xmath35 is the total sky coverage of all bursts from one progenitor ) , frb progenitors may be a rare type of nss .
for example , the birth rate of magnetars is @xmath153 of core - collapse rate @xcite , so `` being a magnetar '' may not be sufficient for frbs .
this research was funded by the named continuing fellowship at the university of texas at austin . | assuming : fast radio bursts ( frbs ) are produced by neutron stars at cosmological distances ; frb rate tracks the core - collapse supernova rate ; and all frbs repeat with a universal energy distribution function ( edf ) @xmath0 with a cutoff at burst energy @xmath1 .
we find that observations so far are consistent with a universal edf with @xmath2 , high - end cutoff @xmath3 and normalization @xmath4 ; where @xmath5 is the integrated rate above the reference energy @xmath6 ( @xmath7 is the radio emission efficiency ) .
implications of such an edf are discussed .
[ firstpage ] radio continuum : general stars : neutron |
the author wishes to thank the participants of the 4th patras workshop on axions , wimps and wisps " and the brainstorming and calculationshop on the physics case for a low energy frontier " for very stimulating discussions in a really friendly ambient .
f. iocco , g. mangano , g. miele , o. pisanti , and p. d. serpico , http://arxiv.org/abs/0809.0631[arxiv:0809.0631 [ astro - ph ] ] . v. simha and g. steigman , http://dx.doi.org/10.1088/1475-7516/2008/06/016[_jcap_ * 0806 * ( 2008 ) 016 ] , http://arxiv.org/abs/0803.3465[arxiv:0803.3465 [ astro - ph ] ] .
s. davidson , s. hannestad , and g. raffelt , _ jhep _ * 05 * ( 2000 ) 003 , http://arxiv.org/abs/hep-ph/0001179[hep-ph/0001179 ] .
z. berezhiani and a. lepidi , http://arxiv.org/abs/0810.1317[arxiv:0810.1317 [ hep - ph ] ] .
e. masso and j. redondo , _ phys .
* 97 * ( 2006 ) 151802 , http://arxiv.org/abs/hep-ph/0606163[hep-ph/0606163 ] .
j. redondo , _ can the pvlas particle be compatible with the astrophysical bounds ?
_ phd thesis , 2008 .
http://arxiv.org/abs/0807.4329[arxiv:0807.4329 [ hep - ph ] ] .
d. j. fixsen _
et al . _ , http://dx.doi.org/10.1086/178173 [ _ astrophys . j. _ * 473 * ( 1996 ) 576 ] , http://arxiv.org/abs/astro-ph/9605054[arxiv:astro-ph/9605054 ]
a. melchiorri , a. polosa , and a. strumia , http://dx.doi.org/10.1016/j.physletb.2007.05.042[_phys .
* b650 * ( 2007 ) 416420 ] , http://arxiv.org/abs/hep-ph/0703144[arxiv:hep-ph/0703144 ] . j. jaeckel , j. redondo , and a. ringwald , _ phys .
_ * 101 * ( 2008 ) 131801 , http://arxiv.org/abs/0804.4157[arxiv:0804.4157 [ astro - ph ] ] .
k. ichikawa , t. sekiguchi , and t. takahashi , http://arxiv.org/abs/0803.0889[arxiv:0803.0889 [ astro - ph ] ] .
s. hannestad , a. mirizzi , g. g. raffelt , and y. y. y. wong , http://dx.doi.org/10.1088/1475-7516/2007/08/015[_jcap_ * 0708 * ( 2007 ) 015 ] , http://arxiv.org/abs/0706.4198[arxiv:0706.4198 [ astro - ph ] ] . s. hannestad , a. mirizzi , g. g. raffelt , and y. y. y. wong , http://dx.doi.org/10.1088/1475-7516/2008/04/019[_jcap_ * 0804 * ( 2008 ) 019 ] , http://arxiv.org/abs/0803.1585[arxiv:0803.1585 [ astro - ph ] ] .
u. seljak , a. slosar , and p. mcdonald , http://dx.doi.org/10.1088/1475-7516/2008/07/008[_jcap_ * 0610 * ( 2006 ) 014 ] , http://arxiv.org/abs/astro-ph/0604335[arxiv:astro-ph/0604335 ] .
g. mangano , a. melchiorri , o. mena , g. miele , and a. slosar , _ jcap _ * 0703 * ( 2007 ) 006 , http://arxiv.org/abs/astro-ph/0612150[arxiv:astro-ph/0612150 ] .
j. hamann , s. hannestad , g. g. raffelt , and y. y. y. wong , http://dx.doi.org/10.1088/1475-7516/2007/08/021[_jcap_ * 0708 * ( 2007 ) 021 ] , http://arxiv.org/abs/0705.0440[arxiv:0705.0440 [ astro - ph ] ] . j. jaeckel and a. ringwald , http://dx.doi.org/10.1016/j.physletb.2007.11.071[_phys .
_ * b659 * ( 2008 ) 509514 ] , http://arxiv.org/abs/0707.2063[arxiv:0707.2063 [ hep - ph ] ] .
j. jaeckel , http://arxiv.org/abs/0809.3112[arxiv:0809.3112 [ hep - ph ] ] .
j. jaeckel , http://arxiv.org/abs/0807.5097[arxiv:0807.5097 [ hep - ph ] ] .
j. jaeckel and j. redondo , http://arxiv.org/abs/0806.1115[arxiv:0806.1115 [ hep - ph ] ] .
g. g. raffelt , _ stars as laboratories for fundamental physics : the astrophysics of neutrinos , axions , and other weakly interacting particles_. the university of chicago press , 1996 .
g. g. raffelt , http://dx.doi.org/10.1007/978-3-540-73518-2_3[_lect .
notes phys . _
* 741 * ( 2008 ) 5171 ] , http://arxiv.org/abs/hep-ph/0611350[arxiv:hep-ph/0611350 ] .
g. g. raffelt , _ phys .
* d33 * ( 1986 ) 897 .
g. g. raffelt and d. s. p. dearborn , _ phys .
rev . _ * d36 * ( 1987 ) 2211
. g. g. raffelt , http://dx.doi.org/10.1016/0370-2693(86)91588-1[_phys .
lett . _ * b166 * ( 1986 ) 402 ] .
j. isern , e. garcia - berro , s. torres , and s. catalan , http://arxiv.org/abs/0806.2807[arxiv:0806.2807 [ astro - ph ] ] .
h. schlattl , a. weiss , and g. raffelt , http://dx.doi.org/10.1016/s0927-6505(98)00063-2[_astropart .
* 10 * ( 1999 ) 353359 ] , http://arxiv.org/abs/hep-ph/9807476[arxiv:hep-ph/9807476 ] .
p. gondolo and g. raffelt , http://arxiv.org/abs/0807.2926[arxiv:0807.2926 [ astro - ph ] ] .
p. sikivie , _ phys .
lett . _ * 51 * ( 1983 ) 1415
. [ erratum - ibid . * 52 * , 695 ( 1984 ) ] .
k. van bibber , p. m. mcintyre , d. e. morris , and g. g. raffelt , _ phys .
* d39 * ( 1989 ) 2089 .
e. a. paschos and k. zioutas , http://dx.doi.org/10.1016/0370-2693(94)91233-5[_phys .
* b323 * ( 1994 ) 367372 ] .
d. m. lazarus _
_ , _ phys .
* 69 * ( 1992 ) 23332336 .
s. moriyama _ et al .
_ , _ phys .
lett . _ * b434 * ( 1998 ) 147 , http://arxiv.org/abs/hep-ex/9805026[hep-ex/9805026 ] . y. inoue _ et al . _ , http://dx.doi.org/10.1016/j.physletb.2008.08.020[_phys
* b668 * ( 2008 ) 9397 ] , http://arxiv.org/abs/0806.2230[arxiv:0806.2230 [ astro - ph ] ] .
collaboration , k. zioutas _ et al .
_ , _ phys .
* 94 * ( 2005 ) 121301 , http://arxiv.org/abs/hep-ex/0411033[hep-ex/0411033 ] .
collaboration , s. andriamonje _ et al .
_ , _ jcap _ * 0704 * ( 2007 ) 010 , http://arxiv.org/abs/hep-ex/0702006[hep-ex/0702006 ] .
j. redondo , http://dx.doi.org/10.1088/1475-7516/2008/07/008[_jcap_ * 0807 * ( 2008 ) 008 ] , http://arxiv.org/abs/0801.1527[arxiv:0801.1527 [ hep - ph ] ] .
s. n. gninenko and j. redondo , http://dx.doi.org/10.1016/j.physletb.2008.05.020[_phys .
lett . _ * b664 * ( 2008 ) 180184 ] , http://arxiv.org/abs/0804.3736[arxiv:0804.3736 [ hep - ex ] ] .
collaboration , e. zavattini _ et al . _ , http://dx.doi.org/10.1103/physrevd.77.032006[_phys .
_ * d77 * ( 2008 ) 032006 ] , http://arxiv.org/abs/0706.3419[arxiv:0706.3419 [ hep - ex ] ] .
collaboration , e. zavattini _ et al .
_ , _ phys .
* 96 * ( 2006 ) 110406 , http://arxiv.org/abs/hep-ex/0507107[hep-ex/0507107 ] .
e. masso and j. redondo , _ jcap _ * 0509 * ( 2005 ) 015 , http://arxiv.org/abs/hep-ph/0504202[hep-ph/0504202 ] .
p. brax , c. van de bruck , and a .- c .
davis , _ phys .
_ * 99 * ( 2007 ) 121103 , http://arxiv.org/abs/hep-ph/0703243[hep-ph/0703243 ] .
r. n. mohapatra and s. nasri , http://dx.doi.org/10.1103/physrevlett.98.050402[_phys .
* 98 * ( 2007 ) 050402 ] , http://arxiv.org/abs/hep-ph/0610068[arxiv:hep-ph/0610068 ] .
j. jaeckel , e. masso , j. redondo , a. ringwald , and f. takahashi , _ phys .
_ * d75 * ( 2007 ) 013004 , http://arxiv.org/abs/hep-ph/0610203[hep-ph/0610203 ] .
m. ahlers , j. jaeckel , j. redondo , and a. ringwald , http://dx.doi.org/10.1103/physrevd.78.075005[_phys .
rev . _ * d78 * ( 2008 ) 075005 ] , http://arxiv.org/abs/0807.4143[arxiv:0807.4143 [ hep - ph ] ] . | many weakly interacting sub - electronvolt particles ( wisps ) are easily accommodated in extensions of the standard model .
generally the strongest bounds on their existence come from stellar evolution and cosmology , where to the best of our knowledge observations seem to agree with the standard budget of particles . in this talk
i review the most demanding constraints for axions and axion - like - particles , hidden photons and mini - charged particles .
there is little doubt in the particle physics community about the need of complementing the already very successful standard model ( sm ) to pursue a completely satisfactory final theory of elementary particles . on the other hand , and with the exception of the dark matter
, our increasingly precise knowledge of the universe shows no trace of physics beyond the sm .
if new light particles exist they should be very weakly interacting , probably only accessible to extremely precise experiments .
experiments such as the ones presented in this conference .
astrophysics and cosmology are often strong probes of weakly interacting particles .
the reason is clear : the huge magnitudes of the typical sizes , time scales , densities or temperatures in the early universe or in stars can convert a tiny microscopic " effect in a big qualitative change in the evolution of the whole system .
this conclusion is specially emphasized when we note that the only weakly interacting sub - ev particles ( wisps ) in the standard model are neutrinos , whose production cross sections are strongly energy - dependent and therefore their role is increasingly inhibited as temperatures drop below the electroweak scale .
thus , in an non - extreme range of temperatures the early universe and stellar plasmas are very opaque to standard particles and wisps can be the most efficient way of energy transfer . whenever such an anomalous energy transfer has an observable implication we can derive strong constraints on the wisp interactions with the standard particles constituting the relevant plasma .
the oldest picture of the universe we have is a dense and hot plasma of elementary particles that expanded against gravity . as this plasma cooled down ,
the three long range forces clustered the particles into the structures which nowadays are found : the color force first confined quarks into protons and neutrons and later merged them into light nuclei ( at bbn ) , the coulomb force combined them with electrons into atoms ( releasing the cmb ) that gravity finally clustered into galaxies , then into clusters , etc ... after the first galaxies formed , the conditions for stars to be born were settled . during all these steps of structure formation ( in a broad sense )
the role of wisps can be constrained .
let us start this review in chronological order .
summary plots on the reviewed bounds are shown at the end of this contribution . _
big bang nucleosynthesis.- _ bbn left an invaluable probe of the early universe environment imprinted in today s observable light nuclei abundances @xcite .
below @xmath0 mev the weak reactions @xmath1 became ineffective , fixing the neutron / proton density ratio to @xmath2 .
all particles present contribute to the energy density @xmath3 which determines the speed of the cosmic expansion @xmath4 and the freeze - out " ratio @xmath5 in turn .
the larger @xmath6 the sooner the @xmath7-@xmath8 freezing and the higher @xmath5 .
later , all neutrons are confined into @xmath9he nuclei whose primordial abundance can be measured today , leading to a bound on the non - standard energy density @xmath10 during bbn , usually expressed as an effective number of thermal neutrino species , @xmath11 @xcite , where we assumed three standard neutrinos . therefore , while a spin - zero particle thermalized during bbn is allowed , this is not the case for other wisps like a mini - charged particle ( mcp ) ( @xmath12 ) or a massive hidden photon @xmath13 ( @xmath14 ) .
the interactions of mcps and @xmath13s with the standard bath should not allow thermalization before bbn .
mcps @xmath15 are produced with a rate @xmath16 ( with @xmath17 the mcp electric charge ) while @xmath13s with @xmath18 with @xmath19 the standard compton scattering rate . here
@xmath20 is the effective @xmath21 mixing in the plasma , which for sub - ev @xmath13 masses is @xmath22 .
the ratio of the @xmath13 mass to the plasma frequency @xmath23 is extremely small before bbn so it suppresses @xmath13 production with respect to other wisps . comparing with the expansion rate @xmath6
we find that mcps with @xmath24 would be allowed @xcite but there are no significant bounds for hidden photons @xcite .
_ cosmic microwave background.- _ the today s measured cmb features an almost perfect blackbody spectrum with @xmath25 angular anisotropies .
it is released at @xmath26 ev but the reactions responsible of the blackbody shape freeze out much earlier , at @xmath27 kev .
reactions like @xmath28wisp@xmath29 will deplete photons in a frequency dependent way , which can be constrained by the precise firas spectrum measurements @xcite .
this has been used to constrain light mcps @xcite and hps with @xmath30 mev @xcite . on the other hand , around @xmath27
ev the primordial plasma is so sparse that wisps would free - stream out of the density fluctuations , diminishing their contrast .
moreover , thermal wisps contribute to the _ radiation _ energy density , delaying the matter - radiation equality and reducing the contrast growth before decoupling . in these matters
they act as standard neutrinos @xcite so @xmath10 ( and the couplings that would produce it ) can again be constrained from the value of @xmath31 inferred from analysis of cmb anisotropies and other with other cosmological parameters such as the dark matter density . ]
large scale structure ( lss ) data @xmath32 @xcite .
this argument has been used to constraint axions @xcite and mev @xmath13s @xcite . in this
bound ly-@xmath33 forest data has been deliberately omitted .
ly-@xmath33 has systematically favored values of @xmath34 larger than zero @xcite which could be revealing the existence of a cosmic wisp relic density .
if this anomaly is due to a population of @xmath13s created through resonant oscillations @xmath35 between bbn and the cmb decoupling it can be tested in the near future by new laboratory experiments such as alps at desy @xcite . _ bounds from stellar evolution.- _ the production of wisps in stellar interiors can substantially affect stellar evolution @xcite .
wisps can be only scarcely produced in the dense plasmas of stellar interiors , but they will easily leave the star contributing directly to its overall luminosity . on the other hand , only photons of the photosphere ( or neutrinos ) contribute to the standard energy loss .
therefore , the wisp luminosity is enhanced at least by a volume / surface factor and a further @xmath36 ( @xmath37 a relevant particle density , @xmath38 ) with respect to the standard luminosity .
this can be a huge enhancement which certainly justifies the typical strong constraints .
stars evolve fusing increasingly heavier nuclei in their cores .
heavier nuclei require hotter environments , and when a nuclear species is exhausted in the core this slowly contracts and heatens up until it reaches the new burning phase .
wisp emission shortens normal burning phases ( the energy loss rate is higher than standard but the total energy is limited by the number of nuclei ) but enlarges the intermediate ( red giant ) phases ( wisp cooling delays reaching the appropriate temperature during the core contraction ) . these effects have been used to constraint a variety of wisps in different stellar environments @xcite for which information on evolutionary time scales is available .
the strongest limits for general axion - like - particles ( alps ) with a two photon coupling and mcps come from observations of horizontal branch ( hb ) stars in globular clusters ( gc ) @xcite .
for the standard qcd axions , the best constraints come from white dwarf cooling @xcite through the coupling to electrons ( dfsv axions ) and from the duration of the sn1987a neutrino burst @xcite through the nucleon coupling ( ksvz axions ) .
the sun is less sensitive than these other stars to wisp emission , even though its properties are better known .
solar bounds have been obtained from studies of its lifetime , helioseismology and the neutrino flux @xcite , but although more precise they are also less demanding . nevertheless , if wisps are emitted from the sun one can detect them with a dedicated laboratory experiment at earth @xcite .
one of the so - called helioscope axion searches @xcite , cast , has recently beaten the hb constraints for alps with a two photon coupling @xcite , and its results have been used to limit a possible solar @xmath13 flux @xcite . following the now disclaimed @xcite pvlas 2005 results @xcite , specific models were recently built that suppress wisp emission from stars
@xcite .
if this idea is realized , helioscope bounds will gain terrain to energy loss arguments @xcite ( @xmath13s are the minimal example of this case @xcite ) . in summary , cosmology and astrophysics
provide the strongest constraints on the ( minimal ) wisp models described elsewhere in these proceedings , with the only exception of sub - mev @xmath13s .
summary plots are shown in figs . 1 and 2 . or decay constant @xmath39 ) @xcite and axion - like - particles ( down ) ( two photon coupling @xmath40 vs. mass @xmath41 ) @xcite .
see the text for details . for comparision , the most notable laboratory limits are also shown . ] or decay constant @xmath39 ) @xcite and axion - like - particles ( down ) ( two photon coupling @xmath40 vs. mass @xmath41 ) @xcite .
see the text for details . for comparision , the most notable laboratory limits are also shown . ] vs. mass @xmath42 ) @xcite and minicharged particles ( down ) ( charge @xmath43 vd .
mass @xmath44 ) @xcite .
see the text for details .
for comparision , some laboratory limits are also shown . ] vs. mass @xmath42 ) @xcite and minicharged particles ( down ) ( charge @xmath43 vd .
mass @xmath44 ) @xcite .
see the text for details . for comparision , some laboratory limits are also shown . ] |
the search for efficient , high - power , inexpensive , compact , and room - temperature methods of generation of coherent teraherz ( thz ) radiation is one of the main topics in modern optoelectronics and photonics @xcite .
its importance is based on the fact that thz radiation has unique potential for a wide range of applications from diagnostics of different materials ( including semiconductors , chemical compounds , biomolecules , and biotissues ) , imaging ( for medical and security purposes ) , to remote atmospheric sensing and monitoring , astronomy , etc . @xcite . in previous works
we have focussed on a new approach to the problem of generation of short coherent thz pulsed radiation by taking advantage of dramatic enhancement of the nonlinear response of a medium via maximal quantum coherence @xcite , created in atomic and molecular gases with different level configurations ( double-@xmath1 , v-@xmath1 , double - v schemes ) by coherent laser radiaiton . as is well - known , quantum coherence can dramatically change the optical properties of media .
for example , lasing without population inversion ( lwi ) @xcite and electromagnetically induced transparency ( eit ) @xcite have been demonstrated @xcite in coherently driven media .
these unusual properties are employed for generation of electromagnetic radiation of different frequencies ranging from ir @xcite to uv @xcite , and gamma - rays @xcite .
enhancement of coherent raman scattering via maximal coherence has been demonstrated experimentally @xcite , and it has applications to enhanced real time spectroscopy ( fast cars ) @xcite .
previously , the double @xmath1 scheme with near - maximal raman coherence was used for highly efficient conversion of blue to ultraviolet light in @xmath2 vapor @xcite .
a new type of gas - phase optical parametric oscillator @xcite was suggested for frequency downconversion to generate 1.88 @xmath3 m radiation in pb vapor , and efficient infrared upconversion @xcite was suggested to convert infrared light with a 100 @xmath3 m wavelength to the visible .
the first experimental demonstration supporting these ideas was performed in @xcite where 5 @xmath3 m ir radiaiton was generated in @xmath4 atomic vapor .
many results in this area have been obtained in gases , but , recently , eit has been generalized and extended to solids with a long - lived spin coherence @xcite , a class of solid materials , namely , rare - earth and transition metal ion doped dielectrics have been suggested as very attractive from the point of view of realization and applications of eit .
several applications of eit to improve performance of solid state lasers have been already proposed @xcite .
lwi , first realized in a gas medium @xcite , recently has been demonstrated in solids as well @xcite . in this paper
we extend the approach of generation of thz radiation via resonantly induced coherence to solid state media .
the coherence at the corresponding transition can be induced by coherent optical fields as shown in fig .
[ fig : ruby - levels ] .
electronic levels of the thz transition are coupled by a pair of optical fields with rabi frequencies @xmath5 and @xmath6 to a common ground state forming a v system of energy levels .
coherence @xmath7 induced by the optical fields at the thz transition gives rise to polarization provided that the corresponding dipole moment is not zero , which will radiate out a thz pulse with rabi frequency @xmath8 .
this method differs from the one proposed in @xcite for gaseous media where the optically prepared thz coherence produces no macroscopic polarization of gaseous media at a thz transition ( this is the manifestation of symmetry with respect to inversion ) , the reason is that an electrically allowed dipole moment at the two - photon transition is zero since it is typically forbidden ( although , it is important to note here that the magnetic dipole moment can be non - zero even in gases , we discuss this opportunity later ) . for doped solids having sites with no inversion symmetry is common , such as the one occupied by cr@xmath9 ( simplified energy level structure is shown in fig.([fig : ruby - levels]a ) ) in ruby , the corresponding two - photon transition dipole moment is not zero , resulting in all three transitions in the @xmath0 scheme being allowed .
it is worth mentioning that using nonlinear optical mixing in a semiconductor heterostructure for generation of few - cycle thz pulses was proposed in @xcite .
in ruby closely spaced r@xmath10 and r@xmath11 lines arise from transitions between the ground state of the cr@xmath9 ion ( @xmath12a@xmath11 ) and its first excited state ( @xmath13e ) .
each of these cubic - field states is split by the trigonal crystal field and spin - orbit coupling into a pair of kramers doublets , 0.38 @xmath14 apart in the ground state and 29.14 @xmath14 apart in the @xmath13e state .
the width of the r lines is about 11 @xmath14 at room temperature and @xmath15 @xmath14 at the temperature of liquid nitrogen .
the transitions from the @xmath12a@xmath11 ground state to levels @xmath13e are spin - forbidden , and have oscillator strength @xmath16 ( r@xmath10 , r@xmath11 line peak cross - section is @xmath17 cm@xmath13 at the temperature of liquid nitrogen and @xmath19 cm@xmath13 at room temperature ) .
they are predominantly electric - dipole in nature , since the inversion symmetry of a cubic crystal field is broken by crystal field trigonal distortions and odd - parity lattice vibrations .
namely , the odd - parity component of the crystal field at the cr@xmath9 ion mixes odd - parity states of high energy into the even parity d states between which r - transitions are observed and makes them weakly allowed . the magnetic - dipole contribution to the r - lines
is estimated to be @xmath20 of the strength of the elecric - dipole transitions @xcite .
the 29 @xmath14 transition in ruby has history which is as rich as the ruby laser itself .
it was considered promising for realization of a quantum counter for far - infrared radiation ( fir ) @xcite on one hand , and as a system very convenient to study interaction processes of non - equilibrium phonons with two - level electronic systems ( so - called phonon spectrometer ) , on the other @xcite .
there was also a proposal to make a far - infrared laser at this transition pumping via the r@xmath11 line , but due to the unfavorable ratio of relaxation rates ( fast relaxation at the fir and slow at the optical r@xmath10 transition ) , the laser could operate only at liquid helium temperature and the estimated gain did not exceed losses @xcite .
it is interesting that 29 @xmath14 fir was produced by non - linear mixing of two ruby laser beams , one emitting at r@xmath10 and another at r@xmath11 , in linbo@xmath21 @xcite and in znte crystals @xcite , with the efficiency @xmath22 .
the method proposed in the present work would allow pulsed thz radiation to be easily produced from a dual - color ruby laser itself , operating at both r@xmath10 and r@xmath11 lines , thus significantly miniaturizing the system .
the outline of the paper is the following : in the next section we describe a theoretical model used to calculate parameters of generated thz pulses . in section iii
we consider possibilities of phase - matching for fields in different regimes ( cw and short pulses ) . in section
iv we make estimates showing the perspective of the suggested scheme as well as discuss a proposed experiment with a standard femto - second laser setup .
next we show that the system we have considered is not limited , i.e. there are many systems to which our theory can be applied .
finally we compare the efficiency of our scheme with the methods of generation of thz radiation used nowadays and possible applications that are opened due to the high efficiency of the proposed technique .
the ruby system shown in fig.([fig : ruby - levels]a ) includes eight states , four in the ground @xmath23 and four in the excited @xmath24 one . in our theoretical analysis
we describe it using only three levels as in fig.([fig : ruby - levels]b ) , since this simple model takes into account all essential physics and allows us to make estimates of expected thz radiation characteristics and required optical field parameters . for the proposed technique to work
we need to make sure that an efficient @xmath0 scheme can be realized . in the case of a single fs pulse ,
driving both optical transitions simultaneously , we will have only one polarization , for example , linear . for two pulses we can choose each pulse polarization separately .
selection rules for the right circular and left circular polarizations are shown in fig.([fig : sel - rules]a ) @xcite .
as one can see , neither for the right , nor for the left circular polarization a @xmath0 scheme involving r@xmath10 and r@xmath11 transitions can be organized . for a linear polarization perpendicular to the optic axis , which is an equal sum or difference of the right and left ones , there are four @xmath0 schemes possible , shown in fig.([fig : sel - rules]b ) .
taking into account the relation between the coefficients c@xmath25 and c@xmath26 : @xmath27 , one can see that the schemes a and c cancel each other , since the products of the matrix elements are opposite ( @xmath28 for a and c , respectively ) . on the other hand , for schemes b and d
the products are the same ( @xmath29 ) .
these two @xmath0 schemes will interfere constructively and induce the thz coherence .
the interaction hamiltonian for the system shown in fig.([fig : ruby - levels]b ) is given by @xmath30\ ] ] @xmath31,\ ] ] where @xmath32 is the rabi frequency of the respective field , @xmath33 and @xmath34 ; @xmath35 and @xmath36 are the electric dipole matrix elements of the optical transitions @xmath37 and @xmath38 , and @xmath39 of the thz transition @xmath40 , respectively ; @xmath41 , @xmath42 and @xmath43 are the frequencies of the electronic and thz transitions ; @xmath44 is the amplitude of the respective electromagnetic field .
the time - dependent density matrix equations are @xmath45-\frac{1}{2}(\gamma \rho + \rho \gamma),\ ] ] where @xmath46 is the relaxation matrix , and @xmath47 is the total hamiltonian . to estimate the efficiency of thz radiation generation and required laser field intensities we first consider the system ( [ eq : den - matr ] ) in the limit of ultrashort pulses , which are short compared to relaxation times of the system , and neglect propagation effects for the optical fields .
in this simplified picture we assume that the system is driven by two resonant fields with equal time - dependent rabi - frequencies @xmath48 ( assumed real ) . in the case of a v - scheme considered in this work
the stirap technique , proposed in @xcite for maximal molecular coherence excitation in a @xmath1 energy system , is not applicable . in the case of the v scheme maximal coherence
is excited by two fields of equal rabi frequency and duration , acting simultaneouly .
the system is then described by the following set of equations @xmath49
@xmath50 @xmath51 @xmath52 @xmath53 @xmath54 due to the symmetry of the model system @xmath55 , @xmath56 and @xmath57 - real .
let us introduce new variables @xmath58 @xmath59 for which , taking into account ( [ eq : sigcb ] ) and ( [ eq : rhoaa ] ) , the following relation can be obtained @xmath60 giving @xmath61 . for the new variables we arrive at the following system of equations @xmath62 @xmath63 which can be further simplified
if we substitute @xmath64 @xmath65 @xmath66 the solution of this system is @xmath67 @xmath68 it follows from eq.([eq : beta - sol ] ) that maximal coherence @xmath69 will be excited by a pair of pulses with area @xmath70 each .
taking pulses of a gaussian shape @xmath71 with the area of the pulse @xmath72 , we will have the maximal coherence excited when @xmath73 .
the excited coherence will give rise to the polarization which will radiate out a coherent thz pulse .
it is worth noting that this mechanism is similar to free - induction decay ( fid ) @xcite in photon echo : excitation of a maximal coherence in a two - level system by a @xmath74 pulse , followed by emission of a coherent pulse .
the difference is that typically in solids the fid decay is governed by the dephasing due to inhomogeneous broadening of the transition leading to decay of the corresponding coherence during the time @xmath75 . in the case of ruby
the thz transition is homogeneously broadened even at liquid helium temperature , so the decay of the thz coherence is determined by the homogeneous width of the corresponding transition . for two long pulses , which can be considered as cw ( in the sense that optical and thz coherences approach the steady - state ) , with approximately equal rabi frequencies @xmath76 , resonant with the corresponding optical transitions
, we get the thz coherence @xmath77 where @xmath78 and @xmath79 are the optical and thz coherence decay rates , respectively . in this case
the maximal thz coherence @xmath80 will be excited provided that @xmath81 , which is a usual steady - state eit threshold condition .
let us now estimate the peak intensity and energy of the optical pulses necessary to excite required coherence .
the peak intensity of the pulse is expressed in terms of the peak rabi - frequency as @xmath82 where @xmath83 is the width of an optical transition , @xmath84 is the refractive index at the optical wavelength , @xmath85 is the optical wavelength , and @xmath86 is the corresponding absorption cross - section . the energy of a single pulse can be calculated as @xmath87 where @xmath88 is the laser beam cross - section .
a complete self - consistent calculation also includes field propagation equations @xmath89 where index @xmath90 indicates all fields and corresponding polarizations , @xmath91 is the corresponding coupling constant , @xmath92 are the frequencies of the optical and thz fields , @xmath93 is the density of the medium , @xmath94 is the speed of light in vacuum , @xmath95 is the corresponding refractive index , and @xmath96 are losses for the field during propagation in the crystal due to scattering , diffraction , or non - resonant absorption . for optical fields these losses
are usually small , but for the thz field in free space the diffraction losses given by @xmath97 should be taken into account . to avoid diffraction losses
the crystal can be placed in a waveguide for thz radiation , then the distribution of the field mode should be taken into account .
propagation effects are important for nonlinear interactions such as wave mixing .
let us consider two regimes : cw , when the spectral widths of pulses is smaller than the splitting between the levels of the thz transition , and pulsed one , when the spectral width of pulses exceeds the splitting between the levels . in the cw regime
the field @xmath8 at the output of the crystal is given by ( assuming that the thz pulse propagates in @xmath98 direction ) @xmath99 @xmath100 if losses are neglected .
here @xmath101 is the length of the crystal , and the prefactor describes phase - matching ( @xmath102 ) .
the direction of thz propagation is given by the phase - matching condition ( see fig.3 ) @xmath103 to satisfy the phase - matching condition @xmath104 for the thz pulse to propagate in @xmath98 direction ( along the crystal ) the angle between the optical fields has to be adjusted according to the equantion @xmath105 rewriting a condition for optical and thz frequencies @xmath106 in the form of @xmath107 , we obtain @xmath108 where @xmath109 is the wavelength of the optical ( thz ) field . for the case of ruby the refractive index for thz radiation is twice larger than the one for optics , @xmath110 @xcite , then the angle between directions of propagation for optical fields @xmath111 .
the angle between the @xmath112 vectors and the thz field propagation direction is @xmath113 in this case .
this provides an interesting opportunity to design a dual - wavelength ruby laser working simultaneously at @xmath114 and @xmath115 lines that can also generate thz radiation , as is shown in fig.[fig : phase - matching]c .
the crystal will simultaneously serve as a thz waveguide in this case .
now we turn to a short pulse excitation regime . for many femtosecond laser experimental setups
it would be an interesting extension to include an option of thz generation . the phase - matching in this case
can be obtained by excitation of thz coherence with two optical beams with the same frequencies but different directions of propagation .
it is instructive to show how it works in this case .
let us consider the problem in the simplest case of lowest order of perturbation in coherent pumping fields .
the density matrix equations look like @xmath116 @xmath117 @xmath118 a solution of this set of equations is given by @xmath119 one can see that the phase - matching between the optical fields and the thz field is achieved by appropriate tilting of the optical beams at the angle @xmath120 ; in particularly , for ruby @xmath121 .
first we analize the optimal conditions at which the maximal thz coherence can be induced in ruby by the optical fields at both room and liquid helium temperatures . for room temperature ruby
@xmath122 @xmath14 @xcite , @xmath123 , @xmath124 nm , and @xmath125 cm@xmath13 @xcite . in the case of ultrashort driving pulses ( 100 fs-1 ps ) the shortest thz pulse that can be generated in this system has a duration @xmath126 ps . with the optical pulses of the same duration @xmath127 ps the rabi frequency @xmath128 s@xmath129 is required to produce the maximal coherence , corresponding to @xmath130 w / cm@xmath13 , which is still below a damage threshold of ruby for such duration of the pulses @xcite , but we make a conservative estimate for lower intensities @xmath131 gw / cm@xmath13 , certainly below the threshold , corresponding to the peak rabi frequency @xmath132 s@xmath129 . for long driving pulses ( 10 ps -1 ns ) to excite the maximal coherence
@xmath80 would require @xmath133 resulting in the rabi frequency @xmath134 s@xmath129 , corresponding to intensities exceeding the ruby damage threshold for pulses of such duration . for long pulses ( 1 - 10 ns )
the damage threshold intensity is in the range @xmath135 gw / cm@xmath13 @xcite , which corresponds to the rabi frequency @xmath136 s@xmath129 .
this value is used in the estimates below . at low temperatures ( @xmath137 k )
the thz coherence lifetime is significantly larger , @xmath138 ps @xcite , which means that with ultrashort pumping pulses ( in this case of 10 ps-1 ns duration ) , the maximal coherence will be excited with only @xmath139 s@xmath129 , requiring @xmath131 mw / cm@xmath13 pulse intensity . for long pulses ( 1 - 10 ns )
the maximal coherence will be excited when @xmath140 , and since at low temperatures @xmath141 @xmath14 @xcite , it results in @xmath142 s@xmath129 , requiring @xmath143 mw / cm@xmath13 .
the next question is the energy of the generated thz pulse , which can be estimated from the propagation equation @xmath144 where @xmath93 is the density of @xmath145 ions , @xmath146 @xmath14 ( @xmath147 thz ) , the dipole moment of the far - infrared transition can be calculated using measured experimentally low - temperature ( lt ) parameters @xmath148 , and @xmath149 is the non - resonant absorption coefficient of the sapphire host at the thz wavelength . assuming that the crystal sample is side - pumped homogeneously , so that the thz coherence @xmath57 does not depend on the propagation coordinate @xmath98 , and phase - matching conditions are satisfied we arrive at the equation @xmath150 where @xmath151 , so that at the end of the crystal the rabi - frequency of the thz pulse is @xmath152 where @xmath153 . for the ultrashort pumping pulses we can model @xmath154 at @xmath155 and @xmath156 at @xmath157 for an estimate .
eq.([eq : thz - rabi - fr ] ) gives @xmath158 for @xmath159 and @xmath160 for @xmath161 .
eqs.([eq : thz - rabi - short1])-([eq : thz - rabi - short2 ] ) then allow one to calculate the energy of the thz pulse @xmath162 @xmath163 where @xmath164 is the crystal cross - section along thz pulse propagation direction . for two pulses of 1 ps duration with @xmath131 gw / cm@xmath13 the maximal induced coherence is @xmath165 . taking @xmath166 @xmath167 for 0.05@xmath168 doped ruby , and known from low - temperature measurements @xmath169 s@xmath129 , @xmath170 cm@xmath13 @xcite , and room temperature parameters @xmath171 s@xmath129 and @xmath172 @xmath14 @xcite , @xmath173 @xcite , and considering the crystal of 1 @xmath1741 @xmath1740.1 cm size with the @xmath175 cm @xmath176 0.1 cm thz emitting cross - section and @xmath177 cm length , we arrive at @xmath178 pj . the energy can be higher for higher optical peak intensities , approaching several nj for @xmath130 w / cm@xmath13 . for low temperatures ( @xmath137 k )
the parameters are : @xmath179 s@xmath129 , @xmath180 cm@xmath13 , @xmath181 @xmath14 @xcite , and the maximal coherence @xmath69 can be easily excited . due to stronger absorption at low temperatures ( @xmath182 cm@xmath13 @xcite )
, half of the optical intensity will be absorbed at @xmath183 mm for @xmath184 cr@xmath9 density , therefore , for this estimate we take a crystal with dimensions 1 @xmath1740.1 @xmath1740.05 cm size with the thz emitting cross - section @xmath185 @xmath174 0.05 cm and @xmath177 cm length which results in @xmath186 @xmath3j .
let us also make an estimate for fs pulse duration . since powerful fs ti
: sapphire lasers are readily available in many laboratories nowadays this is of interest . in this case
the coherence can be excited by just one pulse of 100 fs duration , since its spectral width is larger than the thz transition splitting .
although , to satisfy phase - matching conditions , it is necessary to split it in two beams hiting the crystal at slightly different angles .
keeping the same peak intensity @xmath187 gw / cm@xmath13 we obtain a smaller pulse area @xmath188 , giving the coherence @xmath189 and resulting in thz pulse energy of @xmath190 fj at room temperature . for the cw
pumping case we have , in analogy with the short pulse pumping , @xmath191 where from eq.([eq : cw - coherence ] ) @xmath192 and @xmath193 . in the limit @xmath194 the maximal coherence is excited @xmath195 , which would require , though , experimentally unfeasible rabi frequency @xmath134 s@xmath129 at room temperature , resulting in @xmath130 w / cm@xmath13 . for such long pulses ( 10 ps -1 ns )
the damage threshold intensity for ruby is @xmath196 gw / cm@xmath13 , which corresponds to the rabi frequency @xmath197 s@xmath129 . for this rabi frequency and room temperature decay rates
@xmath198 s@xmath129 , the ratio @xmath199 and the resulting thz coherence @xmath200 for a 10 ps pulse
. the thz pulse energy will be @xmath201 pj . at low temperatures
the maximal coherence @xmath195 can be excited by long pulses ( 1 - 10 ns ) , which would give energy of a thz pulse @xmath202 @xmath3j .
finally we estimate the total energy of optical pulses using eq.([eq : opt - energy ] ) and overall conversion efficiency to thz radiation .
for room temperature 1 ps pulses pumping the side of the crystal with @xmath203 @xmath1740.1 cm size , the energy in one pulse according to eq.([eq : opt - energy ] ) is @xmath204 mj , for two pulses , @xmath205 mj , respectively .
we have to take into account that only @xmath206 of the incident energy is absorbed , where @xmath207 is the size of the crystal along which the optical pulses propagate .
for room temperature @xmath125 cm@xmath13 , @xmath208 @xmath167 and @xmath209 cm , the absorbed fraction is @xmath210 , so about twice the energy calculated above is required , resulting in @xmath211 mj total energy . for a pair of 100 fs pulses the same reasoning leads to the total energy of @xmath212 mj ( @xmath213 mj in one pulse from eq.([eq : opt - energy ] ) ) . for long pulses
the total required energy is @xmath214 mj .
the expected values of the energy of thz pulses , the peak amplitude of the thz field if focused to a spot of @xmath214 @xmath3 m size , required pumping radiation energy and duration and efficiency of optical to thz energy conversion for room temperature ruby are summurized in table [ tab : tablei ] . [ cols="^,^,^",options="header " , ]
it is important to note that the proposed technique can be applied to other solid materials with suitable transitions in the thz range .
similar schemes can be found in other ions as well , since for a typical rare - earth or transition metal ion the level structure is complex , having many crystal field and spin - orbit split components , separated by tens to hundreds of wave numbers .
for example , this method is readily applicable to cr@xmath9:beal@xmath11o@xmath215 ( alexandrite ) , having r - lines splitting of 41 @xmath14 ( @xmath216 thz ) in a non - inversion cr@xmath9 site @xcite , and r - lines absorption / emission cross - section ten times larger than ruby ( @xmath217 cm@xmath13 at room temperature ) .
it allows one to decrease peak optical intensities to induce maximal thz coherence and results in higher convertion efficiency .
the technique can also be applied to rare - earth ion doped materials , for example pr@xmath9 doped hosts , such as laf@xmath21 @xcite , y@xmath11sio@xmath218 @xcite ,
caf@xmath11 @xcite with the ground and excited state stark splittings being in a broad range 17 - 100 @xmath14 and higher , the same splittings are found in nd@xmath9 doped crystals , such as yag @xcite and yvo@xmath215 @xcite , and a number of others , for example tm : yag @xcite and er : yag @xcite and er : ylif@xmath215 @xcite .
also it is important to note that the technique can be applied to gases as well , but the coupling to thz radiation in this case occurs by a magnetic - dipole moment ( macroscopic polarization due to an electric - dipole moment is zero because of symmetry reasons ) . typically the magnetic dipole moment is smaller than the electric dipole one , so efficiency is smaller , but for higher densities and pressures this approach can still produce thz radiation .
the dipole moments for optical transitions can be stronger than for ruby , so the coherence can be larger .
this possibility is an extension of the method proposed in @xcite , although its analysis is beyond of the scope of the present paper .
the next important remark is related to the case when the thz transition of interest belongs to the ground state .
then there are a lot of new opportunities to use more sophisticated methods of coherence preparation .
in particular , the method of pulsed coherence production via stimulated raman adiabatic passage ( stirap ) @xcite can be used to control the duration of the thz pulse similarly to @xcite .
generally , all methods for coherence preparation via femtosecond pulse shaping developed to improve the sensitivity of coherent raman scattering @xcite , in particular , frequency chirping@xcite and fractional stirap @xcite , matched pulses @xcite , etc . can be applied .
it is interesting to compare the method of thz generation proposed in recent works @xcite and in this paper with currently available ones .
there are a variety of methods already considered successful for generation of short thz pulses , so it is worth to compare the above estimates with the currently achieved parameters . at the moment
the most impressive results in generation of short thz pulses were achieved using electron beam based sources , free - electron lasers ( fel ) @xcite and synchrotrons @xcite . for fels ,
typically thz pulses of about 1 ps duration with 1 - 40 ns distance between micropulses are generated , grouped into a few-@xmath219s trains .
energies of the micropulses are about 1 - 50 mj ( the parameters of felbe laser , rossendorf , germany ) .
recently , substantially higher powers of coherent broadband thz pulses , produced by synchrotron emission , were obtained from the electron beamline @xcite .
hundreds of fs - short half - cycle thz pulses were generated with energies up to 100 @xmath219j .
thz pulses with ns durations were achieved in thz semiconductor lasers ( quantum cascade lasers , p - ge and n - si lasers @xcite-@xcite ) , limited by the need of cryogenic cooling and in the case of p - ge lasers by high electric and magnetic fields required for laser operation . in p - ge lasers
@xcite-@xcite 20 ps pulse durations were obtained in a mode - locked regime with peak power up to several watts in a few-@xmath219s train of pulses . the obvious advantages of the optical crystal based thz emitters discussed in this work over fel and synchrotron sources are compactness ( typical optical solid - state materials are significantly smaller ) , cost , and ease of handling .
compared to thz qc lasers the crystals are much easier to grow and handle , they also can provide better coupling between the thz mode and the generating material due to a larger size of crystals .
other methods of short thz pulse generation are based on interaction of different materials with ultrashort laser pulses .
the most popular ones , giving subpicosecond thz pulses , are the photocurrent method using the auston - switch technique @xcite and the optical rectification technique @xcite .
since the proposed crystal - based thz sources are closer in their characteristics to these methods , we will make a more detailed comparison .
recently thz sources based on amlifier - laser systems such as ti : sapphire , utilizing different methods of thz radiation generation such as optical rectification in nonlinear crystals and laser - produced plasma and photocurrent thz generation in semiconductor antennas , became popular .
the transition metal and rare - earth ion doped crystals proposed in this work as thz emitters are suitable for amplifier - laser - based thz systems working in a single - shot regime @xcite , since the typically long ( hundreds @xmath219s - milliseconds ) population decay time of excited states of these ions limits the repetion rate .
crystals are also better in terms of saturation , able to withstand high laser pulse fluence and thus very compatible with low - repetiotion rate table - top laser systems producing pulses with energies in the 10 - 100 mj range with @xmath220 hz repetion rates .
fig.[fig : comp-1 ] compares the predicted in @xcite performance of different amplifier - laser - based thz sources and ruby at room temperature , at optical pulses energy of @xmath220 mj .
it shows that ruby is expected to perform as good as the biased gaas antenna with 1 kv / cm dc bias field and the plasma thz source with an external bias field , slightly yielding them in conversion efficiency .
fig.[fig : comp-2 ] compares the predicted in @xcite performance of different amplifier - laser - based thz sources at room temperature and ruby at liquid helium one , at optical pulses energy of @xmath221 mj .
the ruby source at liquid helium temperature can compete with the the plasma source utilizing the fundumental and second harmonic optical fields in terms of thz pulse energy and conversion efficiency , and yields it in peak thz field amplitude .
new methods of thz generation proposed in @xcite for atomic and molecular gases and in the present work for doped crystals potentially have very high efficiency , they offer the possibility of thz pulses with controlled durations including femtosecond region and the possibility of generation of pulses with high energy , already comparable with synchrotron or free - electron laser based thz sources .
implementation of the method , therefore , will open exciting opportunities in many fields of thz applications such as the time - resolved thz spectroscopy . using intense short pulses of thz radiation generated by the proposed method will significantly increase the performance of spectroscopic measurements , allowing maximal ( among currently existing methods ) temporal resolution .
extremely high efficiency of the proposed technique opens a way for orders of magnitude increase in sensitivity of the spectroscopic methods , thz tomography , non - destructive quality control , medical diagnostics and biomaterial characterization .
another advantage of the proposed scheme is a better control over the phase of the generated thz radiation , resulting in lower ( pulse duration @xmath176 bandwidth ) product compared to the non - coherent techniques .
resulting thz pulses have narrower bandwidth , thus offering an opportunity of better resolution measurements .
coherence of the generated thz pulses combined with their high energy opens the way to various non - linear phenomena in the thz range , as well as to coherent phenomena similar to the ones observed with lasers in the visible range , such as rabi oscillations , coherent transients , etc .
the proposed methods have , therefore , high potential for investigation of nonlinear effects in chemical and biological objects , in medical applications , which is a substantial , practically unexplored field of research .
in the present work we proposed a new technique for generation of short intense thz pulses in coherently driven doped optical crystals . the method is based on excitation of maximal thz coherence by a pair of resonant optical pulses , resulting in polarization build - up in the medium , which will radiate a thz pulse . as an example , a well - known laser material cr@xmath9:al@xmath11o@xmath21 ( ruby ) is considered and numerical estimates of expected thz and required optical field parameters are given . a number of other doped crystals , having transitions in the thz range , are suggested .
comparison is made with the existing methods of short thz pulses generation . in summary ,
the optical crystal based thz sources proposed in this work can potentially produce short high energy ( hundreds pj - tens @xmath219j ) thz pulses with durations ranging from 1 ps up to several ns
. the generated thz radiation will be coherent , thus offering the possibility of coherent interaction between the thz pulses and probed media .
this allows one to study various types of nonlinear penomena .
obvious advantages of crystals are their compact size , ease of growth and handling , robustness and low cost .
we gratefully acknowledge the support from the air force office of scientific research , the national science foundation , the defense advanced research projects , the office of naval research under award no .
n00014 - 03 - 1 - 0385 , and the robert a. welch foundation ( grant no .
a1261 ) . o. kocharovskaya and ya.i .
khanin , pisma zn . eksp .
. fiz . * 48 * , 581 ( 1988 ) , ( sov
jetp lett .
* 48 * , 630 ( 1988 ) ) ; s.e .
harris , phys .
lett . * 62 * , 1033 ( 1989 ) ; m.o .
scully , s.y .
zhu , and a. gavrielides , phys .
* 62 * , 2813 ( 1989 ) .
zibrov , m.d.lukin , d.e.nikonov , l.hollberg , m.o.scully , v.l.velichansky , h.g.robinson , phys .
lett . * 75 * , 1499 ( 1995 ) ; g.g .
padmabandu , g.r.welch , i.n.shubin , e.s.fry , d.e.nikonov , m.d.lukin , m.o.scully , phys .
lett . * 76 * , 2053 ( 1996 ) .
g.alzetta , a.gozzini , l.moi , g.orriols , nuovo cimento b * 36 * , 5 ( 1976 ) ; h.r.gray , r.m.whitley , c.r.stroud , jr . , opt .
lett . * 3 * , 218 ( 1978 ) ; k.j.boller , a.imamoglu , s.e.harris , phys .
. lett . * 66 * , 2593 ( 1991 ) ; j.e.field , k.h.hahn , s.e.harris , phys .
* 67 * , 3062 ( 1991 ) . | we have shown that a coherently driven solid state medium can potentially produce strong controllable short pulses of thz radiation .
the high efficiency of the technique is based on excitation of maximal thz coherence by applying resonant optical pulses to the medium .
the excited coherence in the medium is connected to macroscopic polarization coupled to thz radiation .
we have performed detailed simulations by solving the coupled density matrix and maxwell equations . by using a simple @xmath0-type energy scheme for ruby ,
we have demonstrated that the energy of generated thz pulses ranges from hundreds of pico - joules to nano - joules at room temperature and micro - joules at liquid helium temperature , with pulse durations from picoseconds to tens of nanoseconds .
we have also suggested a coherent ruby source that lases on two optical wavelengths and simultaneously generates thz radiation .
we discussed also possibilities of extension of the technique to different solid - state materials . |
despite the widespread recent theoretical and experimental interest in quantum phase transitions , a direct quantitative confrontation between theory and experiment has been difficult to achieve for systems in two and higher spatial dimensions .
a major obstacle is that it is often difficult to tune system parameters over the range necessary to move across a quantum critical point .
furthermore , for many examples where such tuning is possible , the theory for the quantum critical point is intractable .
consequently , the analysis of the data is often limited to the testing of general scaling ansatzes , without specific quantitative theoretical predictions . a class of quantum phase transitions
have recently been exceptionally well characterized in a variety of experiments .
these experiments study the influence of a strong applied magnetic field on insulating spin - gap compounds@xcite .
the low lying spin excitations behave like spin @xmath7 canonical bose particles , and the energy required to create these bosons vanishes at a critical field @xmath8 , which signals the position of a quantum phase transition@xcite with dynamic critical exponent @xmath9 . in spatial dimensions
@xmath10 , the quantum critical fluctuations are well described by the familiar bose - einstein theory of non - interacting bosons , and no sophisticated theory of quantum criticality is therefore necessary to interpret the experiments .
the upper - critical dimension of the quantum critical point is @xmath11 , and the boson - boson interaction vanishes logarithmically at low momenta .
so naively , one expects that the @xmath11 case is also weakly coupled , and no non - trivial quantum critical behavior obtains .
the primary purpose of this paper is to show that the above expectation for the quantum - criticality of @xmath11 bose gas is incorrect .
while the pairwise interactions between the bosons are indeed weak , the collective properties of the finite - density , thermally excited bose gas pose a strong - coupling problem .
we will demonstrate here that an effective classical model , which can be readily numerically simulated , provides a controlled description of this problem .
this will allow us to present predictions for the evolution of the dynamic spectrum of the @xmath11 bose gas across the quantum critical point at non - zero temperatures .
our results can be applied to two - dimensional spin - gap antiferromagnets , in the vicinity of the gap - closing transition induced by an applied magnetic .
recent experiments@xcite on piperazinium hexachlorodicuprate ( phcc ) , ( c@xmath12h@xmath13n@xmath14)cu@xmath14cl@xmath15 , will be compared with our results in section [ sec : conc ] .
these experiments are able to easily access the finite temperature quantum - critical region , which we will describe by our theory of the dilute bose gas in @xmath11 ; this bose gas is in a regime of parameters which is typically not examined in the context of more conventional atomic bose gas systems @xcite .
results similar to those presented here @xcite also apply to other quantum critical points with upper - critical spatial dimension @xmath11 _ i.e. _ while the zero temperature properties can be described in a weak - coupling theory , all non - zero temperature observables require solution of a strong - coupling problem .
a prominent example of such a quantum - critical point is the hertz theory of the onset of spin - density - wave order in a fermi liquid@xcite .
results for the thermodynamic and transport properties can be obtained by the methods presented here @xcite .
we maintain that such results are necessary for understanding experiments , and that previous theoretical results@xcite are inadequate for a quantitative analysis .
our results will be presented in the context of the phase diagram of the @xmath11 bose gas shown in fig .
[ fig1 ] as a function of the boson chemical potential , @xmath2 , and temperature , @xmath1 . as a function of the chemical potential @xmath2 and the temperature @xmath1 of a dilute bose gas in two spatial dimensions .
the energy scale @xmath16 , defined in eq .
( [ defu ] ) , is a measure of the pairwise interaction between the atoms , while @xmath17 , defined in eq .
( [ defr ] ) , is an energy scale controlling collective excitations .
the physical properties of systems with the same value of @xmath18 are the same , apart from the change in the value of the energy scale @xmath17 .
the kosterlitz - thouless transition occurs at the contour labelled @xmath19 at @xcite @xmath20 , and the superfluid phase is present for larger @xmath18 .
the contours shown are for equal spacings of values of @xmath21 , with the yellow region on the lower right including values @xmath22 .
our primary results here are in the quantum critical region , with intermediate values of @xmath18 at @xmath3 and @xmath23.,width=307 ] at @xmath24 , there is a quantum critical point at @xmath3 . for @xmath25 ,
the ground state is simply the vacuum with no bosons , while for @xmath26 , there is a finite density of bosons in the ground state . in the spin gap antiferromagnets , @xmath27 , where @xmath28 is the applied field , @xmath29 is the critical field , @xmath30 is the gyromagnetic ratio , and @xmath31 is the bohr magneton .
[ fig1 ] presents results in terms of contours with equivalent physical properties , up to an overall energy scale ( @xmath17 ) whose @xmath1 and @xmath2 dependence will be explicitly presented below , as will the equations determining the shape of the contours .
one of the contours is the position of the kosterlitz - thouless ( kt ) transition of the @xmath11 bose gas , which is present only for @xmath26 .
we are primarily interested here in the quantum - critical region,@xcite which is roughly the region with @xmath0 .
in particular , we will be able to quantitatively examine the signature relaxation rate @xmath32 of the quantum - critical regimes : the `` bose molasses '' dynamics .
our results and methods also describe the crossover to the kt transition , as well as ( in principle ) the region with @xmath33 .
we summarize here the universal aspects of our results of the quantum - critical bose dynamics in two spatial dimensions .
the continuum quantum field theory of the critical point has logarithmic corrections to scaling ; consequently , the properties of the continuum theory do have a logarithmic dependence upon a non - universal ultraviolet cutoff .
we will show that this cutoff dependence can be isolated within a single parameter ; all other aspects of the theory remain universal , and can be accurately computed
. we will be interested in the continuum bose gas theory with the partition function @xmath34 .
\label{zb}\end{aligned}\ ] ] for phcc , the mass @xmath35 can be directly determined from the dispersion of the @xmath7 excitation .
we will present results primarily in the small @xmath36 quantum critical region of fig [ fig1 ] , although our formalism can be extended to other regions , including across the kosterlitz - thouless transition into the `` superfluid '' phase .
the bare interaction between the bose particles , @xmath37 , is renormalized by repeated interactions between the particles , in the @xmath1 matrix , to the value @xmath38 where @xmath5 is a high energy cutoff , and the square - root function in the argument of the logarithm is an arbitrary , convenient interpolating form .
the parameter @xmath5 is the sole non - universal parameter appearing in the predictions of the continuum theory . for a spin gap antiferromagnet like phcc
, we expect @xmath39 , where @xmath40 is a typical exchange constant .
for our purposes , it is convenient to rescale @xmath41 to a parameter @xmath16 , which has the dimensions of energy : @xmath42 our universal results for the continuum theory are predicated on the assumption that the logarithm in eqs .
( [ vr ] ) and ( [ defu ] ) is `` large '' . at first glance
, it would appear from eq .
( [ vr ] ) that the quantum theory of the bose gas is weakly coupled in the leading - logarithm approximation .
however , as will be clear from our analysis ( and has been noted in earlier works @xcite ) , this is _ not _ the case : although pairwise interactions are weak , the collective static and dynamic properties of the gas remain strongly coupled even when the logarithm is large .
we shall argue , that to leading order in the logarithm , these strong coupling effects can be captured in an effective classical model .
the latter model is amenable to straightforward numerical simulations , and so accurate quantitative predictions become possible , whose precision is limited only by the available computer time .
previous analyses @xcite of the dilute bose gas in two dimensions assumed ( when extended to the quantum critical region ) that @xmath43 was large ; a more precise version of this condition appears below , from which it is clear that this condition is essentially impossible to satisfy in practice .
we will not make any assumption on the value of such a double logarithm , and only make the much less restrictive assumption on a large single logarithm .
the characteristic length and time scales of the quantum - critical bose gas are set by a dimensionful parameter , which we denote @xmath17 . like @xmath16 ,
we choose this to have dimensions of energy ; so a characteristic length is @xmath44 , while a characteristic time is @xmath45 .
the value of @xmath17 is determined by the solution of the following equation @xmath46 note that this defines a @xmath47 for all @xmath48 . to understand the magnitude of the various scales , we now discuss approximate solutions of eqs .
( [ defu ] ) and ( [ defr ] ) at the quantum critical point , @xmath3 .
the estimates below should not be used in place of the full solutions in applications of our results to experiments . for the energy scale associated with @xmath17
we obtain the estimate @xmath49 note that the energy scale @xmath17 is logarithmically smaller than @xmath50 : this will be the key in justifying an effective classical description of the dynamics .
numerically , we can easily go beyond eq .
( [ rapprox ] ) , and obtain the full solution of eq .
( [ defr ] ) at @xmath3 ; this is shown in fig .
[ rcrit ] .
dependence of @xmath17 at the quantum critical point , @xmath3 , obtained by solving eq .
( [ defr ] ) .
@xmath5 is a non - universal ultraviolet energy cutoff.,width=268 ] with the two dimensionful parameters , @xmath16 , and @xmath17 , at hand , the reader will not be surprised to learn that the effective classical theory is characterized by the dimensionless coupling @xmath18 .
indeed , as we will see , the ratio @xmath18 behaves like an effective ginzburg parameter for the classical theory . from eqs .
( [ defr ] ) and ( [ rapprox ] ) we estimate @xmath51 in the quantum - critical region . again , we can go beyond the asymptotics , and obtain precise values for @xmath18 by numerical solution of eqs .
( [ defr ] ) and ( [ defu ] ) , and the result appears in fig [ urcrit ] .
dependence of the dimensionless ratio @xmath18 at the quantum critical point , @xmath3.,width=268 ] so , unless @xmath52 is astronomically large , the ratio @xmath18 is not small , and the classical theory is strongly coupled .
as we have already noted , we will not make the assumption of large double logarithms here , but instead obtain numerical results for the strongly coupled theory .
the primary result of this paper is that the _ low energy properties of the dilute bose gas are universal functions of the ratio @xmath18_. this key result is illustrated in fig [ fig1 ] where we plot the loci of points with constant @xmath18 , obtained from eqs .
( [ defu ] ) and ( [ defr ] ) .
the physical properties of systems along a fixed locus are the same , apart from an overall re - scaling of energy and distance scales which are set by the value of @xmath17 . in most cases , the universal dependence on @xmath18 can be determined by straightforward numerical simulations .
as has been shown earlier , in a different context@xcite , the effective classical model undergoes a kosterlitz thouless transition @xcite ( see fig [ fig1 ] ) at @xmath53 .
the values of @xmath18 in the quantum - critical region are smaller ( see eq .
( [ ur ] ) and fig [ fig1 ] ) and the focus of our attention will be on these smaller values . with an eye to neutron scattering observations in phcc , we focus in this paper on the frequency dependence of the boson green s function . in the antiferromagnet ,
this green s function is proportional to the two - point spin correlation function in the plane orthogonal to the applied field _
i.e. _ the correlator of @xmath54 and @xmath55 .
the spectral density of this green s function yields the neutron scattering cross - section .
in particular , we will consider the green s function ( in imaginary time ) @xmath56 after analytic continuation to real frequency , and application of the fluctuation - dissipation theorem , we obtain the corresponding dynamic structure factor @xmath57 . one of our main results is that this structure factor obeys the scaling form @xmath58 we will numerically determine the universal function @xmath59 here for a range of values of @xmath18 in the quantum - critical regime .
our results for @xmath57 are obtained directly in real time , and so do not suffer ambiguities associated with analytic continuation . for a significant range of values of @xmath18 of relevance to quantum criticality , we found that our numerical results could be fit quite well with the following simple lorentzian functional form for @xmath60 @xmath61 where the scaling form in eq .
( [ sscale ] ) implies that the dimensional numbers @xmath62 , @xmath63 , and @xmath64 are _ all universal functions of the ratio @xmath18_. notice that this describes a neutron resonance at frequency @xmath65 with width @xmath66 ; our primary purpose here is to provide theoretical predictions for the temperature dependence of these observables .
our numerical results for the values of @xmath62 , @xmath63 and @xmath64 as a function of @xmath18 appear in figs .
[ zval ] , [ omega_ur ] , and [ gamma_ur ] later in the paper .
this section will obtain the properties of the continuum theory in eq .
( [ zb ] ) which were advertized above . the analysis of the static properties of eq .
( [ zb ] ) has been outlined in ref . .
the key step is the integrate out all the @xmath67 modes of @xmath68 to obtain an effective action only for the zero frequency component . among the important effects of this
is to replace the bare interaction @xmath37 by a renormalized interaction @xmath41 obtained by summing ladder diagrams @xmath69 the co - efficient of @xmath70 is also renormalized , as we specify below .
the resulting effective theory for the zero frequency component is most conveniently expressed by defining @xmath71 and rescaling spatial co - ordinates by @xmath72 this yields the following classical partition function @xmath73.\end{aligned}\ ] ] here the energy @xmath16 is as defined in eq .
( [ defu ] ) , while the ` mass ' @xmath74 is given by @xmath75 the most important property of this expression for @xmath74 is that the integral of @xmath76 is not ultraviolet finite , and has a logarithmic dependence on the upper cutoff .
however , this is not a cause for concern .
the theory @xmath77 is itself not a ultraviolet finite theory , and its physical properties do have a logarithmic dependence on the upper cutoff . fortunately ( indeed , as must be the case ) , the cutoff dependence in @xmath74 above is precisely that needed to cancel the cutoff dependence in the correlators of @xmath77 so that the final physical results are cutoff independent .
this important result is demonstrated by noting that the only renormalization needed to render @xmath77 finite is a ` mass ' renormalization from @xmath74 to @xmath17 , as defined by @xmath78 here , the @xmath17 in the propagator on the r.h.s .
is arbitrary , and is chosen for convenience .
we could have chosen a propagator @xmath79 where @xmath80 is an arbitrary numerical constant ; this would only redefine the meaning of the intermediate parameter @xmath17 , but not the value of any final observable result .
combining eq .
( [ e3 ] ) with eq .
( [ e1 ] ) , we observe that the resulting expression for @xmath17 is free of both ultraviolet and infrared divergences .
the momentum integrals can be evaluated , and lead finally to the expression for @xmath17 already presented in eq .
( [ defr ] ) .
when expressed in terms of the renormalized parameter @xmath17 , the properties of the continuum theory @xmath77 are universal ( _ i.e. _ independent of short - distance regularization ) .
they are defined completely by the length scale @xmath81 and the dimensionless ratio @xmath18 .
the field @xmath82 is also dimensionless , and acquires no anomalous dimension .
this means _
e.g. _ the equal time correlations obey the scaling form @xmath83 where @xmath84 is a universal function .
note that here , and in the remainder of the paper , we have rescaled momenta corresponding to eq .
( [ rscale ] ) @xmath85 so that @xmath86 has the dimensions of energy .
the function @xmath84 can be easily computed in a perturbation theory in @xmath18 , and results to order @xmath87 appear in ref . .
however , we are interested in values of @xmath18 which are not small , and for this we have to turn to the numerical method described in the following subsection
. we will analyze numerically @xmath77 by placing it in a square lattice of spacing @xmath88 , and verifying that the correlations measured in monte carlo simulations become @xmath88 independent and universal in the limit @xmath89 .
the partition function on the lattice is @xmath90 .
\nonumber\end{aligned}\ ] ] the parameter @xmath91 is _ not _ equal to the parameter @xmath74 above . instead , the mapping to the quantum theory has to be made by requiring that the values of the renormalized @xmath17 are the same .
in the present lattice theory we have @xmath92 we can measure lengths in units of @xmath17 , and for each value of @xmath16 , determine @xmath91 from eq .
( [ e10 ] ) , and test if the monte carlo correlations are independent of @xmath88 in the limit of small @xmath88 .
the resulting correlations then determine the scaling function @xmath84 in ( [ e5 ] ) .
we used this method to determine the values of the function @xmath93 for a sample set of values of @xmath18 appropriate to the quantum - critical region .
note that @xmath94 , where @xmath62 is the amplitude appearing in the dynamic function in eq .
( [ chires ] ) .
we used the wolff cluster algorithm ( as described in ref . ) to sample the ensemble specified by @xmath95 .
measurement of the resulting correlations led to the results shown in fig [ zval ] . .
the parameter @xmath62 is that appearing in eq .
( [ chires]).,width=268 ] we now extend the classical static theory above to unequal time correlations by a method described in some detail in refs . and .
as argued there , provided the scale @xmath96 , the unequal time correlations can also be described by classical equations of motion . in the context of perturbation theory , the reduction to classical equations of motion is equivalent to the requirement that it is a good approximation to replace all bose functions by their low energy limit : @xmath97 from eq .
( [ rapprox ] ) we observe that the requirement on @xmath17 is satisfied in the quantum - critical region .
it also holds everywhere in the superfluid phase , with @xmath26 , where @xmath17 becomes exponentially small in @xmath98 ( as can be shown from eq .
( [ defr ] ) ) .
however , it does fail in the low @xmath1 ` spin - gap ' region with @xmath99 , where @xmath100 .
we will not address this last region here , although a straightforward perturbative analysis of the full quantum theory is possible here , as noted in ref . .
some results on the perturbation theory appear in appendix [ app : pert ] .
the classical equations of motion obeyed by @xmath82 are merely the c - number representation of the heisenberg equations of motion obeyed by @xmath68 . with the rescalings in eqs .
( [ psiscale ] ) and ( [ rscale ] ) , these are @xmath101 following the reasoning leading to eq .
( [ e5 ] ) , it follows that the correlations of the @xmath82 evolution described by these equations of motion obey the scaling form @xmath102 where the scaling function @xmath103 can be determined numerically , as we describe below .
after the rescalings in eqs ( [ psiscale ] ) and ( [ rscale ] ) , we conclude that the fourier transform of @xmath104 to @xmath105 is related to the dynamic structure factor @xmath57 defined below eq .
( [ defchi ] ) by @xmath106 we now describe our numerical computation of @xmath103 .
we begin by sampling the ensemble of @xmath107 values specified by @xmath95 , as in the previous section .
once the ensemble is thermalized , we choose a typical set of values of @xmath107 as the initial condition .
these are evolved forward in time deterministically by solving the equations of motion @xmath108 where @xmath109 . for each initial condition , this defines a @xmath110 .
then @xmath111 for a lattice of @xmath112 sites , and where the average is over the ensemble of initial conditions .
sample results from such simulations appear in figs .
[ time1 ] and [ time2 ] . obtained from our numerical simulation .
the results above are for @xmath113 .
two different values of @xmath17 are shown , and the axes have been scaled to demonstrate the scaling collapse as required by eq .
( [ e7 ] ) .
the fit with the functional form in eq .
( [ sinedamp ] ) is also shown.,width=249 ] , but for the imaginary part of the correlation function.,width=249 ] we have fit each observed time evolution to the functional form @xmath114 where @xmath62 , @xmath63 , @xmath64 are numbers determined from the fit .
a sample fit is shown in figs .
[ time1 ] and [ time2 ] .
as is clear from the figures , this form provides an excellent fit over a substantial time window . taking the fourier transform of this result , and using eq .
( [ fdt ] ) , we obtain a result for @xmath57 in the form in eq .
( [ chires ] ) , with the same parameters @xmath62 , @xmath63 , @xmath64 .
the results for @xmath62 appeared earlier in fig .
[ zval ] , and our results @xmath63 and @xmath64 as a function of @xmath18 appears in figs .
[ omega_ur ] and [ gamma_ur ] . on @xmath18 .
the points are the results of the numerical simulation , while the line is a best fit polynomial used for interpolation.,width=249 ] , but for the dimensionless damping @xmath64 .
the full line at small @xmath18 is the result of the perturbation theory obtained in eq .
( [ c1]).,width=249 ] we can now combine the results in these figures with those in fig [ urcrit ] , and obtain predictions for the @xmath1 dependence of the resonant frequeny and damping in the dynamic structure factor in eq . ( [ chires ] ) at the quantum - critical point .
these results appear in fig [ frequency_crit ] and [ damping_crit ] .
on @xmath1 at the quantum - critical point @xmath3 , obtained from figs .
[ urcrit ] and [ omega_ur].,width=288 ] on @xmath1 at the quantum - critical point @xmath3 , obtained from figs .
[ urcrit ] and [ gamma_ur].,width=288 ]
this paper has argued that the quantum - critical dynamics of the two - dimensional bose gas ( and of other quantum critical points in two spatial dimensions ) represents a strong - coupling problem .
nevertheless , an effective classical description was obtained to leading logarithmic order , allowing tractable numerical simulation .
the primary results of this simulation at the quantum critical point appear in figs .
[ frequency_crit ] and [ damping_crit ] , which specify the parameters appearing in eq .
( [ chires ] ) . in comparing our results to experiments , we have to set the value of the unknown high energy cutoff @xmath5 . for the range of parameters shown in figs .
[ frequency_crit ] and [ damping_crit ] , we find a roughly linear dependence of @xmath115 and @xmath116 on @xmath1 , with @xmath117 and @xmath118 .
the experiments @xcite on the quantum - critical point of phcc , also observe a roughly linear dependence of peak frequency and width on temperature , but with different co - efficients ; the peak frequency @xmath119 , while the width is @xmath120 .
choosing a different range of @xmath5 for the theory , _
e.g. _ assuming the experiments are in the range @xmath121 , will change the theoretical predictions for @xmath115 and @xmath116 , but does not improve the agreement with experiments . for the peak frequency , the work of ref .
suggests an origin for the above discrepancy .
a similar theory was used in that paper to obtain the predictions of the peak frequency , but keeping the full lattice dispersion for the spin excitations , and the quantum bose function values for the occupation numbers : good agreement was found between such a theory and the experimental observations . the continuum and classical limits taken in the present paper
were avoided . for the width of the spin excitation
, we expect that the damping is more strongly dominated by the low energy and low momentum excitations , and so the present continuum , classical theory should yield a more accurate description of the experiments .
this is indeed the case , relative to the poor accuracy of the peak frequency in the continuum theory .
nevertheless , a discrepancy of a factor of @xmath122 remains between our present quantum critical theory and the experimental observations .
it is possible that taking the classical theory also on the lattice will improve agreement with experiments .
however , there are also corrections to the present continuum theory which could improve the situation : in particular , at higher order in @xmath41 , there appear renormalization of the time - derivative term in eq .
( [ zb ] ) .
this renormalization would change the time - scale in the classical equations of motion , and so change the overall frequency scale of the results in figs .
[ frequency_crit ] and [ damping_crit ] .
we thank t. hong and c. broholm for useful discussions .
this research was supported by nsf grant dmr-0537077 .
here we present the results of a direct perturbative computation of @xmath123 . to order @xmath124 , perturbation theory yields@xcite @xmath125 where the frequency summation is done most easily by partial fractions and yields @xmath127 \left[n(k_2
^ 2 + r ) - n((k_1 + k_2)^2 - k_1 ^ 2)\right ] } { i \omega + ( k_1 + k_2)^2 - k_1 ^ 2 - k_2 ^ 2 - r}. \nonumber\end{aligned}\ ] ] now we analytically continue to real frequencies , and take the imaginary part at @xmath128 , which is the leading order position of the pole in eq .
( [ chipert ] ) .
this yields @xmath129 \nonumber \\ & ~&~~~~~~~~\times\left[n(k_2 ^ 2 + r ) - n((k_1 + k_2)^2 - k_1 ^ 2)\right ] \delta ( ( k_1 + k_2)^2 - k_1 ^ 2 - k_2 ^ 2 ) \nonumber \\ & = & - 2 \pi \left ( \frac{u}{k_b t } \right)^2 \int \frac{d^2 k_1}{4 \pi^2 } \frac{d^2 k_2}{4 \pi^2 } \left [ n(k_1 ^ 2 + r ) - n(k_1 ^ 2 + k_2 ^ 2 + r ) \right ] \left[n(k_2 ^ 2 ) - n(k_2 ^ 2 + r)\right ] \nonumber \\ & ~&~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\times \delta ( ( k_1 + k_2)^2 - k_1 ^ 2 - k_2 ^ 2).\end{aligned}\ ] ] now we can do the angular integration because the angle appears only in the argument of the delta function , and obtain @xmath130 this result is a function of @xmath131 which has to be evaluated numerically .
however , it is instructive to examine its value in the classical limit , upon applying the approximation in eq .
( [ capprox ] ) , when we obtain @xmath132 notice that explicit factors of @xmath50 have dropped out , and consequently eqs .
( [ chipert ] ) and ( [ c1 ] ) are consistent with the scaling form eq .
( [ chires ] ) .
it is also interesting to compare the value of the classical limit in eq .
( [ c1 ] ) with that obtained from the full quantum expression in eq .
( [ q1 ] ) . for @xmath133 ( which are roughly the quasiparticle energy values obtained in the experiments on phcc @xcite ) , eq .
( [ q1 ] ) evaluates to a value which is 16% smaller than eq .
( [ c1 ] ) .
this gives us an estimate of the error made by representing the quantum - critical theory by classical equations of motion . | the dilute , two - dimensional bose gas exhibits a novel regime of relaxational dynamics in the regime @xmath0 where @xmath1 is the absolute temperature and @xmath2 is the chemical potential .
this may also be interpreted as the quantum criticality of the zero density quantum critical point at @xmath3 .
we present a theory for this dynamics , to leading order in @xmath4 , where @xmath5 is a high energy cutoff .
although pairwise interactions between the bosons are weak at low energy scales , the collective dynamics are strongly coupled even when @xmath6 is large .
we argue that the strong - coupling effects can be isolated in an effective classical model , which is then solved numerically .
applications to experiments on the gap - closing transition of spin gap antiferromagnets in an applied field are presented . |
the experimental studies of nuclear response in charge exchange reactions were extended in the eighties to high excitation energies where a first nucleon resonance the @xmath0-isobar can be excited @xcite .
the detailed studies were done for the ( @xmath1he , t ) charge
exchange reaction at different projectile energies @xcite and different targets @xcite .
the properties of the @xmath0 excited in a nucleus were found different compared to the case of the reaction on a single nucleon .
the difference was both in the peak position and the width of the resonance excited in a complex nucleus .
the review of the observed phenomena can be found in the recent paper @xcite .
the appealing explanation of this phenomenon is related to medium effects , namely , the excitation of a pionic nuclear mode @xcite , @xcite , @xcite , @xcite although another explanation has been proposed as well @xcite . in this picture
the @xmath0 in nuclear matter does not exist as separate resonance but forms a collective excitation consisting of pionic , @xmath0 and nucleon degrees of freedom . at first sight
one should not expect sizeable medium effects for ( @xmath1he , t ) reaction since the reaction takes place at the surface of the target @xcite .
for inelastic reactions , however , the absorption is smaller and one should use absorption factor different from that used for elastic scattering @xcite providing both the medium effects and the magnitude of inelastic cross section .
another important point is the account of the finite target size in the response function . as it will be shown below different @xmath0-hole multipoles
are peaked at different energies so , part of the observed width can be attributed to this displacement .
besides , on finite nucleus the process of coherent pion production is possible .
the process is absent in infinite nuclear matter and it contributes to the shift of the peak position as well .
here we present the results of the absolute cross section calculations of the @xmath2c(@xmath1he , t ) reaction at 0@xmath3 and at the kinetic energy of the @xmath1he t@xmath4=2 gev .
in the next section we discuss the model for the reaction amplitude that is a driving force for the nuclear response . in the section [ resp ]
the pionic response function of a finite nucleus is discussed and the cross
section is calculated in the section [ cross ] .
we shall start from the discussion of the models for the elementary charge
exchange reaction p(p , n)@xmath5 . from the early sixties
it was shown this reaction can be described by ope model @xcite , at least at low momentum transfer .
this analysis has been extended for wide range of the proton energies in @xcite , and has been repeated with some minor modifications in connection to the analysis of p(@xmath1he , t ) @xmath5 reaction in @xcite . as it was shown all existing data in the region of low momentum transfer are well described by ope with the soft monopole @xmath6nn- and @xmath6n@xmath0- form factors @xmath7 the parameter @xmath8 mev at low proton energy and slightly decreases with the proton energy reflecting increase of the absorption effects at high energy . with these soft form factors ( [ form ] ) the main contribution to the @xmath0-production comes from the direct graph shown in fig.1 .
the exchange contribution is small for the p(p , n)@xmath5 reaction . in this model
the amplitude is completely longitudinal with respect to the momentum transfer . in the other model used for the description of the @xmath0- production at 800 mev proton energy @xcite the transverse part of the amplitude
was described by @xmath9- exchange and hard form factors with @xmath10 = 1.2 gev and @xmath10 = 1.7 gev were used for @xmath6n@xmath0- and @xmath9n@xmath0- vertexes .
the magnitude of the cross section and the momentum transfer dependence are reproduced in this model due to cancellation between the direct and the exchange parts of the amplitude .
the cancellation is rather delicate and at higher proton energy it can be broken resulting in wrong momentum transfer dependence @xcite . at very high energy
the situation is different .
the @xmath6-exchange contribution decreases as @xmath11 , where _ s _ is the center of mass energy squared , while for the @xmath9-exchange the decrease is slower .
its contribution falls down like @xmath12 for small momentum transfer where @xmath13 is the corresponding regge trajectory .
thus , in the asymptotic region at high energy one should expect the dominance of the @xmath9-exchange even at forward angles .
below 20 gev the cross
section for forward scattering follows the @xmath14 law @xcite so , the contribution of @xmath9-exchange at intermediate energies is believed to be small . for the p@xmath15he ,
t)@xmath16 reaction the situation is similar to the ( p , n ) case .
the @xmath6-exchange with the soft form factors ( [ form ] ) gives reasonable description of the absolute cross section and the tritium spectrum at forward angle for all existing data @xcite .
nevertheless , at the kinetic energy of @xmath1he 2 gev , which is close in kinematics to 800 mev ( p , n ) one can get good description using @xmath17- exchanges as well @xcite . it would be very desirable to extend the last analysis to higher @xmath1he energies .
it is convenient to start with the plane waves for both projectile and ejectile in order to obtain an expression for the cross section that can be easily generalized to the distorted waves . in pwia
the cross
section is proportional to the matrix element , shown in fig.1 , squared and summed over final nuclear and @xmath0 - states .
@xmath18 for plane waves @xmath19 , it gives for the cross
section @xmath20 the expression under the sum is just imaginary part of the pionic self energy in nuclear medium . @xmath21 using the pionic self energy ( [ self ] ) we obtain the final expression for the cross section suitable for inclusion of medium effects . @xmath22
the main effect of nuclear medium is the renormalization of the pion propagator by intermediate @xmath0-hole loops giving the major contribution to the pionic self energy ( [ self ] ) near the @xmath0-resonance . to take it into account one must change in ( [ fsec ] ) the bare pion propagator @xmath23 for the dressed one @xmath24 , where @xmath25 making this change we are going out of the scope of the impulse approximation .
the imaginary part of the bare pion propagator is equal to zero for negative @xmath26 .
using it we obtain @xmath27 with these changes the cross section ( [ fsec ] ) becomes @xmath28 it is clear from ( [ res ] ) the pion propagator @xmath24 in nuclear medium is just the response function to a virtual pion probe .
the excitation created by a virtual pion is no more pure @xmath0-hole but a superposition of the @xmath0-hole and pionic degrees of freedom , which is usually called the pionic mode . the unquenched @xmath0-hole self energy ( [ self ] ) produces too much of attraction giving unreasonably low excitation energy for the pionic mode .
in order to make the description more accurate several effects should be taken into account .
first of all , more correct @xmath6n- scattering amplitude reproducing @xmath29 and @xmath30 partial waves should be used since we are interested in the energies lower than the @xmath0 in vacuum . for this purpose
one should add the born diagrams with a nucleon intermediate state , _
u_-channel @xmath0 diagram and a @xmath31-term arising from the @xmath31-commutator @xcite .
second , the short
range n@xmath0- correlations must be taken into account .
@xmath32 in nuclear matter the effect of short range correlations can be accounted in the following way .
let us define the @xmath0-hole response function @xmath33 by @xmath34 then , @xmath35 similar effect should be taken into account for the nucleon
hole response function as well .
but , in the region of interest in excitation energy its contribution is negligible and will be omitted below .
in contrast , the backward @xmath0-hole loops and the @xmath31-term must be retained since they have their own dependence on the virtual pion mass @xmath26 that influences the position of the pionic branch in nuclear matter .
finally , the virtual pion can be absorbed in nuclear medium emitting two or more nucleons . in the resonance region
the absorption goes via intermediate @xmath0 and it can be described by a @xmath0-nucleus optical potential . at lower energies
the other mechanisms with different intermediate states contribute to the absorption . to take it into account we use the ericson
ericson optical potential from pionic atoms @xcite @xmath36 where @xmath37 is the nuclear matter density . for pionic atoms @xmath38 .
it was obtained from the fit of the mesoatomic level width and contains all absorption mechanisms including the intermediate @xmath0 . using this value in our case
will give double counting since intermediate @xmath0 is considered explicitly .
thus , @xmath39 should be considered as a free parameter accounting another absorption mechanisms . with these corrections
the pion self energy in nuclear matter is @xmath40 where @xmath41 is the pion decay constant @xmath42 mev and @xmath43 is the @xmath31-commutator for forward pion scattering .
@xmath44 includes both forward and backward @xmath0-hole loops corrected for short range correlations .
for a finite nucleus it is convenient to work in the configuration space where the self energy ( [ full ] ) and the pion propagator become the functions of two distinct variables instead of functions of the distance between coordinate points in nuclear matter . @xmath45
@xmath46 the @xmath47 - vertex in the configuration space is @xmath48 where @xmath49 . for a spherical nucleus the self energy ( [ full ] )
has simple multipole expansion @xmath50 the similar expansion exists for the @xmath51-vertex @xmath52 where the tensor operator @xmath53_{jm } , \label{tens}\ ] ] and the radial vertex @xmath54 is @xmath55 here @xmath56 and @xmath57 are the spherical bessel functions with an imaginary argument .
this is the advantage of using the simple scalar formfactors ( [ form ] ) .
the @xmath0-hole response function @xmath58 can be expanded using the set of tensor operators ( [ tens ] ) @xmath59\cdot n_{j_{n}l_{n}}\cdot r_{j_{n}l_{n}}(r)r_{j_{n}l_{n}}(r ' ) , \label{fresp}\end{aligned}\ ] ] where @xmath60 are the nucleon occupation numbers , @xmath61 the radial wave function of bounded nucleon and the @xmath62 is its energy .
@xmath63 is the green function of the radial schrdinger equation for the @xmath0 moving in the mean nuclear optical potential .
it was calculated using two independent solutions of the radial schrdinger equation .
the @xmath0-hole contribution to the pion self
energy was calculated using the following expression @xmath64 where @xmath65 related to @xmath66 via linear integral equation accounting the short range correlations ( [ corr ] ) @xmath67
for numerical calculations it is convenient to come back from expression ( [ res ] ) to more complex one similar to ( [ fsec ] ) @xmath68 the product sign means integration over all coordinates in the configuration space and the overline is the averaging and summing over spins of a @xmath1he and a triton . in infinite matter ( [ ddsec ] ) would be the only contribution to the inclusive cross section .
in finite nucleus , however , the cut of a bare pionic line gives nonzero contribution corresponding to coherent pion production .
@xmath69 in the plane wave approximation the @xmath70 vertex is @xmath71 where the effective momentum transfer @xmath72 in lab .
system is @xmath73 here @xmath74 is the total energy of @xmath1he and @xmath75 is its mass . at first order in @xmath76
it can be rewritten as @xmath77 the effective momentum transfer squared coincides with the four momentum transfer squared .
@xmath78 is the ( @xmath1he , t ) transition form factor . the multipole expansion of the vertex looks as follows @xmath79 where @xmath80 are the tensor operators analogous to the ( [ tens ] ) @xmath81_{jm}.\ ] ] for plane waves the radial vertex is @xmath82_{jm}{f_{n}(q^2)\over\mu}f(q^2)\ ] ] the distortion of the incoming and outgoing waves has been taken into account via inelastic distortion factor @xcite .
with this factor the @xmath83het vertex becomes @xmath84 { f_{n}(q^2)\over\mu}\exp(\imath{\bf qr})\exp(-{1\over2}\chi_{in}({\bf r } _ { \perp},{\bf q } ) ) .
\label{dvert}\ ] ] the distortion factor @xmath85 is @xcite @xmath86 where @xmath87 is the thickness function @xmath88 and @xmath89 is the target density . the @xmath90 is related to the elastic nucleon nucleon scattering amplitude at given energy per nucleon and @xmath91 is the wave function of the @xmath1he or the triton depending on the internal coordinates @xmath92 .
the value of @xmath93 used in calculations is @xmath94 @xcite .
two features of the distortion factor ( [ chi ] ) should be mentioned .
first , the @xmath1het form factor can not be separated from the effects of distortion .
second , since the vertex ( [ dvert ] ) has a gradient coupling and the distortion factor ( [ chi ] ) depends on the transversal coordinates , some transversal components arise in the reaction amplitude even if it was before pure longitudinal amplitude . the multipole expansion of the distorted vertex ( [ dvert ] ) can be obtained directly by multiplying it on @xmath95 , taking trace over spin matrices , and integrating over angles of the unit radius vector @xmath96 .
@xmath97 the separate multipoles contribute independently into cross
section ( [ ddsec ] ) , which becomes @xmath98 similar expansion exist for the coherent pion contribution ( [ cohpi ] ) . for numerical calculation
it is convenient to define the function @xmath99 which is the pion field at the reaction point generated by the source @xmath100 .
the integration in ( [ int ] ) is not well defined numerically since the integrand is oscillating function that is not decreasing at infinity .
the indirect integration has been done in the following way .
the function @xmath101 satisfies the integro differential equation @xmath102 which was solved numerically using the condition for feynman propagator @xmath103 at positive energy . since @xmath104 has an outgoing wave at infinity it fixes the solution of the equation ( [ gout ] ) .
the cross section expressed in terms of @xmath101 is @xmath105 in the expression ( [ dsecw ] ) the integration over coordinates goes effectively in a finite range , inside the target nucleus .
the nucleon single particle potential used for the wave functions of the bound nucleons has been taken in the standard woods saxon form .
@xmath107 where @xmath108 , @xmath109 is the pion compton wavelength , and @xmath110 is the coulomb potential for protons that was taken as the potential of a uniformly charged sphere .
the parameters of the potential are listed in the table 1 .
the response function ( [ fresp ] ) were found not very sensitive to the parameters of the nucleon potential .
the situation is , however , different for the optical @xmath0 nucleus potential that has been taken in similar woods saxon form .
@xmath111 where @xmath112 are the spin @xmath113 matrices .
2 shows two triton spectra for the @xmath0h contribution demonstrating sensitivity to the real part of the @xmath0nucleus optical potential . in the absence of the @xmath0nucleus interaction the peak position coincides with the one in the reaction on free proton . the attraction produced by the real part of the potential in the final state increases cross - section and shifts the peak position on 15 - 20 mev down .
the parameters of the optical potential are listed in table 2 .
they were taken mainly from @xcite , @xcite , and @xcite .
the radius @xmath114 and the diffuseness @xmath115 were kept the same as for the nucleons .
the main feature of the pionic self energy near the resonance is its large imaginary part .
it is instructive to study separately the effects of real and imaginary parts of the self
energy on the triton spectra .
fig.3 shows three spectra where either real or imaginary part of the self energy were accounted in comparison to the quasifree case .
the real part of the self energy is attractive and it brings more strength to the @xmath0-hole peak . in analogy with particle
hole excitations one can say the real part of the pionic self energy makes the @xmath0-hole peak more collective .
the peak position , however , changes not much . the imaginary part , takes all this collectivity back decreasing cross section due to incoherent @xmath0 decay so , when both parts are taken into account the cross section appeared close to its quasifree magnitude as it is shown in fig.3 and fig.4 . in fig.3 one
can see that the imaginary part of the self energy produces also some shift in the peak position .
the origin of the shift is , however , different from the shift due to the real part . to understand its origin
let us return to nuclear matter and compare two expressions for the cross - sections with and without renormalization of the pion propagator . without renormalization the cross
section is proportional to @xmath116 while in the other case it is proportional to @xmath117 for simplicity the real part is omitted in this expression .
the first case corresponds to quasifree mechanism and the peak position is at the same place as in the reaction on a proton . in the second case
the cross - section starts to grow at the same threshold but at the resonance where @xmath118 we have the cross
section proportional to @xmath119 thus , instead of maximum the cross
section appeared to be small at this energy .
since the cross section is growing from threshold one can immediately conclude that the maximum of the cross - section will be below the resonance position .
fig.5 demonstrates this effect for the @xmath120 multipole in the case of finite nucleus .
this effect is not so much pronounced for higher multipoles that determine the magnitude of the cross section .
the higher multipoles undergo smaller medium effects since they are peaked at nuclear surface where the density is small .
therefore , the overall shift of the peak position is smaller than for @xmath120 .
the imaginary part of the optical @xmath0 - nucleus potential makes this effect stronger as it is shown in fig.6 .
the contribution of separate multipoles to the triton spectrum is shown in the fig.7 .
the contribution of the low multipoles @xmath120 and @xmath121 is almost negligible due to strong absorption of the incoming and outgoing ions .
the main contribution comes from the multipoles between @xmath122 to @xmath123 although higher multipoles , at least up to @xmath124 , have to be cosidered at higher excitation energy .
another feature clearly seen in the fig.7 is rather wide spreading of the different multipole contributions .
the @xmath122 contribution is most sensitive to the medium effects shifting down the transition strength . it has the largest downward shift in the peak position .
the absorption of the @xmath1he and t is smaller for @xmath122 compared to @xmath120 resulting in sizeable contribution to the cross section .
higher multipoles have smaller medium effects and their peak positions are at more and more high excitation energies .
this produce large displacement width of summed triton spectrum .
fig.8 shows the final triton spectrum toghether with separate contributions of the @xmath0hole excitations and the coherent pion production .
the process of coherent pion production gives sizeable contribution to the inclusive triton spectrum .
the maximum of the cross
section is at about 240 mev excitation energy and it also contributes to the shift of the inclusive peak .
the coherent pion production is absent in infinite nuclear matter , therefore one should expect decrease of the relative yield of coherent pions for heavier nuclei .
for ( @xmath1he , t ) reaction in the @xmath0-region strong deviations from impulse approximations were demonstrated .
the deviations come from the medium effects of renormalization of the pion propagator in the ope mechanism of the elementary charge
exchange reaction .
the medium effects change both the peak position and its height .
several effects , besides pion renormalization , contribute to the shift of peak position including @xmath0nucleus optical potential , and coherent pion production .
the possible effects of collectivity in the @xmath0hole excitations are strongly suppressed by incoherent @xmath0 decay .
the effects of virtual pion propagation manifest itself only in the process of coherent pion production . the finite size of a target nucleus produces together with the medium effects large spreading of the observed peak .
the absorption in initial and final states strongly supresses the lowest multipols of the angular momentum transfer .
the glauber approach to the distortion in initial and final states gives reasonable description both the size of the medium effects and the absolute value of the cross section . c. ellegaard et al .
* 50 * ( 1983 ) 1745 .
ableev et al . , jetp lett .
* 40 * ( 1984 ) 763 .
d. contardo et al.,phys
168b * ( 1986 ) 331 . c. gaarde , annu .
nuc . part .
sci.,*41 * ( 1991 ) 187 . g. chanfray and m.ericson , phys . lett .
* 141b * ( 1984 ) 163 .
dmitriev and toru suzuki , nucl . phys .
* a438 * ( 1985 ) 697 .
dmitriev , yad . fiz .
* 46 * ( 1987 ) 770
hong , f. osterfeld and t. udagawa , phys .
* 245b * ( 1990 ) 1 .
j. delorm and p.a.m .
guichon , phys . lett .
* 263b * ( 1991 ) 157 .
e. oset , e. shiino and h. toki , phys . lett .
* 224b * ( 1989 ) 249 . h. esbensen and t .- s.h .
lee , phys .
rev . * c32 * ( 1985 ) 1966 .
dmitriev , phys . lett .
* 226b * ( 1989 ) 219 .
e. ferrari , f. selleri , phys.rev .
* 7 * ( 1961 ) 387 .
g. wolf , phys .
* 182 * ( 1969 ) 1538 .
dmitriev , o.p .
sushkov and c. gaarde , nucl.phys .
* a459 * ( 1986 ) 503 .
vervest , phys.lett . *
83b * ( 1979 ) 161 .
karmanov et al . , sov.journ.nucl.phys . *
18 * ( 1973 ) 1133 .
p. desgrolard , j. delorme and c. gignoux , preprint inst
. de phys .
nucl . de lyon , lycen 9205 ( 1992 ) .
olsson , e.t.osypowsky , nucl.phys .
bf b101 ( 1975 ) 136 . m. ericson and t.e.o
. ericson , ann .
* 36 * ( 1966 ) 323 .
h. lee and k. ohta , phys.rev . c*25 * ( 1982 ) 3043 .
j. h. koch , e. j. moniz and n. ohtsuka , ann . of phys .
* 154 * ( 1984 ) 99 .
stephen j. wallace , in advances in nuclear physics , plenum press , new york , vol .
12 , 1981 . | the excitation of a @xmath0-isobar in a finite nucleus in charge
exchange ( @xmath1he , t ) reaction is discussed in terms of a nuclear response function .
the medium effects modifying a @xmath0- and a pion propagation were considered for a finite size nucleus .
the glauber approach has been used for distortion of a @xmath1he and a triton in the initial and the final states .
the effects determining the peak positions and its width are discussed .
large displacement width for the @xmath0 - h excitations and considerable contribution of coherent pion production were found for the reaction on @xmath2c .
iu / ntc 9234 november 1992 + + in the ( @xmath1he , t ) reaction nuclear theory center , indiana university , bloomington in 47408 + + budker institute of nuclear physics , novosibirsk , 630090 , russia |
the precise study of the hadron spectroscopy would help us understand the substructure of observed hadrons and verify the theory , quantum chromodynamics ( qcd ) , which describes the strong interactions of colored quarks and gluons .
qcd predicts the new particles beyond the quark model , _
e.g. _ , glueballs , hybrids and multiquark states .
experimental study of the decay modes of known hadrons and the search of new hadrons would provide helpful information for further verifying the qcd theory . in the last decades , considerable progresses on the hadron spectroscopy have been made .
since the upgrade was completed in 2008 , the besiii @xcite detector at the bepcii collider , has collected the world s largest data samples at @xmath7-charm energy region , including 1.3 billion @xmath1 events , 4.5 million @xmath2 events , 2.9fb@xmath8 at the peak of the @xmath3(3770 ) resonance , and lots of data samples above 4.0gev/@xmath9 , which offers us a unique opportunity to study the hadron spectroscopy and search for the new hadrons .
the article reviews several structures around 1.85gev/@xmath9 found or confirmed by besiii , the particle wave analysis results of @xmath10 , the study of @xmath11 and @xmath12 anomalous decays , and the observation of new excited baryons .
in 2003 besii reported the first observation of @xmath13 mass threshold enhancement in @xmath14 , which attracted both experimental and theoretical attentions to investigate its nature . using the 225@xmath15 events collected by the besiii detector in 2009 , a partial wave analysis ( pwa ) of @xmath16 @xcite with @xmath172.2 gev/@xmath18
was performed to determine its spin - parity , mass and width . in the analysis ,
the final state interactions were considered by including the julich formulation @xcite .
the pwa results indicate that the @xmath19 assignment fit for this structure is better than that for @xmath20 or other @xmath21 assignments .
the mass , width and product branching fraction of the @xmath22 are measured to be @xmath23 = @xmath24 ( model ) mev/@xmath9 , @xmath25(model)mev ( a total width of @xmath26mev at the 90% c.l . ) and @xmath27(model))@xmath28 , respectively .
we also observed the structures around 1.85gev in the other @xmath1 decays .
the @xmath29(1835 ) was observed from the invariant mass of @xmath30 in @xmath31 decay with a statistical significance of @xmath32 by the besii experiment @xcite .
the same process was studied at besiii , which confirmed @xmath29(1835 ) with the statistical significance larger than 20@xmath33 @xcite . to further understand the nature of @xmath13 mass threshold structure ( @xmath29(1860 ) and @xmath29(1835 ) ) , the decay of @xmath34 was studied @xcite . the process of @xmath35 was first observed with the signal significance estimated to be 7.2@xmath33 . the @xmath36 and @xmath37 were also clearly observed in the @xmath38 mass spectrum , as shown in fig .
[ fig:4plots](a ) . to understand the nature of above mentioned particles , @xmath22 , @xmath29(1835 ) and @xmath40 , further study
is strongly needed , in particular , in searching for new decay modes . since the @xmath41 was confirmed to be a pseudoscalar particle @xcite and it may have properties in common with the @xmath42 .
six charged pions is a known decay mode of the @xmath42 .
therefore , @xmath1 radiative decays to 3(@xmath43 ) may be a favorable channel to search for the @xmath29 states in the [email protected]/@xmath18 region .
[ fig:4plots](b ) shows the 3(@xmath43 ) invariant mass spectrum in @xmath45 .
a structure at 1.84gev / c@xmath46 was observed with a statistical significance of 7.6@xmath33 .
the mass and width were measured to be @xmath23 = @xmath47 mev/@xmath18 and @xmath48mev .
the product branching fraction was determined to be @xmath49 .
a study of the doubly ozi suppressed decays of @xmath50 @xcite was performed .
a strong deviation(@xmath51 30@xmath33 ) from three - body phase space for @xmath50 near the @xmath52 invariant - mass threshold was observed , as shown in fig .
[ fig:4plots](c ) .
assuming the enhancement is due to the influence of a resonance , the @xmath29(1810 ) , a partial wave analysis with a tensor covariant amplitude determined that the spin - parity of the @xmath29(1810 ) is @xmath20 .
the mass and width of the @xmath29(1810 ) were determined to be @xmath5319(model ) mev/@xmath18 and @xmath54 = 95@xmath5510@xmath5675(model ) mev and the product branching fraction was measured to be @xmath57 = ( [email protected]@xmath581.30(model))@xmath5910@xmath60 . these results are consistent within errors with those from the besii experiment @xcite . the comparison to the mentioned besiii results of the masses and widths are shown in fig .
[ fig:4plots](d ) .
the mass of @xmath29(1840 ) is in agreement with @xmath22 , while its width is significantly broader .
therefore , based on these data , one can not determine whether @xmath29(1840 ) is a new state or the signal of a 3(@xmath43 ) decay mode of @xmath22 .
further study , including an amplitude analysis to determine the spin and parity of the @xmath29(1840 ) , is needed to establish the relationship between these experimental observations .
m - etapipi ( 23,75)(a ) m - kkpipipi0 ( 21,78)(c ) m-6pi ( 23,60)(b ) comp ( 23,60)(d )
radiative @xmath1 decay is a gluon - rich process and has long been regarded as one of the most promising hunting grounds for glueballs . in particular , for a @xmath1 radiative decay to two pseudoscalar mesons , it offers a very clean laboratory to search for scalar and tensor glueballs because only intermediate states with @xmath61 are possible .
an early study of @xmath10 was made by the crystal ball collaboration @xcite with the first observation of @xmath62(1710 ) , but the study suffered from low statistics . using 225 million @xmath1 events collected with the besiii detector , a pwa of @xmath10 has been performed .
[ fig:4pi](a ) shows the projection of pwa in the invariant mass spectrum of @xmath63 .
the scalar contributions are mainly from @xmath62(1500 ) , @xmath62(1710 ) and @xmath62(2100 ) , while no evident contributions from @xmath62(1370 ) and @xmath62(1790 ) are seen .
recently , the production rate of the pure gauge scalar glueball in @xmath1 radiative decays predicted by the lattice qcd @xcite was found to be compatible with the production rate of @xmath1 radiative decays to @xmath62(1710 ) ; this suggests that @xmath62(1710 ) has a larger overlap with the glueball compared to other glueball candidates ( _ e.g. _ @xmath62(1500 ) ) . in this analysis , the production rate of @xmath62(1710 ) and @xmath62(2100 ) are both about one order of magnitude larger than that of the @xmath62(1500 ) and no clear evidence was found for @xmath62(1370 ) , which are both consistent with , at least not contrary to , lattice qcd predictions .
the tensor components , which are dominantly from @xmath64(1525 ) , @xmath65(1810 ) and @xmath65(2340 ) , also have a large contribution in @xmath10 decays . the significant contribution from @xmath64(1525 )
is shown as a clear peak in the @xmath63 mass spectrum ; a tensor component exists in the mass region from 1.8gev/@xmath18 to 2gev/@xmath18 , although we can not distinguish @xmath65(1810 ) from @xmath65(1910 ) or @xmath65(1950 ) ; and the pwa requires a strong contribution from @xmath65(2340 ) , although the possibility of @xmath65(2300 ) can not be ruled out .
the @xmath6 system provides a unique stage for understanding the distinct symmetry - breaking mechanisms .
furthermore , the @xmath6 decays play an important role to explore the effective theory of qcd at low energy , especially for the chiral perturbation theory ( @xmath66pth ) .
its main decay modes , including hadronic and radiative decays , have been well measured , but the study of anomalous decays is still an open field .
besiii had collected 1.3 billion @xmath1 events , one can obtain large @xmath6 samples ( @xmath67 ) from processes @xmath68 or @xmath69 .
it is a good place to study the anomalous decays of @xmath6 .
besiii had published results of @xmath6 invisible decays @xcite , weak decays of @xmath70 @xcite and @xmath71 @xcite . here
we present other two studies briefly .
the @xmath12 meson is much heavier than the goldstone bosons of broken chiral symmetry , and it has a special role in hadron physics because of its interpretation as a singlet state arising due to the axial @xmath74(1 ) anomaly @xcite . discovered in 1964 @xcite , it remains a subject of extensive theoretical studies aiming at extensions of chiral perturbation theory @xcite .
based on a sample of 1.3 billion @xmath1 events taken with the besiii detector , we observe the decay modes @xmath72 and @xmath73 with a statistical significance of 18@xmath33 and 5@xmath33 , respectively , as shown in fig . [ fig:4pi](b ) and fig .
[ fig:4pi](c ) .
we measure their product branching fractions : @xmath75 = ( [email protected]@xmath550.30)@xmath76 and @xmath77 = ( [email protected]@xmath550.89)@xmath76 . using the pdg world average value of @xmath78 @xcite , the branching fractions of @xmath79 are determined to be @xmath80 = ( [email protected]@xmath550.64)@xmath81 and @xmath82 = ( [email protected]@xmath550.18)@xmath83 , which are consistent with the theoretical predictions based on a combination of chiral perturbation theory and vector - meson dominance @xcite , but not with the broken su@xmath84 o@xmath85 quark model @xcite . m-2eta ( 80,70)(a ) m_4pi_new ( 80,70)(b ) m_2pipi0_new ( 80,70)(c ) the decay @xmath87 , in the frame of @xmath66pth , both at @xmath88 and @xmath89 ( @xmath90 is the typical four - momentum transfer in the decay , as defined in ref .
@xcite ) , is forbidden because there is no direct coupling of photons to the neutral @xmath91 and @xmath11 . the second order @xmath89 , is much suppressed due to the kaon masses from the loops involving kaons , or suppressed due to the g - parity violating transitions from the pion loops .
the first sizable contribution comes at @xmath92 @xcite .
since the coefficients involved at @xmath92 are not precisely determined , different models have been used to obtain the coefficients and predict the decay rate of @xmath87 in the framework of @xmath66pth .
this decay provides therefore a unique opportunity to test directly the correctness of the calculations of third order @xmath66pth , since the first order is zero and the second order is very small . besides
, the branching fraction of @xmath87 decay has been also calculated with pure vector meson dominance model ( vmd ) @xcite , the quark - box diagrams @xcite and the coherent sum of the linear @xmath33 model and vmd contributions @xcite , respectively . with a sample of 1.3 billion @xmath1 events in the besiii detector ,
the rare , doubly radiative decays @xmath86 have been studied .
the branching fraction of @xmath93 is measured for the first time to be @xmath94 = ( [email protected]@xmath550.58)@xmath83 .
the measured value is much lower than that of the theoretical predictions @xcite but consistent with the upper limit in pdg @xcite .
no evidence for the decay of @xmath87 is found .
although symmetric non - relativistic three - quark models of baryons are quite successful in interpreting low - lying excited baryon resonances , they tend to predict far more excited states than are found experimentally @xcite . from the theoretical point of view , this could be due to a wrong choice of the degrees of freedom , and models considering di - quarks have been proposed @xcite .
experimentally , the situation is very complicated due to the large number of broad and overlapping states that are observed .
moreover , in traditional studies using tagged photons or pion beams , both isospin @xmath95 and isospin @xmath96 resonances are excited , further complicating the analysis .
an alternative method to investigate nucleon resonances employs decays of charmonium states such as @xmath1 and @xmath2 .
[ [ observation - of - two - new - n - resonances - in - psirightarrow - pbarppi0 ] ] observation of two new @xmath5 resonances in @xmath97 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ to search for new excited @xmath5 baryons , we performed a pwa of @xmath97 , to study the @xmath5 intermediate resonance coupling to @xmath98 or @xmath99 .
the @xmath100 resonances are excluded due to isospin conservation . as a consequence ,
the reduced number of states greatly facilitates the analysis @xcite . after a partial wave analysis ,
resonances with significance greater than 5@xmath33 are taken as significant ones , including @xmath101(940 ) and seven @xmath5 resonances .
the mass and width of @xmath5 states are varied , and the values with the best fitting result are taken as the optimized values . table [ table - n * ] lists the optimized values for the seven @xmath5 states . in this table ,
the first five @xmath5 resonances are consistent with the values in the particle data book @xcite , while the last two states can not be identified with @xmath102 or @xmath103 . however , the significance of these two states are 15@xmath33 and 11.7@xmath33 , respectively . as a consequence
, we label these two states as @xmath104 and @xmath105 , with @xmath106 assignment of @xmath107 and @xmath108 , respectively .
using these eight significant resonances , the fit result agrees well with the data .
the contribution of each intermediate resonance including interference effects with other resonances are extracted .
[ fig : n*-ppbarpi0](a ) shows the contributions of @xmath101(1440 ) , @xmath101(1520 ) , @xmath101(1535 ) and @xmath101(1650 ) in which we can see clear peaks and also tails at the high mass region from the interference effects .
[ fig : n*-ppbarpi0](b ) shows the contributions of @xmath101(940 ) , @xmath101(1720 ) , @xmath101(2300 ) and @xmath101(2570 ) . for @xmath101(2300 ) and @xmath101(2570 ) , their peak positions are below the breit - wigner mean values reported in table [ table - n * ] because of the presence of interference contributions , as well as phase space and centrifugal barrier factors .
.the optimized mass , width and significance ( sig . ) of the seven significant @xmath5 resonances .
@xmath109 represents the change of the log likelihood value .
@xmath110 is the change of the number of free parameters in the fit . in the second and third columns ,
the first error is statistical and the second is systematic .
the names of the last two resonances , n(2100 ) and n(2200 ) , have been changed to @xmath101(2300 ) and @xmath101(2570 ) according to the optimized masses . [
cols="^,^,^,^,^,^",options="header " , ] using the 106 million @xmath2 events taken at the besiii detector , a full pwa on the 745 @xmath111 candidates is performed .
the pwa results , including the invariant mass spectra of @xmath13 , @xmath112 , @xmath113 are shown as histograms in fig .
[ fig : ppeta ] , and are consistent with the data .
the results indicate that the dominant contribution is from @xmath114 .
the mass and width of @xmath115 are determined to be @xmath116mev/@xmath18 and @xmath117 mev , respectively , which are consistent with those from previous measurements listed in the pdg @xcite .
the product of the branching fractions is calculated to be @xmath118 = ( @xmath119 . the @xmath13 mass enhancement observed by besii @xcite
is investigated , and the statistical significance of an additional @xmath13 resonance is less than 3@xmath33 .
based on 106 million @xmath2 events collected with the besiii detector , the decays @xmath121 and @xmath122 are analyzed .
the @xmath123 and @xmath124 invariant mass spectra , shown in fig .
[ fig : lbdsgmpi](a ) and fig .
[ fig : lbdsgmpi](b ) , indicate @xmath125 and @xmath126 structures , @xmath127 peaks around 1.4gev/@xmath18 to 1.7gev/@xmath18 in the invariant mass distributions of @xmath123 and @xmath124 , that clearly deviate from what is expected according to phase space . as shown in fig .
[ fig : lbdsgmpi ] , the background contamination is small and is ignored . sixteen possible intermediate excited states
are included in the pwa .
a comparison of the data and global fitting results , shown in fig .
[ fig : lbdsgmpi ] , indicates that the pwa results are consistent with data .
a similar pwa is also performed for the decays @xmath128 , and the results are also in agreement with data . , ( a ) @xmath129 , ( b ) @xmath130 and ( c ) @xmath131 .
points with error bars are data , the solid histograms are pwa projections , the dashed histograms are phase space distributions from mc simulation , and the shaded histograms are the background contributions estimated from the @xmath132 sidebands .
, title="fig:",scaledwidth=30.0% ] , ( a ) @xmath129 , ( b ) @xmath130 and ( c ) @xmath131 .
points with error bars are data , the solid histograms are pwa projections , the dashed histograms are phase space distributions from mc simulation , and the shaded histograms are the background contributions estimated from the @xmath132 sidebands . , title="fig:",scaledwidth=30.0% ] , ( a ) @xmath129 , ( b ) @xmath130 and ( c ) @xmath131 .
points with error bars are data , the solid histograms are pwa projections , the dashed histograms are phase space distributions from mc simulation , and the shaded histograms are the background contributions estimated from the @xmath132 sidebands . , title="fig:",scaledwidth=30.0% ] | the besiii experiment has collected world s largest direct @xmath0 colliding beam data at the @xmath1 , @xmath2 , @xmath3(3770 ) and at and around the @xmath4(4260 ) .
these data have enabled the study of hadrons at unprecedented precisions .
we present results on the light hadron spectroscopy , production of hadrons in radiative decays of charmonia , observation of new excited @xmath5 baryons and of anomalous decays of the @xmath6 states . |
already in the early 1960s it became clear that the strongest radio sources associated with disk galaxies are by about 3 orders of magnitude weaker than the strongest radio sources hosted by giant elliptical galaxies , and that the most radio luminous objects in the universe are quasars @xcite .
soon it was realized , however , that the majority of quasars is radio quiet @xcite .
initial optical imaging indicated that their radio - bimodality may be related to the host galaxy morphology , with radio quiet quasars preferentially hosted by disk galaxies and radio - loud quasars hosted by elliptical galaxies @xcite .
this suggested a possible association of radio quiet - quasars with seyfert galaxies and radio - loud quasars with radio galaxies .
such a bimodality is visualized by @xcite in the radio - luminosity versus [ oiii]-luminosity representation . since radio structures in agns
are powered by jets , the agn radio dichotomy is most likely related to the efficiency of a jet production . according to the spin paradigm ,
this efficiency depends on the value of the black hole spin @xcite . in the simplest version of this paradigm
the observed radio dichotomy is directly related to the cosmologically determined distribution of bh spins .
assuming that the growth of supermassive bhs is dominated by their mergers , @xcite demonstrated that the excess of radio quiet quasars could be explained by the rarity of major galaxy mergers , resulting in a ` bottom - heavy ' distribution of bh spins .
this is because in such a scenario , fast rotating bhs can only be produced by mergers of two bhs with comparable masses , and this in turn implies similar masses of merging galaxies . however , as was pointed out by @xcite , the growth of bhs in agns is very likely dominated by accretion processes , which are known to be able to spin up bhs very efficiently @xcite .
moderski & sikora argued nevertheless that the spin paradigm could still be at work , and showed that the bottom - heavy distribution of the bh spin might be obtained provided the accretion history of most of agns was marked by many small - mass accretion events with randomly oriented angular momenta ( see also * ? ? ?
such events , due to bardeen - petterson effect @xcite , lead to the formation of co- and counter - rotating disks and , in consequence , to bh spins fluctuating around zero ( with very small amplitudes ) . since during their evolution disk galaxies have avoided major mergers @xcite , bhs in spiral - hosted agns had much larger chance avoiding spinning - up by massive accretion events than their cousins in giant ellipticals .
however , these simple versions of the spin paradigm were challenged , both from the observational and theoretical perspectives .
observationally , this is because : ( i ) luminous quasars , irrespective on their radio - loudness , were recently found to be hosted by giant ellipticals ( see floyd et al .
2004 , and references therein ) ; ( ii ) several independent investigations of the agn accretion radiation efficiency using the sotan s type of argument @xcite indicate that the majority of quasars are powered by fast rotating bhs @xcite ( but see * ? ? ?
* ) ; ( iii ) low luminosity agns ( llagns ) , including local seyfert galaxies and liners , are found to have higher than previously thought radio - to - optical nuclear flux ratios , placing them rather in the category of radio loud objects @xcite .
theoretically , this is because : ( iv ) the blandford - znajek mechanism was claimed to be not efficient enough to explain jet energetics in the most radio - loud quasars @xcite ; ( v ) the possibility of formation of counter - rotating accretion disks in agn systems was questioned @xcite .
in addition , recent investigations of the jet ( radio ) activity in xrbs could suggest that the accretion rate is the only parameter controlling the jet production efficiency in these systems @xcite . as a consequence of all the above , the ` main stream '
agn models in the last years became those with a jet production related to accretion processes exclusively @xcite .
does this mean that the spin paradigm is ` dead ' ?
not at all . as was shown and discussed by (
* hereafter ssl07 ) and @xcite , several challenges listed above are not well justified , while other can be overcome after adopting certain modifications to the spin paradigm . in particular , @xcite demonstrated analytically and @xcite confirmed numerically that the formation of counter rotating disks in agns is possible . also , @xcite and @xcite showed ( using different gr mhd simulations ) that the efficiency of extraction of the bh rotational energy can be much larger than indicated by the formula derived originally by @xcite using 1st order perturbation method and adopting the @xcite disk model . finally , ssl07 demonstrated that the host - related radio bimodality of agns _ remains _ real at _ all _ accretion rates . on the other hand ,
modifications of the spin paradigm are indeed required in order to reconcile the observed radio - bimodality of powerful ( elliptical - hosted ) quasars with the requirement that all bhs in quasars have large spins . addressing this problem
, ssl07 proposed that the fact that most of the elliptical - hosted quasars are radio quiet results from the suppression ( quenching " ) of jet production at high accretion rates .
such a suppression is in fact directly observed in the transient xrbs : the observations of micro - quasar grs 1915 - 105 indicate for example that two different accretion modes may exist at high accretion rates , and that only during one of them the efficient jet production proceeds @xcite . in their modified version of the spin paradigm ,
ssl07 anticipated such a two accretion - mode scenario for the elliptical - hosted agns accreting at high ( eddington ) rates , assuming in particular that the quasar jets produced by the rotating supermassive bhs can avoid suppression only if the efficient collimation by the mhd outflows from the outer parts of the accretion disk is the case .
this article is based in its large parts on the work of ssl07 , and is organized as follows . in 2 , the dependence of the agn radio - loudness on the eddington ratio is presented ; in 3 , the multi - accretion - event scenarios are investigated ; in 4 , the challenges to the spin - paradigm are discussed and critically re - examined ; and in 5 the final conclusions are listed .
in order to quantify the jet production efficiency in agns , @xcite introduced a ` radio - loudness ' parameter defined as the ratio of radio ( @xmath0ghz ) and optical ( b - band ) spectral flux densities , @xmath1 .
based on the claimed bimodality of opticaly - selected pg quasars , they established @xmath2 as a borderline between the radio - loud and radio - quiet objects .
the methods of estimating masses of bh in galactic nuclei developed in the 1990s ( see * ? ? ?
* and refs . therein ) allowed to study the dependence of this radio - loudness parameter on the eddington ratio @xmath3 , which is defined as the ratio of the bolometric accretion luminosity to the eddington luminosity .
such a dependence was first presented by @xcite for an agn sample composed of low luminosity sources ( llagns : weak local seyfert galaxies and liners ) , ` classical ' seyferts selected from the palomar and cfa surveys , and pg quasars .
he showed that the radio - loudness increases with decreasing eddington - ratio and that , according to the kellermann s et al .
definition , practically all the llagns are radio - loud . around the same time , a trend for the increasing radio - loudness with the decreasing accretion luminosity was discovered also in the low / hard states of xrbs @xcite . following these results
, @xcite constructed a ` fundamental plane ' which attempts unifying the dependence of radio activity on accretion rate for all the bh ( i.e. , galactic and extragalactic ) accretion systems .
such studies have been extended by ssl07 by considering an agn sample enlarged by the addition of radio selected quasars , broad line radio galaxies ( blrgs ) and fr i radio galaxies . when compared with the objects considered previously by @xcite , the newly included sources form a separate pattern ( though of a similar shape ) on the @xmath4 plane , being in particular @xmath5 orders radio louder than llagns and seyferts with comparable accretion luminosity ( see figures 1 and 2 ) . at the first glance
, one might suspect that such a two - pattern structure is a result of selection effects .
however , noting that no agn in the upper , ` radio - loud ' branch is hosted by a disk galaxy , one can conclude that the two revealed upper and lower patterns represent , at least , the upper bounds for the radio - loudness of agns hosted by elliptical and disk galaxies , respectively .
the relatively complete samples of llagns and low luminosity radio galaxies analyzed by @xcite , @xcite , and @xcite , indicate that these are indeed the real distributions of the parameter @xmath6 in the case of low-@xmath3 objects .
ghz luminosity vs. b - band nuclear luminosity in the eddington units .
blrgs are marked by filled circles , radio - loud quasars by open circles , seyfert galaxies and linears by crosses , fri radio galaxies by open triangles , and pg quasars by filled stars ( from ssl07).,width=302 ] vs. eddington ratio @xmath3 .
different types of agns are denoted in the same way as in fig .
1 ( from ssl07).,width=302 ] the situation changes for the high - accretion rate ( high-@xmath3 ) sources , represented by quasars and nls1 galaxies . in particular
, the radio - loudness distribution of quasars is very broad and ` bottom - heavy ' , despite that probably most of them are hosted by giant ellipticals @xcite .
this can not be entirely due to selection effects .
hence , the significant differences in the jet production efficiency in these objects , as observed in figures 1 and 2 , seems not to be related to the morphologies of their hosts .
we note , that a bimodal distribution ( instead of a continuous one ) of the radio - loudness in quasar sources was claimed by @xcite , @xcite , and @xcite , although the most recent studies on this issue , based on the deep radio and optical surveys , are not very conclusive @xcite . at the intermediate eddington - ratios ,
agns hosted by giant elliptical galaxies and located in the lower ( ` radio - quiet ' ) sequence of figures 1 and 2 , are represented in our sample by only four objects .
however , recent discoveries of many elliptical galaxies with very broad balmer lines and very massive black holes but very weak radio emission @xcite strongly indicate that rarity of such objects in the available agn catalogs might be due to selection effects . hence , it is plausible that also at the intermediate accretion luminosities most of agns hosted by giant ellipticals are radio - quiet .
ssl07 studied also the dependence of the radio - loudness parameter on the bh mass .
the results are presented in figure 3 .
one can see that agns with the black hole masses @xmath7 reach values of @xmath6 three orders of magnitude larger than agns with black hole masses @xmath8 .
a relatively smooth transition between those two populations is caused , most likely , by the overlap between the sources hosted by disc and elliptical galaxies .
the errors in the black hole mass estimations can also have a similar effect .
it is then interesting to compare figure 3 with the analogous figures restricted to high eddington - ratio objects , and presented by @xcite and @xcite .
one can see that in all the cases there is a difference of about @xmath9 orders of magnitude between the maximal radio - loudness of agns with @xmath10 , and those with less massive bhs .
however , because in the sample studied by ssl07 the objects with very low values of the @xmath3 parameter are included as well , the boundaries of maximal radio - loudness for lower and higher mass bhs are now located at much larger values of @xmath6 .
this effect is a simple consequence of the increasing radio - loudness with the decreasing eddington - ratio . because of this
, the upper boundaries on figure 3 are determined by low-@xmath3 objects , i.e. , by seyferts and liners at low bh masses , and by fr i radio galaxies at high values of @xmath11 . vs. black hole mass @xmath11 in the solar units .
different types of agns are denoted in the same way as in fig .
1 ( from ssl07).,width=302 ] let us mention in this context the case of nearby galaxies which show marginal ( if any ) signatures of the central ( accretion ) activity , and for which precise determinations of the profiles of the stellar distribution are available . in particular , the most recent studies led to the discovery of a strong correlation between the type of the surface brightness distribution in the nuclear parts of the galaxy and its radio luminosity .
namely , the ` core - galaxies ' i.e. , the ones with shallow central stellar density distributions are much radio - louder than the power - law / cuspy galaxies @xcite .
it is then interesting to note that the core - galaxies are on average much more massive than the power - law galaxies , being usually identified as giant ellipticals , while the power - law stralight profiles are common in disk galaxies or disky - ellipticals @xcite .
ssl07 investigated the possibility that the parameter responsible for the host - morphology related split of the upper radio - bounds of agns in the @xmath4 plane , as discussed above , is the spin of a central black hole , @xmath12 , where @xmath13 is the angular momentum of the black hole and @xmath14 .
the conditions required for that are : 1 .
bhs in disk galaxies should avoid efficient spinning - up by the accretion disks ; 2 . it should be possible to reconcile the spin paradigm with the mismatch between the spin and radio - loudness distributions of the high-@xmath3 agns hosted by elliptical galaxies .
as was demonstrated by @xcite , bhs can avoid too intensive spinning - up if their evolution is composed of many small - mass accretion events with randomly oriented angular momenta .
in such a case , provided that the event masses are small enough to avoid an alignment of the bh with distant portions of the accretion disk @xcite , the accretion proceeds via a comparable number of co- and counter - rotating disks . as a result
, the bh spin undergoes fluctuations around the zero value with a very low amplitude , or quickly drops to small values if starting from a large one ( see figures 4 & 5 ) . and
the accretion event masses @xmath15 ( from * ? ? ?
* ) .,width=302 ] and @xmath16 .
the number of accretion events is marked on the curves ( from * ? ? ?
* ) .,width=302 ] quantitatively , the condition to avoid the bh alignment during an accretion event of a mass @xmath17 is @xmath18 where @xmath19 , @xmath20 is the bardeen - petterson warp radius , i.e. the distance at which the accretion time scale is equal to the time scale of the lense - thirring precession @xcite , and @xmath21 is the schwarzschild radius . for the @xcite disk model , for example
, the warp is formed in the ` middle region of the disk ' , namely at @xmath22 where @xmath23 , @xmath24 is the accretion rate expressed in the eddington units @xmath25 , @xmath26 is the shakura - sunyaev parameter , @xmath27 is the viscosity related to the ` planar ' shear within the disk , and @xmath27 is the viscosity related to the ` vertical ' shear . according to @xcite , @xmath28 ) . with this
, the alignment mass reads as @xmath29 such an event - mass limit is severe but , on the other hand , consistent with the observations indicating very short lifetimes of individual accretion events in seyfert galaxies @xcite , and revealing random orientations of jets relative to the host galaxy axis @xcite .
it should be also noted that fuelling of agns in disk galaxies is presumably not related to the galaxy mergers , and can be provided by molecular clouds @xcite .
alternatively , low values of bh spins in such systems could be assured if the bh growth thereby is dominated by mergers with intermediate mass bhs relics of population iii stars , or bhs formed in young stellar clusters ( see * ? ? ? * and references therein ) .
in contrast to spiral galaxies , giant ellipticals underwent at least one major merger in the past ( see , e.g. , * ? ? ? * ) .
such mergers are followed by accretion events which involve too much mass to satisfy the condition given by equation 3 .
then , regardless of whether the accretion disk was initially counter- or co - rotating , all the disks end up as co - rotating due to the alignment process .
provided that @xmath30 , such disks spin - up black holes to large values of @xmath31 , very likely up to even @xmath32 if only @xmath33 @xcite .
this scenario is in agreement with the large average spin of bhs in quasars as inferred from the comparison between the local bh mass density and the amount of radiation produced by luminous quasars @xcite .
however , since the majority of quasars is radio - quiet , the jet production in most of them should be suppressed .
such a bimodality of a jet suppression at high accretion rates can be connected with a bimodality of a jet collimation , and this , in turn , with a bimodality of the accretion states .
the collimation of a central relativistic outflow can be provided by a non - relativistic mhd wind / jet produced by an accretion disk .
such a double jet structure was originally proposed by @xcite , and was investigated further by @xcite , @xcite , and @xcite .
the quasar systems with such efficiently collimated jets can be rare indeed due to the difficulties in developing the required large - scale poloidal magnetic fields in the geometrically thin accretion disks .
such fields may , however , develop stochastically , leading to an intermittent jet production @xcite .
alternatively , large - scale poloidal fields may be carried to the inner portions of the disks from larger distances by the drifting isolated patches of the accreting matter @xcite .
recent studies of the dependence of the radio - loudness on the eddington - ratio in agns and xrbs indicate that the jet - to - accretion power ratio increases with the decreasing accretion rate , and that at the high accretion rates the jet production is strongly suppressed in majority of objects @xcite . in addition , the previous claims that powerful extragalactic radio sources avoid disk galaxies @xcite have been confirmed , and , as shown by ssl07 , such a radio bimodality persists over an entire investigated range of the eddington - ratios . assuming that jets are powered by rotating bhs ,
the galaxy - morphology related radio - bimodality can be explained if bh spins in giant ellipticals are , on average , much larger than in disk galaxies .
the best and the most direct way to verify this conjecture is to measure bh spins in different types of galaxies .
first attempts are already undertaken using profiles of the fluorescent iron x - ray line emitted in some agns by the innermost portions of the accretion disks ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
unfortunately , the quality of the available data is too poor to disentangle the effects connected with the bh spin from the effects of a warm absorption @xcite , and/or the effects of a superposition of differently shaped spectra produced in different flux states @xcite .
we would like to point out that mcg-6 - 30 - 15 , considered often to be the best evidenced case showing the fast rotating bh in a seyfert galaxy , is hosted in fact by the e / s0 galaxy @xcite .
also , the eddington ratio in this system is very high , @xmath34 @xcite , and so the jet production thereby is likely to be suppressed , even if mcg-6 - 30 - 15 hosts indeed the fast rotating bh .
ssl07 and @xcite investigated possible evolutionary scenarios which may lead to the galaxy - morphology related bh spin bimodality of agns .
assuming that the growth of supermassive bhs is dominated by the accretion , they showed that , in order to keep small bh spins in the multi - accretion event scenario @xcite , the event masses must be smaller than @xmath35 .
such small - mass accretion events can be provided by stochastically captured molecular clouds @xcite .
these may form from a cold gas streaming / dropping onto the galaxy from cosmological filaments .
such an inflow is predicted to take place in galaxies with the dark matter halo masses smaller than @xmath36 @xcite , and is supposed to protect the disk galaxies against destruction by the multiple minor mergers @xcite .
the modified spin paradigm described above is consistent with the recent finding that the radio - loudness correlates with stellar brightness profiles in the nuclear portions of active galaxies @xcite .
namely , the inner regions of radio - loud galaxies display star deficient cores . such cores , in turn , reside preferentially in giant ellipticals ( see , e.g. , * ? ? ?
on the other hand , radio - quiet galaxies , including nearby low - luminosity seyferts , display instead cuspy brightness profiles . and
these are found preferentially in disk galaxies .
hence , noting that the core stellar nuclei result from merging bhs following galaxy mergers @xcite , balmaverde & capetti s discovery supports our conjecture that the galaxy - morphology related spin bimodality results from the different evolutionary tracks of bhs in disk and elliptical galaxies : in the former case being dominated by randomly oriented small - mass accretion events , in the latter case by massive accretion events which follow galaxy mergers .
in summary , we conclude that : 1 .
the maximal values of the radio - loudness parameter of agns hosted by giant elliptical galaxies are by @xmath37 orders of magnitude larger than of agns hosted by disc galaxies
both populations of spiral - hosted and elliptical - hosted agns show a similar dependence of the upper bounds of the radio - loudness parameter on the eddington ratio ; the radio - loudness increases with the decreasing eddington ratio , faster at the high accretion rates , and slower at the low accretion rates .
the very large , host - morphology related difference between the radio - loudness reachable by agns in disc and elliptical galaxies can be explained by the scenario according to which * the spin of a black hole determines the jet power ; * central black holes can reach large spins only in early type galaxies ( following major mergers ) , and not ( in a statistical sense ) in spiral galaxies .
the broad , ` bottom - heavy ' distribution of the radio - loudness in quasars is not related to the distribution of the bh spin ; however , it is still the bh spin which mediates launching of a jet and determines the upper bound of the radio - loudness , whereas the ( intermittent ) suppression of a jet production can be connected with the absence of the jet collimation by an mhd wind from an accretion disk .
.s . acknowledges support by the mein grant 1-p03d-00329 . m.s .
thanks kipac / slac for the hospitality during the completion of this manuscript . | active galactic nuclei ( agns ) form two distinct sequences on the radio - loudness eddington - ratio plane .
the ` upper ' sequence contains radio selected agns , the ` lower ' sequence is composed mainly of optically selected agns .
the sequences mark the upper bounds for the radio - loudness of two distinct populations of agns , hosted respectively by elliptical and disk galaxies .
both sequences show the same dependence of the radio - loudness on the eddington ratio ( an increase with decreasing eddington ratio ) , which suggests that another parameter in addition to the accretion rate must play a role in determining the efficiency of jet production in agns .
we speculate that this additional parameter is the spin of the black hole , assuming that black holes in giant elliptical galaxies have ( on average ) much larger spins than black holes in disc galaxies .
possible evolutionary scenarios leading to such a spin dichotomy are discussed .
the galaxy - morphology related radio - dichotomy breaks down at high accretion rates where the dominant fraction of luminous quasars being hosted by giant ellipticals is radio quiet .
this indicates that the production of powerful jets at high accretion rates is in most cases suppressed and , in analogy to x - ray binary systems ( xrb ) during high and very high states , may be intermittent .
such intermittency can be caused by switches between two different accretion modes , assuming that only during one of them an outflow from the central engine is sufficiently collimated to form a relativistic jet .
accretion disks , black hole physics , galaxies : active , galaxies : jets 98.54aj , 98.54 cm , 98.54gr , 98.58fd , 98.62js , 98.62mw , 98.65fz |
since gibbs @xcite established the ensemble theory of statistical mechanics , there has arisen the problem as to whether it has the ability to describe phase transitions @xcite .
let us investigate the problem through the famous example , i.e. , the conventional low-@xmath6 superconductivity , which can be described by the so - called bardeen - cooper - schrieffer ( bcs ) hamiltonian @xcite , @xmath7where @xmath8 denotes the energy relative to fermi level , @xmath9 the coupling strength , @xmath10 and @xmath11 ( @xmath12 and @xmath13 ) the destruction ( creation ) operators for the electrons with up and down spins , respectively . here
, the notation @xmath14 expresses that the hamiltonian is a function of @xmath10 , @xmath15 , @xmath12 , and @xmath16 . for bcs superconductivity
, what is the most important physically is that the hamiltonian @xmath14 will remain invariant under the transformation,@xmath17where @xmath18this invariance is known as gauge symmetry . as a direct consequence of this symmetry ,
one has@xmath19where @xmath20 denotes the gibbs ensemble average with respect to the hamiltonian @xmath14 , and @xmath21is the statistical density operator where @xmath22 with @xmath23 and @xmath24 being the boltzmann constant and temperature , respectively . to prove eq .
( [ zero ] ) , it is sufficient to heed that @xmath25 eq .
( [ zero ] ) indicates that the electron - pair amplitude @xmath26 can never become nonzero according to gibbs ensemble theory , in other words , there can not happen superconductivity for the bcs hamiltonian of eq .
( [ bcs ] ) within gibbs ensemble theory no matter how low the temperature is .
this proves definitely that gibbs ensemble theory has no ability to describe phase transitions .
even so , it is still hoped to do , as least as possible , some modification to gibbs theory so that it could describe phase transitions . up to now , there are two kinds of modification , they both arise from the theoretical studies on the so - called ising model @xcite , and become the popular and predominant believes about phase transitions in the current world of statistical physics .
the first is to introduce the thermodynamic limit in the end of calculation @xcite , that is,@xmath27where the @xmath28 means to take the thermodynamic limit ( @xmath29 , @xmath30 , and @xmath31 constant ) after performing the trace .
because@xmath32irrespective of the magnitudes of @xmath33 , @xmath34 , and @xmath35 ( @xmath14 is invariant under gauge transformation , and that is true independently of @xmath33 , @xmath34 , and @xmath36 , namely , the gauge symmetry has nothing to do with @xmath33 , @xmath34 , and @xmath35 . ) , one has@xmath37therefore , this kind of modification can not make gibbs ensemble theory have the ability to describe phase transitions .
indeed , the thermodynamic limit is important for calculating the statistical average of observable , but itself alone is insufficient to act as the physical criterion for phase transitions .
the second is to introduce both the thermodynamic limit and an auxiliary external field @xcite , @xmath38 where , as pointed out by huang @xcite , the limit @xmath39 must be taken after the thermodynamic limit @xmath29 . the purpose of introducing an auxiliary external field @xmath40 is to break the gauge symmetry of the system , as can be easily seen from the term,@xmath41since the gauge symmetry has been broken with regard to the entire hamiltonian , @xmath42 in so far as the auxiliary external field @xmath40 is infinitesimal but nonzero , the trace and the limit @xmath29 of eq.([pair ] ) become nonzero .
further , if the limit @xmath39 does not tend to zero , the electron - pair amplitude @xmath43 will get nonzero .
one thus concludes that there appear cooper pairs in the system , and that the system has gone into the superconducting phase .
however , this conclusion can not hold in physics , for the modification suffers the serious problem : the limit procedure employed by eq.([pair ] ) is physically equivalent to the scenario that an infinitesimal auxiliary external field is first added to the system so as to induce the symmetry of the system to break down , and taken off from the system finally .
but , in the actual situation , the inducement of such an external field is unnecessary , the symmetry itself breaks down spontaneously and does not need any help of external force .
anyhow , this kind of modification contains unphysical operations , and thus can not be accepted in principle .
here , it is also significant and worth while to give a brief discussion to the two - dimensional ising model because it can be solved exactly within the two modifications,@xmath44 where @xmath45 denotes the spin variable on site @xmath46 , the symbol @xmath47 means that the sites @xmath46 and @xmath48 are the nearest neighbors , and @xmath49 is the exchange coupling between a nearest - neighbor pair of spins .
as is well known , onsager @xcite proved rigorously that the specific heat of the ising hamiltonian @xmath50 is singular at @xmath51 within the first modification of gibbs ensemble theory .
it hints that the system might undergo a paramagnetic - ferromagnetic phase transition at @xmath6 in the thermodynamic limit .
however , just as eqs .
( [ zero1][zero2 ] ) , the magnetization @xmath52 , @xmath53is always equal to zero due to the fact that the parity symmetry of @xmath54 with respect to @xmath55 and @xmath56 has nothing to do with the magnitude of @xmath34 where @xmath34 represents the total number of sites .
( [ mising ] ) proves definitely that there is no paramagnetic - ferromagnetic phase transition at any temperature within the first modification of gibbs ensemble theory even if the specific heat is singular .
the above proof also shows that it can not be solved within the framework of the first modification whether or not the onset of the singularity at @xmath57 manifests a phase transition . in order to justify that the phenomenon occurring at @xmath58 is a phase transition , yang @xcite went beyond the first modification and turned to the second one , which was , in fact ,
suggested originally by ising himself @xcite in studying the one - dimensional ising model .
wonderfully , yang succeeded in proving rigorously that@xmath59where @xmath60 stands for an external magnetic field .
mathematically , it seems reasonable to say from eq .
( [ mnzero ] ) that a paramagnetic - ferromagnetic phase transition will occur at @xmath58 . nevertheless , as pointed out above , eq .
( [ mnzero ] ) imports unphysical operations from the introduction of the external field @xmath60 , and this leads to the result that the phase transition will not occur automatically but has to be driven and decided from outside the system , which directly contradicts the basic experimental fact that any phase transition occurs itself spontaneously .
therefore , the second modification , or rather ising s criterion , can not be used as a physical criterion to justify whether there exists a paramagnetic - ferromagnetic phase transition at @xmath58 .
it should be stressed again that it is the two rigorous works of refs .
onsager and @xcite that establish the two kinds of modification to the gibbs ensemble theory . on all accounts , the two kinds of modification must both be discarded , and it is necessary to modify or extend gibbs ensemble theory anew .
that is just the main purpose of this paper .
recently , gibbs ensemble theory has been extended by c. tsallis and his followers with the conception of nonextensive entropy tsallis1,tsallis2,tsallis3 , which is being under controversy . in this paper , as a personal review , we would like to suppose another possible extension of the gibbs ensemble theory so that it can provide a reasonable description to phase transitions and spontaneous symmetry breaking ( ssb ) . also , there is a macroscopic theory for phase transitions , i.e. , landau phenomenological theory @xcite .
phenomenological as it is , landau theory succeeds in providing a unified picture for all the second - kind phase transitions , e.g. , superconductivity , superfluidity , magnetism , and structural phase transitions .
physically , landau picture consists of two basic notions : order parameter and variational principle .
order parameter describes the degree of order of the system , it is zero in the disordered phase , and nonzero in the ordered one .
variational principle yields the equation of motion of order parameter and controls the evolution of order parameter with temperature , a system must arrive at the minimum of helmholtz or gibbs free energy in its stable state .
according to this picture , a system will evolve with the evolution of its order parameters , and produce phase transitions spontaneously , i.e. , without any help or drive from outside the system . in short , order parameter and variational principle are two characteristics of landau theory .
on the other hand , landau pointed out that a phase transition of the second kind reflects physically the change in symmetry of the system : the disordered phase has a higher symmetry than the ordered one .
he found that this change in symmetry can be described through the representations of the symmetry group , a phase transition corresponds to a representation transformation , different phases will realize different representations , respectively .
accordingly , landau established the relation between the order parameter and the representation . using this relation ,
it can be easily determined from the variational principle which representation will be realized at a certain temperature . representation transformation and spontaneous symmetry breaking are the other two characteristics of landau theory . to sum up ,
a system will select and realize automatically different representations at different temperatures according to the variational principle of order parameter .
that is the mechanism for phase transitions discovered by landau theory , it crystallizes landau s ideas about phase transitions : order parameter , variational principle , representation transformation , and spontaneous symmetry breaking .
theoretically , this mechanism gives a reasonable answer to the problem why a phase transition or ssb can happen at a certain temperature .
in applications , it also agrees quite well with various phase transitions , its effectiveness and universality are well known to condensed matter physicists . in a word ,
landau s ideas on phase transitions are successful and of great value in physics ! regrettably , it is impossible to deduce landau theory from gibbs ensemble theory because there is no variational principle of order parameter within the latter .
this demonstrates again the deficiency of gibbs ensemble theory .
in fact , this deficiency was pointed out by born and fucks @xcite early in 1938 . in 1937
, mayer @xcite published his famous paper on gas - liquid transition , which is based directly on gibbs ensemble theory .
this work was seriously doubted and questioned by born and fucks @xcite : how can the gas molecules ` know ' when they have to coagulate to form a liquid and solid ? obviously , owing to lack of landau mechanism , it is hard for gibbs ensemble theory to answer the question posed by born and fucks .
gibbs ensemble theory should be extended to incorporate the landau s ideas so as to describe phase transitions and answer the question posed by born and fucks .
that is another intention of this paper .
in order to describe phase transitions and spontaneous symmetry breaking , we shall , as a personal review , extend gibbs ensemble theory with three hypotheses , which can be stated , taking the bcs superconductivity as an instance , as follows .
the system hamiltonian is represented by @xmath61 , @xmath62 where @xmath63 here @xmath64 is an internal field of the system , it is also the order parameter for bcs superconductivity .
evidently , @xmath61 is not invariant under the gauge transformation of @xmath65 as long as @xmath66 , namely , the gauge symmetry will be broken for @xmath67 if @xmath68 .
from now on , we shall call @xmath69 the phase - transition operator , for sake of convenience .
the statistical average of an observable @xmath70 is defined as @xmath71 we remark that the average is now a function of both temperature and order parameter , which is expressed explicitly by the two arguments of @xmath72 . ( 3 ) .
the order parameter @xmath40 is determined by the minimum of the entropy of the system , @xmath73where @xmath74 denotes the entropy , @xmath75as a function of the order parameter , the entropy controls the evolution of state of the system with temperature .
the first hypothesis means that the system hamiltonian can take different representations at different temperatures , e.g. , it can take the symmetric representation ( @xmath76 ) at a high temperature , and the asymmetric representation ( @xmath68 ) at a low temperature . which representation it will take is determined by the third hypothesis .
after the representation is so determined , the statistical average can be calculated with respect to this representation , as stated in the second hypothesis .
obviously , the extended theory will reduce to the original one if @xmath77 , that is to say , the original theory holds only for the normal phase of the system .
the broken - symmetry phase will be described by the extended theory . within the framework of the extended ensemble theory
, a system will realize different representations of the same system hamiltonian at different temperatures according to the principle of least entropy with respect to order parameter , this mechanism for phase transitions is , in spirit , the same as that given by landau phenomenological theory ; hence , it can be said that landau mechanism has been incorporated into the extended ensemble theory . also , because the order parameter is an internal field of the system itself , symmetry breaking will occur in a completely spontaneous way rather than forced by an external field , there is no unphysical operation within the extended theory . in a word ,
the extended ensemble theory is conceptionally in accordance with landau s ideas on phase transitions : order parameter , variational principle , representation transformation , and spontaneous symmetry breaking . in sec .
[ popt ] , we shall show further that phase transitions originate physically from the wave nature of matter .
now , let us establish the relationship between the extended ensemble theory and thermodynamics , it can be implemented as follows .
first , the internal energy @xmath78 can be obtained from eq .
( [ average ] ) , replacing @xmath70 by the hamiltonian @xmath14 .
it is a function of temperature @xmath79 and volume @xmath33 , that is , @xmath80 from the internal energy @xmath81 , the specific heat at constant volume , @xmath82 , can be calculated through@xmath83 then , we can obtain the thermodynamical entropy @xmath84,@xmath85 it is worth paying attention to the significant difference between the thermodynamical entropy @xmath84 and the statistical entropy @xmath74 defining in eq .
( [ entropy ] ) : @xmath74 measures the degree of order of the system , whereas @xmath84 measures the amount of heat absorbed or rejected by the system , i.e. , @xmath86 ; besides , @xmath74 contains the basic information of the system , it controls the evolution of state of the system , as a consequence , @xmath87 derives itself from @xmath74 . at last , we arrive at@xmath88where @xmath89 is the helmholtz free energy of the system , with @xmath24 and @xmath33 as its natural variables . as a thermodynamic potential of the system
, @xmath89 can generate other thermodynamic quantities .
by such a way , the extended ensemble theory connects with thermodynamics . in the end of this section ,
let us explain why we choose the bcs model rather than the ising model as the starting instance to develop the theory for phase transitions and spontaneous symmetry breaking .
as is well known , ising model @xmath50 is a classical discrete hamiltonian .
simple as it is mathematically , it is , however , ill defined from the point of view of physics .
on one hand , a classical hamiltonian , @xmath90 where @xmath91 and @xmath92 denote the generalized coordinates and momenta of the system , must be continuous in phase space , that is to say , @xmath93 is a continuous function of @xmath91 and @xmath92 ; there does not exist any classical system which corresponds to a discrete hamiltonian ; therefore , @xmath94 does not correspond to a classical system . on the other hand ,
the discrete quantity @xmath95 is not a quantum operator but a pure integer variable , therefore , @xmath50 does not correspond to a quantum system , either . in a word
, ising model can not correspond to any physical system ; its solution , whether rigorous or not , has no real physical relevance @xcite , which was , in fact , realized and pointed out as the `` ising disease '' by people early in 1940 s and 1950 s @xcite . to understand that point further ,
let us first suppose that ising model could represent a real physical system , and then analyze the statistical properties of this system in detail .
first , if the system is not placed in an external magnetic field , then , according to the onsager s solution ising , huang , its specific heat will diverge logarithmically at @xmath51 ; however , as indicated by eq .
( [ mising ] ) , there appears no spontaneous magnetization bellow @xmath6 , namely , the state of the system will remain nonmagnetic at any temperature , whether @xmath96 or @xmath97 .
there is no paramagnetic - ferromagnetic phase transition in the ising system if it is not placed in an external magnetic field , though its specific heat is singular .
that is the first characteristic behavior of the ising system .
secondly , if the system is placed in an external magnetic field , then , according to the yang s solution @xcite , there will appear an induced magnetization bellow @xmath6 .
that is to say , the ising system will transform from a paramagnetic state at @xmath96 into a ferromagnetic state at @xmath97 when it is placed in an external magnetic field , here , the transformation is an induced transformation other than a spontaneous phase transition .
that is the second characteristic behavior of the ising system . in sum ,
an ising system will exhibit two distinct behaviors , respectively , depending on whether the system is placed in an external magnetic field or not .
it will exhibit the first characteristic behavior in the case without the application of any external magnetic field and the second one in the case with the application of an external magnetic field .
the two behaviors should be checked experimentally with respect to the same physical system .
however , such a physical system that can simultaneously show both the two characteristic behaviors of the ising model has never been observed and reported experimentally . in other words , there does not exist the supposed ising system in nature at all !
that is not surprising , it just reflects the ill definition of the ising model . here , it should be pointed out that the solutions of onsager and yang are not identical but essentially different from each other , whether in the sense of mathematics or in the sense of physics : in the limit @xmath98 , the yang s solution can not reduce to the onsager s solution , as can be easily seen by comparing eq .
( mising ) and eq .
( [ mnzero ] ) ; they represent the statistical properties of the ising system with and without the interaction of an external magnetic field , respectively . needless to say ,
an ising system is completely distinguishable from a real physical system of phase transition . for the latter
, the transition happens itself spontaneously , needing no inducement of a corresponding external field .
evidently , this nature of phase transitions is neither the same as the first characteristic behavior of the ising system , i.e. , a singular specific heat but no phase transition , nor the same as the second characteristic behavior of the ising system , i.e. , a transformation induced but not spontaneous .
there are clear and definite discrepancies between the statistical behavior of a real physical system of phase transition and those of the ising system .
any phase transition can not be described by ising model .
of course , if one disregards physically the ill definition of ising model , the distinction between the solutions of onsager and yang , and the discrepancies between the statistical behavior of a real physical system of phase transition and those of the ising system , he can simulate some phase transitions and critical phenomena quite well with ising model kadanoff1,kadanoff2,kadanoff3,wilson1,wilson2,wilson3,wilson4,wilson5,fisher . theoretically , ising model is ill defined ; experimentally , ising system does not exist in nature at all .
that is the reason why we do not choose it as the starting instance to develop the theory of phase transitions and ssb . as for the bcs model @xcite ,
quite the contrary , it is of high value in theory because it contributes a new quantum concept , i.e. , cooper pairs , which is profound and rather hard to understand theoretically but yet verified exactly by the quantized flux experiments .
actually , it is the bcs theory that began our microscopic understanding of phase transitions .
although ising s criterion for phase transitions is problematic in physics , it is much suggestive in mathematics , indeed .
to some degree , it can be said that the present modification is just equivalent to substituting the external field proposed by ising with an internal field determined by the system itself , as can be seen from expanding the right - hand side of eq .
( hprime ) to the linear term of @xmath40 and then comparing it with eq .
( exfield ) .
let us first apply the extended ensemble theory to the ideal fermi gas .
it is the simplest case , and we shall see that this case is exactly solvable . for the ideal fermi gas , there is no interaction ( @xmath99 ) , eq .
( [ bcs ] ) reduces to @xmath100where @xmath101 with @xmath52 and @xmath102 being the mass of the fermions and the chemical potential of the system , respectively . to facilitate the calculation of the entropy , we reformulated eq .
( entropy ) as @xmath103by use of eq .
( [ dfactor ] ) , we find [ ctrans ] @xmath104where @xmath105 and @xmath106 .
substituting them into eq .
( [ sphi ] ) , we have @xmath107 , \label{sphi1}\ ] ] where@xmath108is the fermi distribution function . from eqs .
( [ sphi1 ] ) and ( [ variation ] ) , it follows that @xmath109 which is the equation of order parameter .
obviously , this equation has the solutions,@xmath110 in addition , one can easily deduce from eq .
( [ sphi1 ] ) that@xmath111where @xmath112 & \geq & 0,\text { } & \theta _ { \mathbf{k}}=0,\text { or }
\pi \\ -\beta \varepsilon ( \mathbf{k})\tanh \!\left ( \frac{\beta \varepsilon ( % \mathbf{k})}{2}\right ) \left [ 1-\cos \left ( 2\delta \theta _ { \mathbf{k}% } \right ) \right ] & \leq & 0,\text { } & \theta _ { \mathbf{k}}=\pi /2,\text { or } 3\pi /2,% \end{array}% \right.\ ] ] with @xmath113 being the deviation of @xmath114 from the corresponding solution . for the solution that all @xmath115 or @xmath116 , we have @xmath117 , which meets the requirement of eq .
( [ increment ] ) , thereby , this solution corresponds to a stable phase .
observe that the order parameter @xmath77 when all @xmath115 , one recognizes immediately that this phase is just the normal phase of the system , which can also be seen from the following equation , @xmath118we are familiar with the properties of this normal phase from the standard books on quantum statistical mechanics .
what it is worth stressing here is that the normal phase is proved to be stable at any temperature within the framework of the extended ensemble theory . of course
, that is reasonable , and just what we expect . as to the other solutions
, they also have physical senses though they correspond to instable phases because the entropies for them are not minimal .
let us first discuss the phase with all @xmath119 or @xmath120 .
we find @xmath121 where @xmath122 with @xmath123 .
that is to say , the temperature of the system goes negative .
this shows that negative temperatures root from the same microscopic mechanism as that for phase transitions , so we call this instable phase the negative - temperature phase . in comparison ,
the normal phase is said to be the positive - temperature phase ( @xmath124 ) .
this negative - temperature phase can not be realized physically because its internal energy is unbounded above .
apart from the negative - temperature phase , all the other instable phases are partially negative - temperature phases , which means that the particles of the system are distributed partially among the negative - temperature ( single - particle ) states , and partially among the positive - temperature ( single - particle ) states , @xmath125that is to say , some of the states take @xmath126 or @xmath116 , the others take @xmath127 or @xmath120 . each @xmath128 with @xmath115 or @xmath116 represents a positive - temperature state , and each @xmath128 with @xmath129 or @xmath120 represents a negative - temperature state . because the partially negative - temperature phase is an instable phase , it can emit photons when particles transfer from a negative - temperature state into a positive - temperature one .
therefore , a partially negative - temperature system can constitute a laser . to make that point more specified ,
let us consider a semiconductor that is described by the following hamiltonian , @xmath130where @xmath131 denotes the energy - band index , @xmath132 the energy of the electrons in the @xmath131th band relative to the fermi level of the system , and @xmath133 the annihilation operator of the electrons with momentum @xmath128 and spin @xmath134 in the @xmath131th band . replacing the @xmath135 of eq .
( [ dfactor ] ) with @xmath136and following the same procedure as for the ideal fermi gas , one can easily get @xmath137where @xmath138 .
suppose that there are , for example , two electrons per site .
if the system is at zero temperature and stays in the normal phase where all @xmath139 or @xmath116 , the valence band , i.e. , the @xmath126th band , is fully filled , the conduction band , i.e. the @xmath140st band , and all the other higher bands are empty .
this is the well - known picture for semiconductors ( or insulators ) , which is depicted in fig . 1 .
now , let us consider such a partially negative - temperature phase , @xmath141if this phase is at zero temperature , the conduction band is fully filled , all the other bands are empty .
it implies that the population is inverted between the valence and conduction bands , which is depicted in fig .
as mentioned above , such a population - inversion phase is instable , it will finally turn into the stable phase , i.e. , the normal phase , with the electrons in the conduction band hopping into the valence band and simultaneously releasing energies to the surroundings .
it will make a laser if the energies are released by means of emitting photons .
that is somewhat an ideal semiconductor laser , it is a two - level system , a more realistic laser can be achieved in the so - called three - level system,@xmath142where the @xmath126th and @xmath143nd bands are partially filled , the @xmath140st band is empty , which is depicted in fig .
, the population is not inverted between the @xmath126th and @xmath143nd bands , i.e. , the population of the @xmath143nd band is less than that of the @xmath126th band , but it is inverted between the @xmath140st and @xmath144nd bands .
this kind of three - levelsemiconductor laser has been realized in experiments @xcite .
for this reason , we call the partially negative - temperature phase the laser phase . obviously , a laser phase can not be described by the original gibbs ensemble theory @xcite . traditionally , negative temperatures are something obscure , and hard to understand why they are higher or hotter than positive temperatures purcell , ramsey , pathria . from the viewpoint of the extended ensemble theory ,
the reason is that both the negative - temperature and laser phases are instable , they will trend , by giving up parts of their internal energy to the surroundings , towards the stable phase , viz . , the positive - temperature phase . that is the microscopic interpretation about negative temperatures .
finally , we note that the ideal fermi gas can not produce superconductivity because there does not exist any cooper pair in any case of the solutions of eq .
( [ theta ] ) , @xmath145this result is just as expected , and right in accordance with the bcs theory @xcite . in summary , it is proved , within the framework of the extended ensemble theory , that the ideal fermi gas can not produce superconductivity , its normal phase is stable at any temperature . besides , the extended ensemble theory gives a microscopic interpretation to the negative - temperature and laser phases : they originate from the same microscopic mechanism as that for phase transitions , and most importantly they are instable .
to study bcs superconductivity with the extended ensemble theory , one needs first to solve eq .
( [ variation ] ) to obtain the order parameter , and then to prove that it can satisfy the requirement of eq .
( [ increment ] ) .
unfortunately , that is rather difficult because it is impossible to calculate @xmath146 rigorously when @xmath14 contains an interaction . as a result , we have to seek approximations . before doing so , it is worthwhile to give a brief survey to the self - consistent mean - field theory due to bcs @xcite , which can be summarized as follows , @xmath147 where@xmath148 eq . ( [ bcsmf1 ] ) and eq .
( [ bcsmf3 ] ) constitute a pair of self - consistent equations to determine both the mean - field hamiltonian @xmath149 and the energy gap @xmath150 .
after @xmath151 is so determined , the average of observable @xmath70 can be evaluated according to eq .
( [ bcsmf2 ] ) . obviously , if @xmath152 , the mean - field hamiltonian @xmath153 is not invariant under the gauge transformation of eq .
( [ gauge ] ) , i.e. , the gauge symmetry is broken with respect to @xmath151 .
it should be noted that , as shown in eqs .
( [ bcsmf2 ] ) and ( [ bcsmf3 ] ) , @xmath154 must be used in place of @xmath14 when calculating statistical averages .
otherwise , @xmath150 becomes zero , there is no cooper pair and superconductivity , as has been demonstrated in eq .
( [ zero ] ) . that is the key point of the bcs mean - field theory .
this kind of mean - field theory is also widely used in studying other phase transitions and ssb , bogoliubov theory of a weakly interacting bose gas being a famous example @xcite . nevertheless , as pointed out by emch in sec .
1.1.f of ref .
@xcite , it contains an inescapable paradox .
that is , the spectrum of the system hamiltonian @xmath151 depends on temperature , which contradicts the fact that the spectrum of an operator is an invariant property of this operator itself and should not depend on temperature .
now , for the extended ensemble theory , the normal and superconducting phases correspond , as stated in sec .
[ eet ] , to the symmetric and asymmetric representations of the same system hamiltonian , respectively , it excludes such paradox absolutely .
although the approximation of eqs . ( [ bcsmf1][bcsmf3 ] ) contains the paradox , its results are astonishingly in good agreement with the experiments .
therefore , we need to search a formally identical but essentially different approximation within the extended ensemble theory so as to interpret the superconductivity anew . in language of green s function ( gf ) , the above approximation can be translated into the following formalism , @xmath155 } , \label{gmf1 } \\
\langle \langle c_{\mathbf{k}\uparrow } |c_{-\mathbf{k}\downarrow } \rangle \rangle _ { \omega } ^{\mathrm{mf}}=\frac{\delta } { \omega ^{2}-\left [ \varepsilon ^{2}(\mathbf{k})+\delta ^{2}\right ] } , \label{gmf2 } \\
\delta = -g\sum_{\mathbf{k}}\int_{-\infty } ^{+\infty } \frac{\mathrm{d}\omega } { \pi }
f(\omega ) \mathrm{im}\langle \langle c_{\mathbf{k}\uparrow } |c_{-% \mathbf{k}\downarrow } \rangle \rangle _ { \omega } ^{\mathrm{mf } } , \label{gmf3}\end{gathered}\]]where @xmath156 denotes the retarded green s function defined with respect to @xmath151 .
now , eqs .
( [ gmf2 ] ) and ( [ gmf3 ] ) constitute a pair of self - consistent equations to determine the two complex - valued functions , @xmath157 and @xmath150 , in contrast to eqs .
( [ bcsmf1 ] ) and ( bcsmf3 ) where an operator is involved .
therefore , eqs .
( [ gmf1]gmf3 ) are formally adaptable to transplanting into the extended ensemble theory . as is well known , bcs superconductivity can be explained in terms of these green s functions @xcite .
we shall reestablish them within the framework of the extended ensemble theory .
of course , they must be redefined , with respect to @xmath158 instead of @xmath159 .
now , we devote ourselves to reestablishing a formalism similar to eqs .
( gmf1[gmf3 ] ) , and interpreting the superconductivity with the extended ensemble theory .
following eq .
( [ sphi ] ) , the entropy of a bcs superconductor can be written as@xmath160where [ dopr ] @xmath161and @xmath162 as the usual mean - field approximation , we decouple the last term on the right - hand side of eq .
( [ bcsetr ] ) as follows , @xmath163it results in @xmath164 \notag \\ & -\frac{1}{4}\beta g\sum_{\mathbf{k}^{\prime } } \sin ( 2\theta _ { \mathbf{k}% ^{\prime } } ) e^{i\varphi _ { \mathbf{k}^{\prime } } } ( \overline{c_{\mathbf{k}% ^{\prime } \uparrow } ^{\dag } c_{\mathbf{k}^{\prime } \uparrow } } -\overline{c_{-% \mathbf{k}^{\prime } \downarrow } c_{-\mathbf{k}^{\prime } \downarrow } ^{\dag } } % ) \times \sum_{\mathbf{k}}\sin ( 2\theta _ { \mathbf{k}})e^{-i\varphi _ { \mathbf{k% } } } ( \overline{c_{\mathbf{k}\uparrow } ^{\dag } c_{\mathbf{k}\uparrow } } -% \overline{c_{-\mathbf{k}\downarrow } c_{-\mathbf{k}\downarrow } ^{\dag } } ) % \text{{\large , } } \label{sups}\end{aligned}\ ] ] where eqs .
( [ dopr ] ) , ( [ ctrans ] ) , and ( [ zero ] ) have been used .
substitution of eq .
( [ sups ] ) into eq .
( [ variation ] ) gives the equations of order parameter,@xmath165 they have a trivial solution,@xmath166and a nontrivial solution,@xmath167where@xmath168the trivial solution always remains zero , it does not change with temperature . in contrast , the nontrivial one will depend on temperature , it may be zero for some temperatures , and nonzero for the others . without loss of generality
, we shall take @xmath169 , the nontrivial solution is then simplified as@xmath170they are a pair of coupled equations to determine both @xmath171 and @xmath172 . if @xmath99 , eq .
( [ suplambda ] ) shows that @xmath173 , then eq .
( suptheta ) comes back to the equation ( [ ifgorder ] ) for the ideal fermi gas .
physically , the bcs hamiltonian of eq .
( [ bcs ] ) can describe only the weak - coupling superconductivity . for the weak - coupling superconductivity
, the coupling strength @xmath174 is quite small , from eqs .
( suptheta ) and ( [ suplambda ] ) it follows that @xmath175 and @xmath176 are also small .
this makes it easy for us to do further approximations , as will be seen from below . in order to solve eqs .
( [ suptheta ] ) and ( [ suplambda ] ) , one needs @xmath177 , it can be obtained by the corresponding retarded green s function @xmath178 defined with respect to @xmath158 @xcite . as is well known , @xmath179 satisfies the equation of motion,@xmath180 \rangle \rangle _ { \omega } , \label{gfmotion}\end{aligned}\]]where @xmath181 denotes the anticommutator of @xmath182 and @xmath60 .
expanding @xmath158 into the power series of @xmath40 , @xmath183 + \frac{i^{2}}{2!}% \left [ d(\phi , c),\left [ d(\phi , c),h(c)\right ] \right ] + \cdots , \ ] ] and substituting it into eq .
( [ gfmotion ] ) , we have@xmath184 \rangle \rangle _ { \omega } -i\langle \langle c_{\mathbf{k}\uparrow } |\left [ c_{-% \mathbf{k}\downarrow } , \left [ d(\phi , c),h(c)\right ] \right ] \rangle \rangle _ { \omega } \notag \\ & -\frac{i^{2}}{2!}\langle \langle c_{\mathbf{k}\uparrow } |\left [ c_{-% \mathbf{k}\downarrow } , \left [ d(\phi , c),[d(\phi , c),h(c)]\right ] \right ] \rangle \rangle _ { \omega } + \cdots .\end{aligned}\]]as stated above , @xmath40 is small for the weak - coupling superconductivity , it is thus rational to maintain only the zeroth - order term,@xmath185 with regard to the second gf on the right - hand side , we do the factorization as in eq .
( [ bcsfact ] ) , @xmath186this leads to@xmath187 \langle \langle c_{\mathbf{k}% \uparrow } |c_{-\mathbf{k}\downarrow } \rangle \rangle _ { \omega } = \lambda \langle \langle c_{\mathbf{k}\uparrow } |c_{\mathbf{k}\uparrow } ^{\dag } \rangle \rangle _ { \omega } .
\label{wgfmotion}\]]the gf on the right - hand side can be obtained by the same procedure as for @xmath188,@xmath189 \langle \langle c_{\mathbf{k}% \uparrow } |c_{\mathbf{k}\uparrow } ^{\dag } \rangle \rangle _ { \omega } = 1+\lambda ^{\dagger } \langle \langle c_{\mathbf{k}\uparrow } |c_{-\mathbf{k}% \downarrow } \rangle \rangle _ { \omega } , \]]where @xmath190from eqs .
( [ wgfmotion][dlambda ] ) , it follows that@xmath191 } , \label{wgfmotion1 } \\
\langle \langle c_{\mathbf{k}\uparrow } |c_{-\mathbf{k}\downarrow } \rangle \rangle _ { \omega } = \frac{\lambda } { \omega ^{2}-\left [ \varepsilon ^{2}(% \mathbf{k})+\lambda ^{2}\right ] } , \label{wgfmotion2 } \\
\lambda = -g\sum_{\mathbf{k}}\int_{-\infty } ^{+\infty } \frac{\mathrm{d}\omega } { \pi } f(\omega ) \text{\textrm{im}}\langle \langle c_{\mathbf{k}\uparrow } |c_{-\mathbf{k}\downarrow } \rangle \rangle _ { \omega } .
\label{dlambda1}\end{gathered}\]]they form a self - consistent mean - field solution to the green s functions responsible for bcs superconductivity within the framework of the extended ensemble theory . comparing them with eqs .
( [ gmf1][gmf3 ] ) , one sees that the present solution is identical to the bcs solution except that the statistical average is now defined with respect to @xmath158 .
thus far , the bcs mean - field results have been transplanted into the extended ensemble theory , the paradox removed . with the help of eq .
( [ wgfmotion2 ] ) , eqs .
( [ dlambda1 ] ) and ( suptheta ) can be simplified as@xmath192where@xmath193eq .
( [ ordereq ] ) together with eq .
( [ gapeq ] ) gives the mean - field solution to @xmath40 ( @xmath194 ) , the order parameter for bcs superconductivity .
( [ gapeq ] ) is familiar in the bcs mean - field theory @xcite , it has the nontrivial solution , @xmath195where @xmath6 is determined by the equation , @xmath196with @xmath197 . from eqs .
( [ pseudoorder ] ) and ( ordereq ) , we find @xmath198since @xmath40 is the order parameter of the system , eq .
( [ realorder ] ) indicates that @xmath6 is the transition temperature for bcs superconductivity .
compared with eq .
( [ realorder ] ) , eq .
( [ pseudoorder ] ) shows that the energy gap @xmath172 behaviors with temperature just like an order parameter .
however , this behavior of the energy gap is not a property of itself , but derives from the property of the order parameter , as can be seen from the equation,@xmath199which demonstrates that the forming of cooper pairs is a direct consequence of the gauge - symmetry breaking caused by the internal spontaneous field @xmath200 . in the extended ensemble theory , the order parameter is a physical quantity more fundamental than the energy gap .
this picture is significantly distinct from the bcs mean - field theory where the energy gap is the most fundamental quantity that is responsible for the superconductivity . in sum ,
( [ wgfmotion1 ] ) , ( [ wgfmotion2 ] ) , ( [ gapeq ] ) , and ( ordereq ) constitute a self - consistent mean - field theory for bcs superconductivity within the framework of the extended ensemble theory , its nontrivial solution reproduces the bcs mean - field results on the conventional low-@xmath6 superconductivity .
we are now confronted with the important task to prove the stability of this nontrivial solution . by expanding the @xmath146 of eq .
( [ sups ] ) into a taylor or volterra series in the neighborhood of the nontrivial solution @xmath201 , one finds @xmath202where @xmath203 represents the @xmath131th power term of the expansion of @xmath204 , or the @xmath131th variation of @xmath74 . specifically , the second variation @xmath205 has the form , @xmath206 where @xmath207 and @xmath113 represent the variations of @xmath208 and @xmath209 from the nontrivial solution @xmath201 , respectively . here , the set of @xmath128 has been separated into two subsets : the subset of @xmath210 and the subset of @xmath211 , where @xmath212 and @xmath213 satisfy @xmath214 and @xmath215 , respectively .
all the contributions from the subset of @xmath211 are included by the term within the curly brackets .
let us prove that this term is definitely positive , it can be done by proving that the eigenvalues of the matrix @xmath216,@xmath217are all positive .
the eigenvalue function for @xmath216 is , @xmath218where@xmath219observe that the second term on the right - hand side of eq .
( [ mkk ] ) is a product of the two factors corresponding to the row @xmath211 and the column @xmath220 respectively , we can reduce eq .
( eigenvalueeq ) into @xmath221where @xmath222
^{2}}\delta _ { \mathbf{k_{2}k}% _ { 2}^{^{\prime } } } + 1.\]]obviously,@xmath223is an eigenvalue of eq . ( [ eigenvalueeq1 ] ) because at least two rows ( @xmath224 ) of the determinant @xmath225 are identical .
apart from those eigenvalues , there are possibly other ones , they satisfy the equation , @xmath226 ^{2}}{% 2\xi _ { \mathbf{k}_{2}}\tanh ( \frac{1}{2}\beta \xi _ { \mathbf{k}_{2}})-\lambda } .
\label{eigenvalueeq2}\]]to derive this equation from eq .
( [ eigenvalueeq1 ] ) , it is sufficient to heed that every element of @xmath227 contains the number @xmath140 . if @xmath228 , one has@xmath229 ^{2}}{2\xi _ { % \mathbf{k}_{2}}\tanh ( \frac{1}{2}\beta \xi _ { \mathbf{k}_{2}})-\lambda } < % \frac{1}{2}g\sum\limits_{\mathbf{k}}\frac{\tanh ( \frac{1}{2}\beta \xi _ { % \mathbf{k}})}{\xi _ { \mathbf{k}}}. \label{inequ}\]]eqs .
( [ eigenvalueeq2 ] ) and ( [ inequ ] ) lead us to@xmath230this inequality contradicts the result of eq .
( [ gapeq]),@xmath231that is to say , the root of eq .
( [ eigenvalueeq2 ] ) must be greater than zero if it has any
. combination of this result with eq .
( [ lambdaroot ] ) demonstrates that all the eigenvalues of the matrix @xmath216 are definitely positive , which ends our proof for the positivity of the term within the curly brackets of eq .
( [ deltas2 ] ) .
this positivity implies that@xmath232when @xmath233 , we know from eq .
( [ deltasbcs ] ) that @xmath234 , it meets the requirement of eq .
( [ increment ] ) .
however , if @xmath235 , we must generally examine @xmath236 , @xmath237 , or even more higher variations of @xmath74 to check whether @xmath238 , which is obviously complicated to handle .
fortunately , we need not do that in the case considered now .
when @xmath239 , eq . ( [ deltas2 ] ) shows that@xmath240with them , one can easily verify that @xmath241 .
namely , @xmath242 if @xmath239 . to sum up ,
@xmath234 if @xmath233 , and @xmath242 if @xmath235 .
since @xmath243 , we conclude that the requirement @xmath117 is satisfied , the nontrivial solution is stable . as regards the trivial solution , @xmath77 ( @xmath244 ) , it is identical to the nontrivial one when @xmath245 , and thus is stable at @xmath245 .
if @xmath97 , we find @xmath246 accordingly , it has positive eigenvalues,@xmath247 in addition , we also have @xmath248 ^{2}}{2\varepsilon ( \mathbf{k}% _ { 2})\tanh \left ( \frac{1}{2}\beta \varepsilon ( \mathbf{k}_{2})\right ) -\lambda } .\]]in contrast to eq .
( [ eigenvalueeq2 ] ) , this equation has a negative root of @xmath1 when @xmath97 . to see it clearly ,
let us consider , for instance , the zero temperature case,@xmath249where , as usual , @xmath250 denotes the density of states at the fermi level , and @xmath251 the debye frequency .
apparently , it has the solution,@xmath252the existence of both the positive and negative eigenvalues implies that the trivial solution is a saddle point at @xmath97 . therefore , the term within the curly brackets of eq .
( [ nds2 ] ) is indefinite , that is , it can change in sign . as a consequence ,
the trivial solution , or rather the normal phase , is unstable at @xmath97 .
that is just the cooper instability bcs1,bcs2,bcs3 . of the two solutions , the nontrivial is the only one that is stable within the whole temperature range . with the nontrivial solution obtained ,
other physical quantities can be easily expressed and calculated in terms of the two green s functions @xmath253 and @xmath254 , as was done in refs .
@xcite , rickayzen and @xcite , with the results unchanged and the same as the bcs theory .
so far , bcs superconductivity has been interpreted within the extended ensemble theory . in this interpretation
, the order parameter for bcs superconductivity evolves with temperature according to the principle of least entropy . at high temperatures ( @xmath245 )
, it is zero and stable , the system hamiltonian realizes the representation with perfect gauge symmetry , the resulting phase is normal and disordered . as temperature decreases and goes below the critical temperature ( @xmath97 ) ,
the zero solution becomes unstable , instead , there arises a new stable solution of the order parameter , it is nonzero . at the same time , the system hamiltonian transforms into a new representation with gauge symmetry broken , the electrons are hence formed into cooper pairs , and the resulting phase gets superconducting . thus and so , the extended ensemble theory answers why , when and how the electrons can coagulate to form cooper pairs and superconductivity , the problem posed by born and fucks @xcite .
lastly , it should be noted that the bcs and landau theories are unified into a single formalism within the framework of the extended ensemble theory .
after the discussion of bcs superconductivity , we proceed now to study bose - einstein condensation ( bec ) .
we shall concentrate on three systems : the ideal bose gas , the photon gas in a black body , and the ideal phonon gas in a solid body .
they are the most important cases of bose systems : the first is a system with the conservation of particles , but the other two are not ; the second belongs to gauge fields whereas the third does not .
in addition , we shall also discuss the quantization of dirac field from a standpoint of the extended ensemble theory .
let us begin with the ideal bose gas .
its hamiltonian reads as follows , @xmath255where @xmath256 ( @xmath257 ) is the annihilation ( creation ) operator for the bosons with a momentum @xmath258 , and @xmath259 the single - particle energy , and @xmath102 the chemical potential .
this hamiltonian is gauge invariant,@xmath260where @xmath261the invariance has an important consequence , @xmath262it signifies that the condensation amplitude will be zero forever if the gauge symmetry does not break down . in order to examine whether the gauge symmetry can break down spontaneously or not , one needs , following the hypotheses in sec .
[ eet ] , to consider the phase - transition operator , @xmath263 where the internal field @xmath264 is the order parameter for bec .
the reason for introducing eq .
( [ becpto ] ) can be seen from comparing eqs .
( becgauge ) and ( [ becpto ] ) with eqs .
( [ zero ] ) and ( [ dfactor ] ) respectively .
we shall consider first the finite volume case where @xmath33 and @xmath34 are both finite , and then the thermodynamic limit case where @xmath265 , and @xmath266 , but @xmath35 is finite .
as will be seen , the two cases are significantly different . in this case
, the wave vector @xmath128 is discrete .
the entropy of the system can be written as @xmath267 where @xmath268 is independent of the order parameter @xmath264 . according to eqs .
( [ variation ] ) and ( [ increment ] ) , we have @xmath269where @xmath270 represents the variation of @xmath271 from the solution given by eq .
( [ orderparam0 ] ) .
( becdelta ) shows that the chemical potential @xmath102 must satisfy the following condition,@xmath272as @xmath273 , this condition is equivalent to @xmath274this inequality is the physical condition for the stability of the system . in combination with eq .
( [ becs ] ) , eq .
( [ orderparam0 ] ) gives us the equation of order parameter , @xmath275it has two solutions : the trivial one,@xmath276and the nontrivial one,@xmath277 for the nontrivial solution , @xmath278 , it causes a contradiction,@xmath279where the left - hand side can not be negative , but the right - hand side is negative because the term of @xmath280 is infinite .
therefore , the nontrivial solution must be discarded . with regard to the trivial solution
, it corresponds to the normal phase of the system .
it does not cause any contradiction,@xmath281this equation can be satisfied at any temperature .
in fact , it is just the equation for the chemical potential @xmath102 , its solution will give @xmath282 . because @xmath34 is finite , @xmath102 will be less than zero irrespective of temperature , i.e.,@xmath283it implies that the normal phase of the system is always stable,@xmath284this proves that there is no phase transition in the finite volume case , the system will stay in its normal phase forever . in this case ,
@xmath29 , we should use the density of entropy instead of the entropy itself , @xmath285where @xmath286is a term irrelevant to @xmath287 .
the limit @xmath29 can be achieved as follows . as usual
, suppose that the system is box normalized , i.e. , @xmath288where @xmath289 denotes the set of integers , and @xmath290 , @xmath291 , and @xmath292 the dimensions of the box along the @xmath293 , @xmath294 and @xmath295 directions respectively ( @xmath296 ) .
as is well known , each @xmath128 corresponds to an eigenfunction , all the eigenfunctions constitute a hilbert space . here
, the hilbert space is a complete space of integrable functions , with the inner product defined as a lebesgue integral .
it should be pointed out that the inner product must be defined as a lebesgue rather than riemann integral because the former can ensure the completeness of an integrable function space whereas the latter can not @xcite .
this implies that all the problems relevant to a hilbert space of integrable functions must be treated with lebesgue theory . on the other hand , lebesgue measure and integration
is itself the mathematical foundation of probability theory kolmogorov , feller and statistical physics @xcite .
we shall therefore perform the thermodynamic limit according to the lebesgue theory of integration .
formally , lebesgue integration includes both the sum over discrete numbers and the integral on a continuous region .
when @xmath33 is finite , a sum over @xmath297 is a lebesgue integral of discrete form .
it will transform into a lebesgue integral of continuous form as @xmath29 .
to show the transformation , let us consider , e.g. , the sum over @xmath128 in eq .
( [ sdensity ] ) . in the first place ,
we write it into an explicit form of lebesgue integral , @xmath298 where@xmath299as usual , @xmath300 represents the characteristic function of the measurable set @xmath301 , and @xmath302 the lebesgue measure of the set @xmath301 . from eqs .
( aset[characterf ] ) , it follows that @xmath303\eta ^{\dag } ( \mathbf{p})\eta ( \mathbf{p% } ) .
\label{vlimit}\end{aligned}\]]now , in the second place , let us take the thermodynamic limit on both the sides of eq .
( [ li ] ) , @xmath304\eta ^{\dag } ( \mathbf{k})\eta ( \mathbf{k})% \mathrm{d}\mathbf{k}.\ ] ] in such a natural way , a discrete sum over @xmath128 will transform into a continuous lebesgue integral in the thermodynamic limit . with the help of the above equation ,
one arrives at@xmath305 \eta ( \mathbf{k})=0 , \label{beccorder } \\
\delta s=\beta \frac{1}{\left ( 2\pi \right ) ^{3}}\int\limits_{% % tcimacro{\u{211d } } % % beginexpansion \mathbb{r } % endexpansion ^{3}}[\epsilon ( \mathbf{k})-\mu ] \delta \eta ^{\dag } ( \mathbf{k})\delta \eta ( \mathbf{k})\mathrm{d}\mathbf{k}\geq 0 .
\label{beccdelta}\end{gathered}\]]the second equation means that the chemical potential must be less than or equal to zero,@xmath306it is the condition for the stability of the system .
the first equation is equivalent to @xmath307\eta ^{\dag } ( \mathbf{k})\eta ( \mathbf{k}% ) = 0 , \label{eta}\]]which can be easily verified with respect to the two cases : @xmath308 and @xmath309 .
( [ eta ] ) has two solutions : the trivial one,@xmath310and the nontrivial one,@xmath311where @xmath312 is a complex which does not depend on @xmath128 , and @xmath313 the dirac @xmath314 function . for the trivial solution ,
the system is in its normal phase , this phase can exist only at high temperatures , which can be easily deduced from the equation of chemical potential,@xmath315where we have reduced the lebesgue integral into an improper riemann integral in the end . because , as shown by eq .
( [ becmu ] ) , @xmath102 must be less than or equal to zero , the above equation can not hold if @xmath97 where @xmath6 is determined by @xmath316as is well known , this gives@xmath317where @xmath318 denotes the riemann zeta function . for the nontrivial solution
, @xmath102 must be zero , the order parameter is determined by@xmath319where we have used the fact that the contribution from the state of @xmath320 is zero when reducing the lebesgue integral into the improper riemann integral in the end . because @xmath321 , the above equation can not hold if @xmath96 . from those discussions
, it follows that@xmath322that is to say , a phase transition will happen at @xmath58 . because @xmath323 when @xmath324 , the normal phase ( @xmath325 ) exists and is stable at @xmath245 , the ordered phase ( @xmath326 ) exists and is stable at @xmath97 . here
, we note that it is a natural consequence of the extended ensemble theory that the chemical potential of the ideal bose gas is fixed at zero ( @xmath278 ) in the ordered phase ( @xmath97 ) , in contrast to the einstein s treatment where it is fixed by physical insight einstein1,einstein2 . obviously , the above results are the same as the bec given by einstein einstein1,einstein2 if @xmath327 were regarded as the density of the bosons condensed onto the state of @xmath280 .
therefore , the transition happening at @xmath58 must be identified physically as the bose - einstein condensation .
it should be pointed out that , generally , the quantity @xmath327 can not be interpreted as the density of condensed bosons , as will be clarified in sec . [ wibg ] .
so far , we have presented a new description to the bose - einstein condensation of the ideal bose gas . in comparison with the einstein s description @xcite , we now describe the bec with two new conceptions : spontaneous symmetry breaking and lebesgue integration . within this new description
, the bec happens simultaneously with the spontaneous breaking of the gauge symmetry ; the integration is defined and performed in lebesgue way .
that is fundamentally different from einstein s description @xcite where the bec happens with no symmetry breaking , and the integration is done in riemann way . here
, it is worth emphasizing the important role played by lebesgue integration in the study of the bec of the ideal bose gas .
as has been seen , there arises the bose distribution function , @xmath328which is unbounded above on the set @xmath329 .
because the integrand of a riemann integral must be bounded everywhere , such a function is not riemann integrable and can not be handled with riemann integration . fortunately , that makes no trouble .
as pointed above , it is , in fact , unnecessary for us to handle the bose distribution function with riemann integration ; what really confronts us in statistical physics is lebesgue other than riemann integration . for lebesgue integration
, it permits more general functions as integrands , and treats bounded and unbounded functions on an equal footing . in the sense of lebesgue integration ,
the bose distribution function is integrable , which can be shown as follows , @xmath330 where the set @xmath331 includes one point only .
@xmath332 is a null set , i.e. , its measure is zero , thus , the integral on it vanishes .
the rest part , i.e. , the integral on @xmath333 , reduces finally into an improper riemann integral .
as is well known , the integral on a null set always vanishes irrespective of the integrand values on this set , even if they are infinite .
thereby , the integral on any single state equals zero , the ground state ( @xmath280 ) is not more advantageous than the other states ( @xmath334 ) . in other words , the singularity of @xmath335 on the point @xmath280 has no special meaning , both in mathematics and in physics .
this aspect makes the present treatment quite distinct from that of einstein , where the integration is handled in riemann way , with a special treatment given to the ground state of @xmath280 .
intuitively , einstein thought that the bosons would gather in the ground state if @xmath97 , and thus the contribution from this state is nonzero @xcite , that is , @xmath336where @xmath337 is the particle density gathering in the ground state . comparing eqs .
( [ lesb3 ] ) , ( [ bose1 ] ) and ( [ bose2 ] ) shows more clearly the difference between the present treatment and einstein s . from this difference
it follows that the quantity @xmath338 needs not to be interpreted as the density of the bosons condensed onto the state of @xmath280 .
in fact , eq . ( [ lesb3 ] )
is physically just an equation of order parameter .
here , it should be noted that the bogoliubov theory of weakly imperfect bose gas @xcite is based on eq .
( [ bose2 ] ) rather than eq .
( [ bose1 ] ) . at last ,
why did einstein not use lebesgue integration ?
there is a historical reason : until 1925 when einstein @xcite studied the bec , lebesgue integration had not yet been introduced to probability theory by kolmogorov , who did it in 1933 @xcite . also , it is significant to compare the two lebesgue integrals represented in eq .
( [ lesb0 ] ) and eq .
( [ lesb3 ] ) : both integrands are unbounded on @xmath339 ; however , the former is infinite on a set with a nonzero measure ( @xmath340 ) , the latter is finite almost everywhere , or infinite just on a null set ( @xmath341 ) .
this difference leads to the result that the nontrivial solution must be discarded in the finite volume case whereas it has physical sense in the thermodynamic limit . in a word
, the ideal bose gas can produce bec in the thermodynamic limit , but it can not in the finite volume case .
that is a rigorous conclusion deduced from the extended ensemble theory . as another instance
, the ideal bose gas shows unambiguously that gibbs ensemble theory needs some kind of extension .
otherwise , the conservation of particles will be broken when @xmath97 , as can be seen by comparing eqs .
( [ bose1 ] ) and ( [ bose2 ] ) .
as the only one of exactly solvable quantum phase transition , the bec in the ideal bose gas demonstrates that the extended ensemble theory postulated in sec .
[ eet ] is feasible in physics for us to describe phase transitions .
lastly , we would like to stress that the condition of eq .
( [ becmu ] ) , i.e. , @xmath324 , is not a mathematical limit but just a physical requirement . like the case of @xmath324 ,
the extended ensemble theory is also well defined mathematically for @xmath342 , as is shown in appendix fsum .
removing the limitation on the chemical potential is physically reasonable , it enables the conservation of particles to be handled freely , and makes it easy to describe the bec in the interacting bose gas , as can be seen in sec .
[ wibg ] .
* remark 1 * : the ideal bose gas is a touchstone for the theory of phase transitions because it is the only one of the many - body systems that can show a quantum phase transition and can be solved rigorously .
any theory of phase transitions must be examined by the ideal bose gas .
it can not be accepted in physics if it can not give a rigorous solution to the ideal bose gas ; and it can be further applied to the interacting bose gas only after it has given a rigorous solution to the ideal bose gas . *
remark 2 * : in 1926 , uhlenbeck @xcite raised an objection to einstein s viewpoint on the condensation , he argued that the ideal bose gas can not produce bec if the volume of the system remains finite .
this objection caused an exciting debate during the van der waals century conference in november 1937 @xcite .
evidently , the present theory has given the exact answer to the debate : uhlenbeck and einstein are both right , which is also in accordance with the agreement reached finally in the conference .
* remark 3 * : in the momentum space used here , the condensed phase is shown to be homogeneous . in sec .
[ wibg ] , using the real space , we shall prove further that the condensed phase can only be homogeneous .
there can not exist any supercurrent or quantized vortex in the ideal bose gas .
* remark 4 * : the rigorous solution for the ideal bose gas shows that the extended ensemble theory holds in the critical region . for the photon gas in a black body , its hamiltonian reads , @xmath343where coulomb gauge is used , @xmath344 ( @xmath345 ) is the destruction ( construction ) operator for the photons with momentum @xmath128 and polarization @xmath346 , and @xmath347 represents the photon energy with @xmath348 the speed of light .
analogous to the ideal bose gas , this system has the phase - transition operator , @xmath349 where @xmath350 is the order parameter for bec . from @xmath351 and @xmath352
, it follows that@xmath353the first is the equation of motion of order parameter , it has the solution,@xmath354where @xmath355 is a complex constant which does not depends on @xmath24 .
owing to the fact : @xmath356 , this solution is stable,@xmath357to determine @xmath355 , it is sufficient to consider the initial state at @xmath358 ( @xmath359 )
. physically , any system must stay in its disordered phase at @xmath358 .
therefore , @xmath355 must be zero at @xmath358 .
since @xmath355 does not depends on @xmath360 , it is always zero .
that is to say , @xmath361this equation shows that the photon gas in a black body can not produce bec , and must always stay in its normal phase .
( [ eds ] ) demonstrates that the normal phase is stable .
this result is also valid for the thermodynamic limit case .
in short , there can not arise bec in the photon gas in a black body , the system stays in its normal phase forever , planck s radiation law holds at any temperature .
obviously , this conclusion agrees completely with the experiment .
as the third system , let us consider the ideal phonon gas in a solid body , whose hamiltonian reads , @xmath362where @xmath363 ( @xmath364 ) is the destruction ( construction ) operator for the phonons with momentum @xmath365 and polarization @xmath134 , @xmath366 represents the phonon energy with @xmath367 the velocity of sound . by comparing eq .
( [ hphonon ] ) with eq .
( [ hphoton ] ) , one concludes at once that the ideal phonon gas in a solid body can not produce bec either , and must also stay in its normal phase forever . at last ,
let us consider the quantization of dirac field from the standpoint of the extended ensemble theory hypothesized in sec .
[ eet ] . if one quantizes dirac field using bose - einstein statistics , he has the following hamiltonian @xcite,@xmath368 where @xmath369 and @xmath370 denote momentum and spin , respectively ; @xmath371 ( @xmath372 ) and @xmath373 ( @xmath374 ) are bosonic destruction ( construction ) operators ; and @xmath375 with @xmath52 and @xmath348 being the electron mass and the speed of light , respectively . evidently , this system has the phase - transition operator,@xmath376where @xmath287 and @xmath312 are the order parameters for bec .
( [ variation ] ) and ( [ increment ] ) imply that@xmath377eqs .
( [ dirac1 ] ) and ( [ dirac2 ] ) show that the system should always stay in its normal phase , but eq . ( [ dirac3 ] ) demonstrates that this normal phase corresponds to a saddle point , and is thus instable . therefore ,
if dirac field were quantized using bose - einstein statistics , the system would be instable and collapse .
in other words , it is not admissible to quantize dirac field using bose - einstein statistics .
this conclusion is in accordance with the spin - statistics theorem in quantum field theory greiner . on the other hand ,
if dirac field is quantized using fermi - dirac statistics , its hamiltonian will read as follows @xcite,@xmath378now @xmath379 ( @xmath380 ) and @xmath381 ( @xmath382 ) are fermionic destruction ( construction ) operators for electrons and positrons . according to sec .
[ eet ] , its phase - transition operator is@xmath383,\end{aligned}\]]where @xmath40 and @xmath384 are order parameters . following the arguments of sec .
[ ifg ] , one recognizes immediately that the system is always stable in its normal phase .
this holds even in the original representation of the hamiltonian @xcite , @xmath385where @xmath386@xmath387.\end{aligned}\]]in contrast to the electron - positron representation of eq .
( [ eprp ] ) , a minus sign appears in the original representation of eq . ( [ ehpr ] ) , but it can not influence the stability of the system . anyway , there is no problem when dirac field is quantized using fermi - dirac statistics . in this section ,
the systems concerned are free boson fields , they can all be solved rigorously .
we shall proceed to the interacting systems in sec .
generally , they can not be solved rigorously , approximations are needed .
before proceeding forward to the interacting bose systems , it is worth making a research into the physical origination for phase transitions .
again , let us take bcs superconductivity as an instance . according to eq .
( [ average ] ) , the statistical average of a one - body operator can be reformulated as @xmath388where @xmath389 is the electron field , @xmath390 stands for the one - body operator in schrdinger picture , and @xmath391after the phase transition ( @xmath392 ) , the electron field @xmath393 is separated into two fields , @xmath394where @xmath395substituting eq .
( [ psiprime ] ) into eq .
( [ single ] ) , we obtain @xmath396f(\mathbf{r})[\psi ( % \mathbf{r})+\varphi ( \mathbf{r})]}\mathbf{. } \label{pattern1}\ ] ] the right - hand side shows that there exists interference , as usual , it is described by the cross terms , @xmath397 in terminology of quantum optics , it represents the single - particle interference .
this single - particle interference will vanish if the system goes into the normal phase where @xmath398 .
a special case is the interference appearing in the particle density where @xmath399 , @xmath400this form of interference is very familiar in optics , eq .
( [ crossterms ] ) being the general case .
in particular , an abnormal form of single - particle interference can appear in the electron - pair amplitude , @xmath401[\psi _
{ \uparrow } ( \mathbf{r})+\varphi _ { \uparrow } ( \mathbf{r})]},\]]it is described by the abnormal cross terms@xmath402which have no counterparts in optics , in contrast to eq . ( [ crossterms1 ] ) .
observe eq .
( [ ctrans ] ) , one recognizes that the field @xmath403 contains two parts,@xmath404where @xmath405 represents the creation part , @xmath406 , \]]with @xmath407 and @xmath408 being connected with the operators @xmath409 and @xmath410 , respectively ; and @xmath411 the annihilation part , which is connected with the operators @xmath10 and @xmath412 .
this separation of two parts leads us to @xmath413 clearly , there exists interference on the right - hand side , this interference induce the abnormal interference on the left - hand side .
the abnormal interference has an important consequence , that is , the electrons are formed into cooper pairs in the superconducting phase . evidently , the abnormal interference vanishes in the normal phase .
also , there exists the two - particle interference in the superconducting phase , which arises from the statistical average of a two - body operator,@xmath414 where @xmath415 stands for the two - body operator in schrdinger picture . in the sense of eqs .
( [ pattern1 ] ) and ( [ pattern2 ] ) , the superconducting phase can be identified as a form of interference pattern .
physically , the onset of an interference pattern will make a system more structured , and thus increases its degree of order and decreases its entropy .
that is why the superconducting phase has a higher degree of order and a less amount of entropy than the normal phase . from the above discussions
, it follows that there will appear interference in the ordered phase whereas there will not in the normal ( disordered ) phase .
obviously , the same conclusion holds for other phase transitions .
therefore , a phase transition amounts to the onset of a form of interference .
as well - known , interference is a characteristic property of wave , we thus conclude that phase transitions originate physically from the wave nature of matter . physically , any system will go in its disordered phase at sufficiently high temperatures , or as @xmath416 . because order parameter equals zero in disordered phase ,
the extended ensemble theory should ensure that there exists a zero solution of order parameter , at least at high temperatures .
so it does , indeed . upon a transformation , eq .
( [ entropy ] ) can be expressed as @xmath417 considering the expansion,@xmath418+\frac{\left ( -i\right ) ^{2}}{2}[d(\phi , c),[d(\phi , c),h\left ( c\right ) ] ] + \cdots , \]]we obtain@xmath419\rho ( h(c))\big)+\cdots , \]]the linear term of @xmath40 vanishes . apart from an irrelevant term , this equation shows that the powers of @xmath40 are equal to or higher than the @xmath143nd in the expansion of @xmath146 . as a result of @xmath420 ,
the powers of @xmath40 are equal to or higher than the @xmath140st in the equation of order parameter .
it implies that there always exists a zero solution of @xmath40 at any temperature , i.e. , the trivial solution .
in fact , this solution is a stable solution as @xmath421,@xmath422 \big)% = 0,\text { } t\rightarrow + \infty .\ ] ] which satisfies the criterion of eq .
( [ increment ] ) . of course
, this conclusion is also valid for other phase transitions . to summarize , there exists a zero solution of order parameter at any temperature in the extended ensemble theory , this solution becomes stable as @xmath359 .
therefore , a system always stays in its normal phase at sufficiently high temperatures , this high - temperature phase is disordered and structureless . as temperature decreases , there can arise spontaneously the quantum interference of matter waves , which makes the system structured and transform into its ordered phase . that is the physical picture for phase transitions within the framework of the extended ensemble theory . * remark * : in ref .
@xcite , yang suggested an important concept , i.e. , off - diagonal long - range order ( odlro ) , to describe quantum phase transitions .
physically , odlro is a generalization of the concept of copper pairing . with this generalized concept
, he developed many beautiful theorems , which are characteristics of quantum phases .
regrettably , in that paper , there is no discussion of the properties of the hamiltonian that is needed to ensure the existence of odlro at low temperatures .
evidently , there exists no odlro if a system hamiltonian is in its symmetric representation
. the requirement can be satisfied if the system hamiltonian takes its asymmetric , or broken - symmetry , representation .
therefore , those theorems in ref .
@xcite still hold in the extended ensemble theory . as a matter of fact , the odlro s of the reduced density matrices @xmath423 and @xmath424 @xcite , e.g. , @xmath425 and @xmath426 , result physically from the single - particle and two - particle interferences , respectively .
that is to say , off - diagonal long - range order is a physical manifestation of the quantum interference of matter waves .
in this section , we focus our attention on the interacting bose systems . we shall first study the weakly interacting bose gas , and then the double - well potential systems where structural phase transitions , goldstone bosons , and higgs mechanism are concerned . as usual
, the hamiltonian for an interacting bose gas reads , @xmath427=\int \mathrm{d}\mathbf{r}\left [ \frac{\hbar ^{2}}{2m}\nabla \psi ^{\dagger } ( \mathbf{r})\cdot \nabla \psi ( \mathbf{r})-\mu \psi ^{\dagger } ( % \mathbf{r})\psi ( \mathbf{r})\right ] + \frac{1}{2}\iint \mathrm{d}\mathbf{r}% \mathrm{d}\mathbf{r}^{\prime } \psi ^{\dagger } ( \mathbf{r})\psi ^{\dagger } ( % \mathbf{r}^{\prime } ) v(\mathbf{r}-\mathbf{r}^{\prime } ) \psi ( \mathbf{r}% ^{\prime } ) \psi ( \mathbf{r}),\ ] ] where @xmath389 represents the boson - field operator , and @xmath428 the interaction .
expressed in terms of @xmath429 and @xmath430 , the phase - transition operator for bec becomes , @xmath431 = \int \mathrm{d}\mathbf{r}\left [ \eta ^{\dagger } ( % \mathbf{r})\psi ( \mathbf{r})+\eta ( \mathbf{r})\psi ^{\dagger } ( \mathbf{r})% \right ] .\ ] ] with @xmath432 $ ] and @xmath433 $ ] , the entropy of the system can be obtained , @xmath434 & = -\mathrm{tr}\big(\ln \left ( \rho ( h\left [ \psi \right ] ) \right ) \!\rho ( h\left [ \psi \right ] ) \big ) \notag \\ & + \beta \bigg\{\int \mathrm{d}\mathbf{r}\left [ \frac{\hbar ^{2}}{2m}\nabla \eta ^{\dagger } ( \mathbf{r})\cdot \nabla \eta ( \mathbf{r})-\mu \eta ^{\dagger } ( \mathbf{r})\eta ( \mathbf{r})\right ] + \iint \mathrm{d}\mathbf{r}% \mathrm{d}\mathbf{r}^{\prime } \eta ^{\dagger } ( \mathbf{r})\eta ( \mathbf{r})v(% \mathbf{r}-\mathbf{r}^{\prime } ) \overline{\psi ^{\dagger } ( \mathbf{r}% ^{\prime } ) \psi ( \mathbf{r}^{\prime } ) } \notag \\ & + \iint \mathrm{d}\mathbf{r}\mathrm{d}\mathbf{r}^{\prime } \eta ^{\dagger } ( % \mathbf{r})\eta ( \mathbf{r}^{\prime } ) v(\mathbf{r}-\mathbf{r}^{\prime } ) % \overline{\psi ^{\dagger } ( \mathbf{r}^{\prime } ) \psi ( \mathbf{r})}+\frac{1}{2% } \iint \mathrm{d}\mathbf{r}\mathrm{d}\mathbf{r}^{\prime } \eta ^{\dagger } ( % \mathbf{r})\eta ( \mathbf{r})v(\mathbf{r}-\mathbf{r}^{\prime } ) \eta ^{\dagger } ( \mathbf{r}^{\prime } ) \eta ( \mathbf{r}^{\prime } ) \bigg\ } , \label{ibecs}\end{aligned}\]]where @xmath435 denotes the statistical average of the operator @xmath89 with respect to @xmath432 $ ] .
its variation yields the equation of order parameter,@xmath436that is a generalized ginzburg - landau or gross - pitaevskii equation . in the simple case where
@xmath437 , it reduces to@xmath438 \eta ( \mathbf{r}% ) + g\left\vert \eta ( \mathbf{r})\right\vert ^{2}\eta ( \mathbf{r})=0,\ ] ] which is the standard ginzburg - landau @xcite or gross - pitaevskii gp1,gp2 equation .
in addition to eq .
( [ iop ] ) , one also needs@xmath439which is the equation of chemical potential .
( [ iop ] ) and eq .
( icp ) constitute the two basic equations for an interacting bose system . for simplicity ,
let us first consider the homogeneous solution,@xmath440where @xmath312 is independent of @xmath441 .
accordingly , eqs .
( [ ibecs ] ) and ( [ iop ] ) get simplified as@xmath434 & = -\mathrm{tr}\big(\ln \left ( \rho ( h\left [ \psi \right ] ) \right ) \!\rho ( h\left [ \psi \right ] ) \big ) \notag \\ & + \beta \int \mathrm{d}\mathbf{r}[-\widetilde{\mu } \xi ^{\dagger } \xi + % \frac{1}{2}v(0)(\xi ^{\dagger } \xi ) ^{2 } ] , \label{ibecs1}\end{aligned}\]]@xmath442where@xmath443 \notag \\ & \times \left ( -\frac{1}{\beta \hbar } \right ) \sum_{n}e^{i\omega _ { n}\eta } % \mathcal{g}(\mathbf{k},i\omega _ { n } ) , \label{rmuw } \\ v(\mathbf{k } ) & = \int \mathrm{d}\mathbf{r}v(\mathbf{r})e^{-i\mathbf{k}\cdot \mathbf{r}},\end{aligned}\]]@xmath444 being the single - particle green s function defined with respect to @xmath432 $ ] .
( [ ibecs1 ] ) has the typical form of landau free energy @xcite . from eqs .
( [ ibecs1 ] ) , ( [ iop1 ] ) and ( [ increment ] ) , it follows that the interaction between two particles must be repulsive for an interacting bose system to be stable , i.e. , @xmath445 .
otherwise , @xmath446 , the system will collapse .
suppose that the repulsive interaction is weak , we can then treat it as a perturbation . as a technique of perturbation , we shall employ gf , and adopt the self - consistent hartree - fock approximation as in ref .
@xcite,@xmath447 } , \label{ibechf1 } \\
\hspace{-1cm}\hbar\sigma^{\star}\left ( \mathbf{k}\right ) & = \int \frac{% \mathrm{d}\mathbf{k}^{\prime}}{\left ( 2\pi\right ) ^{3}}\left [ v(0)+v(\mathbf{% k}-\mathbf{k}^{\prime})\right ] \notag \\ & \times\left ( -\frac{1}{\beta\hbar}\right ) \sum_{n}e^{i\omega_{n}\eta } % \mathcal{g}(\mathbf{k}^{\prime},i\omega_{n } ) , \label{ibechf2}\end{aligned}\ ] ] where @xmath448 represents the proper self - energy . by use of eq .
( [ rmuw ] ) , @xmath449 can be expressed as@xmath450 } , \label{ibecgf}\ ] ] where@xmath451 \notag \\ & \times\left ( -\frac{1}{\beta\hbar}\right ) \sum_{n}e^{i\omega_{n}\eta } % \mathcal{g}(\mathbf{k}^{\prime},i\omega_{n } ) .
\label{renergy}\end{aligned}\ ] ] eq .
( [ ibecgf ] ) shows that @xmath452 and @xmath453 are the renormalized energy and chemical potential , respectively . with the help of eqs .
( [ homoksi ] ) and ( [ ibecgf ] ) , eq .
( [ icp ] ) can be reduced as@xmath454 where @xmath455 is the particle density of the system . when @xmath456 , eq .
( [ iop1 ] ) has the only solution @xmath457 . as shown by eq .
( [ ibecs1 ] ) , it is a stable solution . in this case , the system is in its normal phase .
the renormalized chemical potential @xmath458 is given by eq .
( [ icp1 ] ) , @xmath459 where @xmath460 denotes the principal value introduced in appendix fsum , and @xmath461 the density of states , @xmath462 if @xmath463 , eqs .
( [ ibecs1 ] ) and ( [ iop1 ] ) show that the system will transform into a condensation phase where @xmath464 substitution of it into eq .
( [ icp1 ] ) yields @xmath465 which gives the renormalized chemical potential @xmath453 at @xmath466 , where @xmath6 is the critical temperature for bec,@xmath467 that is to say , @xmath468 . in sum ,
( [ cpgtc ] ) and ( [ cpltc ] ) yield together the renormalized chemical potential of the system at @xmath96 and @xmath97 , respectively . in order to deduce @xmath469 and @xmath312 in more detail ,
let us suppose further that @xmath415 is of short range .
following ref .
@xcite , we can approximate @xmath470 as@xmath471 , \label{ibecsr}\]]where@xmath472 inserting eq .
( [ ibecsr ] ) into eq .
( [ renergy ] ) and eq .
( [ rmuw ] ) leads us to@xmath473
\notag \\ & \times \mathcal{p}\int_{-\infty } ^{+\infty } \mathrm{d}\omega \mathcal{n}% ( \omega , \widetilde{\mu } ) \frac{\left ( \omega + \widetilde{\mu } \right ) } { % e^{\beta \omega } -1 } , \label{iapprox2}\end{aligned}\]]where @xmath474 is the renormalized mass,@xmath475 , \label{iapprox3}\]]and @xmath476 the density of states defined by eq .
( [ ibecdos ] ) , @xmath477 from eqs .
( [ itc ] ) , ( [ iapprox3 ] ) and ( [ iapprox4 ] ) , it follows that@xmath478 where @xmath479 stands for the critical temperature of the bec happening in the corresponding ideal bose gas , @xmath480 eq .
( [ lowtc ] ) manifests that a repulsive interaction will lower the transition temperature of bec,@xmath481 also , we have by eq .
( [ cpgtc]),@xmath482where @xmath483 and @xmath484 .
this equation gives the @xmath469 at @xmath96 . if @xmath97 , eq . ( [ cpltc ] ) must be used in place of eq .
( [ cpgtc]),@xmath485where @xmath486 , and @xmath487 .
this equation will yield both the @xmath469 and @xmath488 at @xmath97 .
the numerical results of @xmath469 and @xmath489 are shown in figs . 4 and 5 , respectively .
we observe that the renormalized chemical potential @xmath469 is a monotonically decreasing function of temperature , at @xmath58 itself continuous but its derivative not .
this agrees with the @xmath245 result given by ref .
fetter and the references therein .
when @xmath97 , the chemical potential @xmath490 is set zero in ref .
@xcite , as einstein einstein1,einstein2 did in the ideal bose gas . in the extended ensemble theory
, there is no mathematical limit on @xmath469 , it will change with temperature even if @xmath97 so as to ensure the conservation of particles , as pointed out in appendix [ fsum ] . the behavior of @xmath491 shown in fig .
5 is just expected for an order parameter .
the internal energy of the system , @xmath329 , can be derived conveniently from @xmath492 , @xmath493where@xmath494combination of eqs .
( [ mugtc ] ) and ( [ egtc ] ) gives @xmath495 , the specific heat at constant volume , @xmath496where@xmath497this result is the same in form as that for the ideal bose gas , which is familiar in standard books on quantum statistical mechanics .
if @xmath498 , @xmath495 must be calculated from eqs .
( [ multc ] ) and ( [ eltc ] ) .
the numerical results of @xmath495 are shown in fig .
the shape of @xmath495 shows that the bec occurring in the interacting bose gas is a @xmath1-transition , as is naturally expected .
in contrast to the ideal bose gas , the weakly interacting bose gas manifests itself with two new features : first , @xmath495 is discontinuous at @xmath57 ; and second , @xmath495 is linear in temperature when @xmath499 .
evidently , this theoretical result for @xmath495 meets the requirement of the third law of thermodynamics : @xmath500 as @xmath501 .
the discontinuity at @xmath58 is obviously due to the discontinuity of the derivatives of @xmath469 and @xmath502 with respect to @xmath24 . as to the linearity at low temperatures ,
it can be explained as follows . observing that @xmath503 we obtain@xmath504 to the second power of @xmath505 as @xmath506 . inserting it into eq .
( [ multc ] ) , we have @xmath507 , \label{yltc}\end{aligned}\]]it is an iterative equation of @xmath294 . at absolute zero ( @xmath508 )
, it reduces to@xmath509this gives the zeroth order solution of @xmath294 , which we denoted by @xmath510 .
iterating of eq .
( [ yltc ] ) to the second power of @xmath505 leads to@xmath511which shows that @xmath294 , i.e. , @xmath469 and @xmath512 , decreases as @xmath513 in the low temperature region ( @xmath514 ) , as can also be seen directly from figs . 4 and 5 . with the same procedure , one finds@xmath515from eqs .
( [ yltc1 ] ) , ( [ x32ltc ] ) , and ( [ eltc ] ) , it follows that @xmath516 increases as @xmath513 .
therefore , the specific heat @xmath495 will be linear in temperature @xmath24 when @xmath499 . if one expands eqs .
( [ x12ltc ] ) and ( [ x32ltc ] ) further to the fourth power of @xmath505 , he arrives at@xmath517 where the expansion coefficients @xmath518 and @xmath519 are functions of @xmath33 .
appendix [ tawibslt ] shows that this expansion does not depend on the approximations used here , it is valid provided that the interaction is weak .
therefore , it is a fundamental property of the weakly interacting bose gas that the specific heat @xmath495 vanishes linearly as @xmath520 . in connection with the @xmath521-transition occurring in liquid @xmath522 , obviously , the shape of @xmath495 agrees well with that of the specific heat observed experimentally @xcite .
however , the latter is @xmath3 , the specific heat at constant pressure rather than at constant volume .
as is well known from thermodynamics , they differ from each other by a temperature - dependent term,@xmath523 for solids at low temperatures,@xmath524 because @xmath525 that is why the debye @xmath4 law for the phonons , the @xmath24 law for the electron gases in normal metals , and the exponential law for bcs superconductors are observed experimentally in @xmath3 albeit they are the laws of @xmath495 at low temperatures . in other words ,
the theoretical result of @xmath495 for a solid can be verified straightforwardly by the experiments . however , for a liquid , the situation is different , both the volume and pressure are much more sensitive to temperature .
therefore , the second term on the right - hand side of eq .
( [ cpcvdiffer ] ) can not be neglected any longer , the difference between @xmath3 and @xmath495 becomes important and should be taken into account . as a result , to compare the theory with the experiment for a liquid , the experimental data for @xmath495 or the theoretical result for @xmath3 are needed .
regrettably , @xmath495 is difficult to measure experimentally , and @xmath3 is difficult to calculate theoretically .
it is thus hard to compare the theoretical result with the experimental data .
fortunately , @xmath495 tends to zero linearly as @xmath526 for a weakly interacting bose gas , we can compare the theoretical result with the experimental data in the limit @xmath520 .
that can be seen clearly from the following analyses .
according to the third law of thermodynamics , we have@xmath527 where , as pointed in sec .
[ eet ] , @xmath84 represents the thermodynamical entropy . substituting eq .
( [ cvtv ] ) into this equation , we obtain@xmath528 at low temperatures . making use of the maxwell relations , @xmath529 one has@xmath530 .\end{aligned}\ ] ] substituting them into eq .
( [ cpcvdiffer ] ) leads us to@xmath531 where the terms with power higher than cube are omitted , and @xmath532 ^{2}\left ( \frac{\partial v% } { \partial p}\right ) _ { t}. \label{soundcontr}\ ] ] eq .
( [ cplltc ] ) presents the form of @xmath3 at low temperatures within the present theory .
it contains a linear and a cubic term of @xmath24 , the former includes only the contribution from @xmath495 , and the latter includes both the contributions from @xmath495 and the second term on the right - hand side of eq .
( [ cpcvdiffer ] ) . heeding that @xmath533 at @xmath534
, we have lifshitz@xmath535 where @xmath536 denotes the velocity of sound .
this implies that at low temperatures the velocity of sound is approximate to@xmath537 that is to say , the factor @xmath538 on the right - hand side of eq .
( [ soundcontr ] ) represents physically the effect of sound .
therefore , the cubic term of @xmath3 includes the contribution from the sound , which is in accordance with the viewpoint of landau on quantum bose liquid @xmath539 ii @xcite .
this @xmath4 law has already been observed experimentally in liquid @xmath2 at low temperatures @xcite , but it is not an intrinsic property of the bec because sound exists in all fluids .
the intrinsic property of the bec is the linear term of @xmath3 , it comes purely from @xmath495 , and is characteristic of a weakly interacting bose gas at @xmath499 , as mentioned above .
there is no such linear term in the landau theory of quantum bose liquid landau1,landau2 . to our knowledge , this linear behavior of @xmath3 has not yet been reported experimentally for liquid @xmath2 , it is thus a prediction of the present theory .
( [ cplltc ] ) indicates that the temperature @xmath24 has to go much lower for @xmath3 to show the linear behavior than to show the @xmath4 law .
we believe that it would be observed if the measuring temperature is lowered enough , and if liquid @xmath2 could be regarded as a weakly interacting bose gas and the @xmath521-transition were a bose - einstein condensation . * remark * : in general , @xmath502 can not be interpreted as the density of condensed particles .
this can be seen from eq .
( [ multc ] ) , it shows that @xmath540 at @xmath541 , that is , @xmath542at absolute zero . only in the case of the ideal bose gas
, @xmath512 can not be greater than the density of particles , i.e. , @xmath543at any temperature .
physically , @xmath312 is the order parameter and internal spontaneous field of the system , there is no _ a priori _ reason why it must be interpreted using the the particle density .
in fact , any statistical average of observable , including the particle density , will depend explicitly on @xmath312 when @xmath97 .
now , let us discuss the correlation of the bec to the superfluidity .
as usual , we set@xmath544where @xmath545 and @xmath546 . with this , we have@xmath547obviously , @xmath548 is the supercurrent density . if @xmath549 , the phase of the bec order parameter , is independent of @xmath550 , there is no supercurrent , @xmath551 ; otherwise , @xmath552 .
the supercurrent velocity @xmath553 is given by @xmath554since the order parameter @xmath555 is a single - valued function of the position @xmath441 , we have@xmath556that is to say , there can exist quantized vortices in the condensed phase if @xmath557 depends on @xmath441 .
we remark that there exists a critical velocity @xmath558 , the velocity of the supercurrent can not be greater than it , i.e. , @xmath559 .
that is because the increase of the supercurrent velocity is unfavorable to the decrease of the entropy of the system ; the larger the @xmath560 is , the greater the entropy will be , which can be easily seen from the first integral over @xmath441 on the right - hand side of eq .
( [ ibecs ] ) with , e.g. , @xmath561 where @xmath562 .
finally , we shall prove that there can not exist any supercurrent or quantized vortex in the ideal bose gas .
for the ideal bose gas , @xmath563 , eq . ( iop ) reduces to@xmath564obviously , it is the eigenvalue equation of the kinetic energy operator .
this equation can be rewritten as@xmath565^{\dagger } \cdot \lbrack -i\hbar \nabla \eta ( \mathbf{r})]=\mu \int \mathrm{d}\mathbf{r\,}\eta ^{\dagger } ( \mathbf{r})\eta ( \mathbf{r}).\]]it indicates that , for any nontrivial eigenfunction , the chemical potential @xmath566 can not be negative , i.e. , @xmath567 . therefore , there can only exist the trivial solution , @xmath568 , when @xmath569 .
the nontrivial solution , @xmath570 , can exist only when @xmath571 for the ideal bose gas . to discuss the stabilities of those solutions of eq .
( [ keig ] ) , one needs to investigate whether @xmath238 . from eq .
( [ ibecs ] ) , we find @xmath572 , \]]where @xmath573 represents the variation of the order parameter @xmath555 from any one of the solutions of eq .
evidently , @xmath238 when @xmath324 , this means that the system is stable when @xmath324 .
since @xmath568 when @xmath569 and @xmath574 when @xmath278 , the normal phase and the homogeneous condensed phase are both stable , which is in accordance with the results of sec .
[ tlc ] . as to the case of @xmath342 , the solution of eq .
( [ keig ] ) is inhomogeneous , which is just the case we are now interested in because an inhomogeneous solution can produce supercurrent and quantized vortices . to determine whether there can exist any supercurrent and quantized vortex in the condensed phase ,
we first reformulate the above equation as follows,@xmath575 , \]]and then examine the eigenvalue problem , @xmath576eq .
( [ boundary ] ) is the boundary condition obeyed by the variation @xmath577 . obviously , this eigenvalue problem always has solutions when @xmath578 , e.g.,@xmath579where @xmath580 is the spherical bessel function .
accompanying those solutions , @xmath581 can be reexpressed as@xmath582clearly,@xmath583it shows that any solution of eq .
( [ keig ] ) for @xmath342 belongs to the saddle points of the system entropy . in other words ,
any inhomogeneous condensed phase is physically instable , needless to say , the supercurrent and quantized vortex .
this proves that there can not exist any supercurrent or quantized vortex in the ideal bose gas . by the way
, the above analyses demonstrate that the condensed phase of the ideal bose gas must be homogeneous .
there is no inhomogeneous condensation in the ideal bose gas .
this result is very natural , it confirms the einstein s deep physical insight on bec , again . from sec .
[ ipgsb ] , it is learned that an ideal phonon gas can not produce bec .
however , the ideal phonon gas is simply a harmonic approximation to an actual solid . generally , the interatomic interaction is anharmonic .
it thus raises the question as to whether an anharmonic system can produce bec .
we intend to discuss the question in this subsection . to that end
, it is helpful to make a representation transformation to the formulation used in sec .
[ ipgsb ] . there , what we used is the fock space where all the observables are expressed in terms of creation and annihilation operators . in the following ,
we prefer to use the phase space where all the observables are expressed in terms of generalized coordinates and momenta .
as is well known , the two spaces are equivalent and can be transformed into each other according to such a rule @xcite , @xmath584and @xmath585where @xmath586 and @xmath587 are the @xmath46th pair of coordinate and momentum ; @xmath588 and @xmath589 the corresponding annihilation and creation operators ; @xmath590 the effective mass ; and @xmath591 the @xmath46th parameter , which can generally take any positive value , and particularly the natural frequency if @xmath586 and @xmath587 constitute a harmonic oscillator . along with eqs .
( [ transform1 ] ) and ( [ transform2 ] ) , the main quantities concerned are transformed as follows,@xmath592where@xmath593here , @xmath312 represents the order parameter for bec in fock space , @xmath594 and @xmath595 the corresponding order parameters in phase space . under the action of @xmath596
, the statistical averages of @xmath597 and @xmath587 will change with temperature as @xmath598 if the system produces a bec , i.e. , @xmath599 and/or @xmath600 change from zero to nonzero with the decreasing of temperature , the average positions of the particles will redistribute in phase space . in this sense ,
the bec is said to be the phase - space bec , it changes the distribution of the system in phase space .
if , however , the bec only causes @xmath599 ( or @xmath601 ) to change from zero to nonzero , the particles will redistribute merely in momentum space ( or configuration space ) , the subspace of phase space .
in such a case , the bec is said to be the momentum - space bec ( or the configuration - space bec ) , it changes only the distribution of the system in momentum space ( or configuration space ) . obviously , once a bec changes the distribution of a system in configuration space ( @xmath602 , @xmath603 or not ) , it induces a structural phase transition ( spt ) blinc . in this connection
, it can be said that a spt is just an instance of bec .
commonly , a system hamiltonian has the form , @xmath604where @xmath605 and @xmath606 stand for the single- and two - particle potentials , respectively .
one can easily verify , using eqs .
( [ variation ] ) , ( [ increment ] ) , ( [ ptd ] ) and ( [ psentropy ] ) , that @xmath607 must be zero at any temperature , i.e. , there can not arise the momentum - space bec . in consequence , the system with such a form of hamiltonian can , at most , produce a configuration - space bec .
therefore , a spt is commonly a configuration - space bec . in the following
, we shall confine our attention within this kind of spt .
the phase - transition operator @xmath608 can now be simplified as @xmath609for @xmath607 vanishes forever . in the field of spt , the hamiltonian of eq .
( [ sphamiltonian ] ) is made concrete as @xcite , @xmath610where @xmath611 , @xmath174 , and @xmath1 are constants , and @xmath612 denotes the nearest neighbor sites .
the system gets anharmonic if @xmath613 or @xmath614 .
the terms included by the second sum represent the single - particle potential that arises from an underlying sublattice of atoms , which do not participate actively in the phase transition , and the one included by the third sum represents the elastic coupling between nearest neighbors , which is described by a harmonic potential with the coupling constant @xmath615 .
obviously , @xmath93 has the discrete symmetry , @xmath616it leads us to @xmath617where @xmath618this discrete symmetry will break down if a spt takes place . let us begin with the simple case of @xmath619 . making use of eq .
( psd ) , we have @xmath620 \zeta_{i}^{2}+\frac{1}{4}% g\zeta_{i}^{4}\right\ } .
\label{pss}\ ] ] it shows that the entropy @xmath621 has the typical form of landau free energy @xcite . from eqs .
( [ pss ] ) , ( [ variation ] ) , and ( [ increment ] ) , one can draw three conclusions . 1 . the case of @xmath622 .
the system is instable and will collapse because the entropy is unbounded from below with respect to the variation of @xmath601 .
the case of @xmath99 .
( a ) if @xmath623 , the trivial solution @xmath624 is the only minimizer of @xmath625 . that is to say
, a system consisting of harmonic oscillators , or rather an ideal phonon gas , can not produce spt , it will stay in the normal phase forever , which is in accordance with the result of sec .
[ ipgsb ] .
( b ) if @xmath626 , @xmath627 can take a temperature - independent value , it represents , as indicated by eqs .
( [ distortions ] ) and ( [ qbar ] ) , the equilibrium position of @xmath628 , i.e. , @xmath629 . as the initial equilibrium position of @xmath586 at @xmath358 , @xmath630 can be chosen as zero .
therefore , a system composed of free distinguishable particles can not produce spt , as is just expected .
( c ) if @xmath631 , @xmath632 is the only solution , but it is a maximizer of @xmath625 , the system is instable and will collapse .
this result is quite natural since all particles go in a reversed harmonic potential .
the case of @xmath9 .
( a ) if @xmath633 , @xmath634 is the only solution , and it is a minimizer of @xmath621 . there is no phase transition in the system , the potential is a single well .
( b ) if @xmath635 , the potential gets double welled , and it must be determined by the position fluctuation @xmath636 whether the system can produce a phase transition . evidently , a transition can occur if the fluctuation is such a decreasing function of temperature that @xmath637 can change in sign , from positive to negative .
if the fluctuation is so strong that @xmath638>0 $ ] up to zero temperature , the system has to stay in the normal phase forever . from the above discussions , it follows that a double - well potential where @xmath639 and @xmath9 is the only possible case for the system to produce a structural phase transition under the condition of @xmath619 . one can easily verify that this conclusion is still valid for @xmath640 .
we are now confronted with performing the evaluation of the position fluctuation @xmath636 so as to ascertain whether a phase transition can occur or not if @xmath631 and @xmath9 . to this end , we need to perform a representation transformation to the hamiltonian of eq .
( [ pshamiltonian ] ) because , as well - known , this form of hamiltonian has no good basic part and is thus unfavorable for doing approximations . complying with quantum field theory ,
the transformation can be achieved as follows , @xmath641where @xmath642 in the sense of classical mechanics , the above transformation amounts to performing a translation so that the new coordinate @xmath643 is referenced from @xmath644 , one of the two minimizers of the double - well potential . along with eq .
( [ qpt ] ) , the new representation @xmath645 of the hamiltonian becomes@xmath646the first sum on the right - hand side represents harmonic oscillators with natural frequencies @xmath647 , it constitutes a basic hamiltonian which is favorable for approximation handling , the second the interaction , and the third an unimportant constant . with @xmath648
, one finds @xmath649 where @xmath650 stands for the statistical average with respect to @xmath651,@xmath652 where @xmath653 . according to eq .
( qfluctuation ) , we now just need to evaluate the fluctuation @xmath654 , it can be worked out by the retarded green s function @xmath655 which is defined with respect to @xmath656 @xcite .
the green s function satisfies the following equation of motion , @xmath657\rangle+\langle\langle\lbrack q_{i},\widetilde { h}% ( q , p)]|q_{i}\rangle\rangle_{\omega } \notag \\ & = \frac{i\hbar}{m}\langle\langle p_{i}|q_{i}\rangle\rangle_{\omega}.\end{aligned}\ ] ] also,@xmath658\rangle+\langle\langle\lbrack p_{i},\widetilde { h}% ( q , p)]|q_{i}\rangle\rangle_{\omega } \notag \\ & = -i\hbar\big[[1 - 2\mu^{2}\langle\langle q_{i}|q_{i}\rangle\rangle_{\omega } + 3g\nu\langle\langle q_{i}^{2}|q_{i}\rangle\rangle_{\omega } \notag \\ & + g\langle\langle q_{i}^{3}|q_{i}\rangle\rangle_{\omega}\big].\end{aligned}\ ] ] we would like to truncate the gf chain with the following decoupling approximation,@xmath659\langle\langle q_{i}|q_{i}\rangle \rangle_{\omega},\end{gathered}\ ] ] where we have used the result @xmath660 , which follows from eq .
( [ qbar ] ) .
this approximation leads us to@xmath661 where @xmath662 ^{1/2 } \label{omegas}\ ] ] represents the renormalized frequency .
according to the fluctuation - dissipation theorem @xcite , the fluctuation @xmath663 can be obtained by@xmath664 , \ ] ] it gives rise to@xmath665 eq .
( [ fluctuation ] ) and eq .
( [ omegas ] ) constitute a self - consistent approximation to the fluctuation @xmath636 . evidently , the first term on the right - hand side of eq .
( [ fluctuation ] ) represents the zero - point fluctuation .
( [ fluctuation ] ) shows that the fluctuation gets weaker and weaker when temperature goes lower and lower . at zero temperature , one has@xmath666{4mg}}\hbar ^{2/3}\sim o(\hbar^{2/3}),\ ] ] and at high temperatures , @xmath667 they imply that there exists a finite temperature @xmath6 such that@xmath668 from eq .
( [ signch ] ) and eq .
( [ pss ] ) , it follows that a spt will occur at @xmath6 , @xmath669 ^{1/2},\text { } & t < t_{c}.% \end{array } \right . \label{zetai}\ ] ] specifically , @xmath670 at zero temperature if the zero - point fluctuation is neglected , i.e. , the order parameter will equal one of the two minimizers of the double - well potential at zero temperature when the zero - point fluctuation is left out of consideration . combining eqs .
( [ zetai ] ) , ( [ distortions ] ) , and ( [ qbar ] ) , we have @xmath671 ^{1/2},\text { } & t
\end{array}% \right.\]]it shows that for each site the distortion @xmath672 can take independently either of the two values at @xmath97 , there is no correlation between the distortions of different sites , therefore , the new phase below @xmath6 is a structural glass .
physically , that is because we have discarded the coupling between nearest neighbors ( @xmath673 ) . if it is taken into account ( @xmath674 ) , one can easily verify that the new phase below the transition temperature will be ferrodistortive : @xmath675where @xmath46 and @xmath48 denote any two different sites .
this result is in agreement with the classical mean - field theory @xcite . in conclusion , we find that the double - well potential system described by the hamiltonian of eq .
( [ pshamiltonian ] ) can produce a structural phase transition .
mechanically , that is because the position fluctuations of particles will decrease with temperature . in the high - temperature regime ,
the fluctuations are relatively strong , the system can only stay in its normal phase .
the fluctuations will get weaker and weaker when temperature goes lower and lower , and finally the system undergoes a transition at a finite temperature and then turns into a new ordered phase in the low - temperature regime .
we learn from here that the position fluctuations of particles play an important role in structural phase transitions .
also , we see that configuration space is very convenient for describing spt , in contrast to the fock space . obviously , the preceding theory for the discrete case can be extended straightforwardly to the continuum case .
let us consider the so - called @xmath676-symmetric vector model @xcite , @xmath677 & = \int \mathrm{d}\mathbf{x}\text { } \big[\frac{1}{2% } \pi ^{\dagger } \pi + \frac{1}{2}\nabla \psi ^{\dagger } \cdot \nabla \psi \notag \\ & + \frac{m^{2}}{2}\psi ^{\dagger } \psi + \frac{g}{4}(\psi ^{\dagger } \psi ) ^{2}\big],\end{aligned}\]]where @xmath678 and @xmath9 are the two constant parameters for a double - well potential , @xmath679 is usually called higgs field , it represents a real - valued vector field with @xmath34 components , and @xmath680 the corresponding momentum , @xmath681=i\delta _ { ij}\delta ( \mathbf{x}-\mathbf{x}^{\prime } ) .\]]as
is well known , this double - well hamiltonian is the typical model for ssb in quantum field theory @xcite .
the arguments given in the preceding subsection are still valid .
they reveal that the physical reason why the @xmath676 symmetry will break down is that the fluctuation of the field @xmath679 decreases monotonically with temperature , and that this ssb belongs essentially to the bec happening in configuration space . in the condensation phase
, the system hamiltonian takes the form , @xmath682 = e^{id\left [ \varphi , \pi % \right ] } h\left [ \psi , \pi \right ] e^{-id\left [ \varphi , \pi \right ] } , \label{hphipsipi}\]]where @xmath683 is the order parameter for higgs field , and @xmath684 $ ] the phase - transition operator , @xmath685 = \int \mathrm{d}\mathbf{x}\text { } \varphi ^{\dagger } ( \mathbf{x})\pi ( \mathbf{x}).\ ] ] in more detail , eq . ( [ hphipsipi ] ) can be written as@xmath682 = \int \mathrm{d}\mathbf{x}\text { } \left\ { \frac{1}{2}\pi ^{\dagger } \pi + \frac{1}{2}\nabla ( \psi ^{\dagger } + \varphi ^{\dagger } ) \cdot \nabla \left ( \psi + \varphi \right ) + \frac{m^{2}% } { 2}(\psi ^{\dagger } + \varphi ^{\dagger } ) \left ( \psi + \varphi \right ) + % \frac{g}{4}[(\psi ^{\dagger } + \varphi ^{\dagger } ) \left ( \psi + \varphi \right ) ] ^{2}\right\ } .
\label{hphipsipi1}\ ] ] at absolute zero , if one neglects the zero - point fluctuation of the field @xmath686 , i.e. , @xmath687 , then he has the stable solution , @xmath688where@xmath689is the minimizer of the double - well potential . without loss of generality
, we shall take @xmath690 .\]]substituting it into eq .
( [ hphipsipi1 ] ) leads us to @xmath682 = \int \mathrm{d}\mathbf{x}\text { } \left [ \frac{1}{2}\pi ^{\dagger }
\pi + \frac{1}{2}\nabla \psi ^{\dagger } \cdot \nabla \psi -m^{2}\psi _ { 1}^{2}+g\nu \psi _ { 1}\psi ^{\dagger } \psi + % \frac{g}{4}(\psi ^{\dagger } \psi ) ^{2}-\frac{g\nu ^{4}}{4}\right ] .\ ] ] that is the zero - temperature representation of the system hamiltonian , it shows that the first component field @xmath691 gets a mass of @xmath692 , the others are massless .
as well - known , those massless particles are the so - called goldstone bosons @xcite . from here , one can see how goldstone bosons are produced in a natural way by spontaneous symmetry breaking within the framework of the extended ensemble theory .
of course , he should also recognize that the zero - point fluctuation can impose a little influence on the goldstone bosons . at last
, it should be emphasized that , in the @xmath676-symmetric vector model , goldstone bosons originate physically from the bose - einstein condensation of higgs field .
if a gauge field is coupled with the higgs field @xmath693 , what will happen to the gauge field ?
can it produce bec as the higgs field @xmath694 ? to clarify this problem , let us consider , for simplicity , the coupling of a complex - valued ( or two - component ) higgs field @xmath695 and an electromagnetic field @xmath696 , @xmath697 = \int\mathrm{d}\mathbf{x}\text { } \left [ \pi^{\dagger}\pi+\left ( \nabla+ie\mathbf{a}\right ) \psi^{\dagger } \cdot\left ( \nabla - ie\mathbf{a}\right ) \psi+m^{2}\psi^{\dagger}\psi + g(\psi^{\dagger}\psi)^{2}+\frac{1}{2}\mathbf{e}^{2}+\frac{1}{2}\left ( \nabla\times\mathbf{a}\right ) ^{2}\right ] , \ ] ] where @xmath698 is the charge of the higgs field @xmath693 , and @xmath699 the electric field strength which plays the role of the canonical momentum corresponding to the canonical coordinate @xmath700,@xmath701=i\delta_{ij}\delta(\mathbf{x}-\mathbf{x}^{\prime}).\ ] ] here , the temporal gauge has been used to perform a canonical quantization to the electromagnetic field , an abelian gauge field .
the reason for choosing this gauge is that it can be easily applied to quantize non - abelian gauge fields , and transformed into other gauges through the path - integral formalism introduced by faddeev and popov @xcite . for this coupled system
, bec can be explored through the phase - transition operator , @xmath702 = \int\mathrm{d}% \mathbf{x}\text { } \left ( \varphi\pi+\varphi^{\dagger}\pi^{\dagger } -\mathbf{% \lambda}\cdot\mathbf{e}\right ) , \label{dpilambda}\ ] ] where @xmath703 and @xmath704 are the order parameters for the higgs field @xmath693 and gauge field @xmath705 , respectively . with @xmath706 $ ]
, one can obtain the entropy of the system , @xmath707 & = -\mathrm{tr}\big(\ln ( \rho(h% \left [ \psi,\pi;\mathbf{a , e}\right ] ) ) \rho(h\left [ \psi , \pi;\mathbf{a , e}% \right ] ) \big)+\int\mathrm{d}\mathbf{x}\big[\left ( \nabla+ie\mathbf{\lambda}% \right ) \varphi^{\dagger}\cdot\left ( \nabla -ie\mathbf{\lambda}\right ) \varphi
\notag \\ & + ( m^{2}+e^{2}\overline{\mathbf{a}\cdot\mathbf{a}}+4g\overline{% \psi^{\dagger } \psi})\varphi^{\dagger}\varphi+g\left ( \varphi^{\dagger}\varphi\right ) ^{2}+e^{2}\overline{\psi^{\dagger}\psi}% \mathbf{\lambda}^{2}+\frac{1}{2}\left ( \nabla\times\mathbf{\lambda}\right ) ^{2}\big ] , \label{coupledphilambda}\end{aligned}\ ] ] which shows that the two order parameters @xmath708 and @xmath709 are coupled together .
( [ coupledphilambda ] ) reminds us of the landau free energy for superconductivity @xcite , indeed , they are both similar in form .
variation of @xmath74 with respect to @xmath708 and @xmath709 yields@xmath710 this form of two coupled nonlinear equations is known as ginzburg - landau equations @xcite . observing that@xmath711 where @xmath712 denotes the magnetic field strength , one can specify@xmath713 which indicates that there are only two independent order parameters for the electromagnetic field .
( [ coulomblambda ] ) can be regarded as the coulomb gauge to the vector order parameter @xmath709 . from eqs .
( [ coupledphilambda][coulomblambda ] ) , one can easily verify that there is no bec for the electromagnetic field @xmath714 if it is not coupled with the higgs field @xmath715 , or it is coupled with @xmath715 but the latter has not condensed yet ( @xmath716 ) .
this also confirms the conclusion of sec .
[ ipgsb ] : an ideal photon gas , or a free electromagnetic filed , can not produce bec . once the higgs field @xmath715 gets condensed ( @xmath717 ) , the electromagnetic field @xmath714 must condense down ( @xmath718 ) simultaneously ; otherwise , there can arise a linear term of @xmath719 in @xmath720 , the coupled system will go instable .
this shows that the gauge field @xmath721 will condense together with the higgs fields @xmath715 when it is coupled with the latter . in such case , the transition is cooperative , one usually says that @xmath722 is the primary order parameter , and @xmath723 the secondary one . with the help of eq .
( [ coulomblambda ] ) , eq .
( [ gl2 ] ) can be reduced as@xmath724 heeding that @xmath725 , eq .
( [ procaeq ] ) is identical to the static proca equation @xcite except that the prefactor of @xmath709 may be a function of @xmath441 , the counterpart in proca equation just a constant .
in general , the prefactor can not be interpreted as the mass of the gauge field . however , once there is a constant occurring in the prefactor , a mass @xmath726 is generated to the gauge field , as can be seen more clearly in the following discussion . owing to the nonlinearity ,
the ginzburg - landau equations ( [ gl1 ] ) and ( [ gl2 ] ) are rather complicated to handle . as usual , let us consider the zeroth - order approximation of eq .
( [ gl1 ] ) at @xmath541 gl , lifshitz , @xmath727 where the effect of @xmath709 and the zero - point fluctuations of the fields @xmath728 and @xmath714 are omitted . from eqs .
( [ procaeq ] ) and ( [ nutzero ] ) , it follows that@xmath729 since the prefactor @xmath730 is constant now , it can be interpreted as the mass of the gauge field at zero temperature,@xmath731 as well - know , this mass can account for the meissner effect @xcite , i.e. , the expulsion of a magnetic field from the interior of a superconducting material , with a london penetration length @xmath732 @xcite .
the above formalism is ginzburg - landau picture of mass producing , in this picture , the mass is represented by the equation of motion .
there is another picture , which is due to higgs @xcite . in higgs picture ,
the mass is represented directly by the system hamiltonian itself .
since an equation of motion is in accordance with its hamiltonian , both pictures are equivalent physically .
to show higgs picture , we aught to consider the zero - temperature representation @xmath733 of the system hamiltonian,@xmath734 } h\left [ \psi,\pi;\mathbf{a , e}\right ] e^{-id\left [ \varphi,\pi \mathbf{;\lambda},% \mathbf{e}\right ] } \notag \\ & \equiv h[\chi,\pi;\mathbf{w , e}],\end{aligned}\ ] ] where@xmath735 } \psi(\mathbf{x})e^{-id\left [ \varphi,\pi\mathbf{;\lambda } , \mathbf{e% } \right ] } , \label{chifield } \\
\mathbf{w}(\mathbf{x } ) & = e^{id\left [ \varphi,\pi\mathbf{;\lambda } , \mathbf{e% } \right ] } \mathbf{a}(\mathbf{x})e^{-id\left [ \varphi , \pi\mathbf{;\lambda},% \mathbf{e}\right ] } .\end{aligned}\ ] ] as is well known , within the temporal gauge , the electromagnetic field still has gauge degrees of freedom , the hamiltonian @xmath736 $ ] remains invariant under any time - independent gauge transformation . as another representation of @xmath736 $ ] , the hamiltonian @xmath737 $ ] is also gauge invariant,@xmath738 = e^{ig\left [ \theta,\chi , \pi,\mathbf{e}% \right ] } h\left [ \chi,\pi;\mathbf{w , e}\right ] e^{-ig\left [ \theta,\chi,\pi,% \mathbf{e}\right ] } , \ ] ] where @xmath739 and@xmath740 = \int\mathrm{d}\mathbf{x}\text { } % [ -ie\theta\left ( \pi\chi-\chi^{\dagger}\pi^{\dagger}\right ) + \nabla \theta\cdot\mathbf{e}].\ ] ] as higgs @xcite did , we can re - gauge the fields @xmath741 and @xmath742 by letting @xmath743 equal to the phase of @xmath744 , that is,@xmath745 so as to eliminate the goldstone bosons represented by the phase of @xmath741 .
upon such a choice of @xmath746 , the hamiltonian @xmath747 can be expressed as@xmath748 , \label{hcoupled}\end{aligned}\ ] ] where @xmath749 } \pi(\mathbf{x})e^{-ig\left [ \theta,\chi,\pi,\mathbf{e}\right ] } , \label{piwave } \\
\widetilde{\mathbf{a}}(\mathbf{x } ) & = e^{ig\left [ \theta,\chi,\pi , \mathbf{e}% \right ] } \mathbf{w}(\mathbf{x})e^{-ig\left [ \theta,\chi , \pi,\mathbf{e}% \right ] } .\end{aligned}\ ] ] from eqs .
( [ chifield ] ) and ( [ dpilambda ] ) , it follows that @xmath750 without loss of generality , we can , from eq .
( [ nutzero ] ) , choose @xmath751 as real,@xmath752 that leads to@xmath753 where@xmath754 } \psi(\mathbf{x})e^{-ig\left [ \theta,\widetilde{\psi},\pi,\mathbf{e}% \right ] } .\ ] ] eq .
( [ rhophi ] ) indicates that @xmath755 is merely a real field .
substitution of eq .
( [ rhophi ] ) into eq .
( [ hcoupled ] ) gives rise to@xmath756 , \label{higgsrepre}\end{aligned}\ ] ] where@xmath757/\sqrt{2 } , \\
p_{0}(\mathbf{x } ) & = i[\widetilde{\pi}^{\dagger}(\mathbf{x})-\widetilde{\pi } % ( \mathbf{x})]/\sqrt{2}.\end{aligned}\ ] ] paying attention to@xmath758 } \left ( \frac{\chi ( \mathbf{x}% ) + \chi^{\dagger}(\mathbf{x})}{\sqrt{2}}\right ) e^{-ig\left [ \theta,\chi,\pi,% \mathbf{e}\right ] } = \phi(\mathbf{x})+\nu,\ ] ] we have @xmath759 = -i\delta ( \mathbf{x}-% \mathbf{x}^{\prime } ) , \\
\left [ p_{0}(\mathbf{x}),h_{0}\right ] = 0.\end{gathered}\ ] ] that is to say , the @xmath755 and @xmath760 become a new pair of canonical field operators whereas @xmath761 is just a constant . physically , that is because @xmath761 corresponds to the eliminated degree of freedom @xmath762 .
( [ higgsrepre ] ) forms higgs picture of mass producing . comparing it with eq .
( [ glpicture ] ) , one sees immediately that the mass of the gauge field in higgs picture is identical to that in ginzburg - landau picture , as is expected .
the above theory for a complex field coupled with an abelian gauge field can be easily generalized to the case of an @xmath34-component field coupled with a non - abelian gauge field , with the same conclusion that the gauge field will obtain a mass through spontaneous symmetry breaking .
this manner of mass producing is the well - known higgs mechanism , which was successively contributed by london @xcite , ginzburg and landau @xcite , anderson @xcite , higgs @xcite , and kibble @xcite .
* remark * : spontaneous symmetry breaking can occur without any fundamental higgs field in superconductivity , it is thus called `` dynamical symmetry breaking '' , which implies that higgs fields are `` normally '' needed for symmetry breaking . from the viewpoint of the extended ensemble theory , that is a prejudice because superconductivity , goldstone bosons and higgs mechanism share the same physical ground .
this is in accordance with the standpoint of huang expressed in his book @xcite .
so far , a possible extension of gibbs ensemble theory has been postulated , as a personal review of the author , to enable a microscopic description to phase transitions and spontaneous symmetry breaking .
the extension is founded on three hypotheses , which root , in physics , from landau s ideas on phase transitions : order parameter , variational principle , representation transformation , and spontaneous symmetry breaking . in this sense
, the extended gibbs ensemble theory can be viewed as a microscopic realization of the landau phenomenological theory of phase transitions . within the framework of the extended ensemble theory ,
a phase transition occurs according to the principle of least entropy , which manifests itself as an equation of motion of order parameter .
this equation determines the evolution of order parameter with temperature , and thus controls the change in representation of system hamiltonian .
different phases correspond to different representations , and vice versa .
a system hamiltonian will realize its symmetric representation in the disordered phase , and asymmetric one in the ordered phase .
the change in symmetry results from the change in representation .
that is the landau mechanism responsible for phase transitions and spontaneous symmetry breaking in the extended ensemble theory , it holds in the whole range of temperature , including the critical region .
physically , phase transitions originate from the wave nature of matter .
a system always stays in its normal phase at sufficiently high temperatures , which is disordered and structureless .
as temperature decreases , matter waves will interfere automatically with one another .
this interference makes the system structured and transform into its ordered phase . that is the physical picture for phase transitions in the extended ensemble theory .
also , the extended ensemble theory has been applied to typical quantum many - body systems , with the conclusions as follows .
for the ideal fermi gas , it can not produce superconductivity , its normal phase is stable at any temperature .
negative - temperature and laser phases arise from the same mechanism as phase transitions , they are both instable , and will finally turn into the positive - temperature phase . that is the microscopic interpretation for negative temperatures and laser phases . for the conventional weak - coupling low-@xmath6 superconductors ,
the mean - field solution due to bardeen , cooper and schrieffer is derived anew , and proved to be stable within the extended ensemble theory . for the ideal bose gas
, it can produce bose - einstein condensation only in the thermodynamic limit , which agrees with einstein s prediction .
but that holds in the sense of lebesgue integration rather than riemann integration . in general
, the order parameter for bec can not be interpreted as the number of condensed particles . besides
, there can not exist any supercurrent or vortex in the ideal bose gas , the condensation is always homogeneous .
the ideal phonon gas can not produce bose - einstein condensation , neither the photon gas in a black body .
it is not admissible to quantize dirac field using bose - einstein statistics .
otherwise , the system is instable . that is a statistical rather than mechanical reason why dirac field has to be quantized using fermi - dirac statistics
this conclusion is in accordance with the spin - statistics theorem in quantum field theory .
a structural phase transition belongs physically to the bose - einstein condensation occurring in configuration space . for a double - well anharmonic system , it is instable at low temperatures , and will undergo a structural phase transition at a finite temperature . physically ,
that is because the position fluctuation is a monotonically decreasing function of temperature .
for the @xmath676-symmetric vector model , the @xmath676 symmetry will break down spontaneously , with the presence of goldstone bosons .
in essence , this ssb is a bose - einstein condensation .
if the system is coupled with a gauge field , it can cause the latter to condense together with itself .
the cooperative condensation can be described by the ginzburg - landau equations .
after the condensation , the gauge field can obtain a mass , which is the so - called higgs mechanism . for an interacting bose gas ,
it is stable only if the interaction is repulsive .
the bec present in this system can be described by the generalized ginzburg - landau equation .
if the interaction is weak , the bec is a @xmath521-transition , and its transition temperature can be lowered by the repulsive interaction . as a characteristic property of this @xmath521-transition
, the specific heat at constant volume @xmath495 will vanish linearly as @xmath520 .
if liquid @xmath2 could be regarded as a weakly interacting bose gas , and if the @xmath1-transition were a bose - einstein condensation , then its specific heat at constant pressure @xmath3 would show a @xmath4 law at low temperatures , which is in agreement with the experiment .
however , if temperature goes lower further , it is predicted that @xmath3 will exhibit a linear behavior at relatively lower temperatures , which is in need of experimental verifications .
the author would like to thank prof .
zheng - zhong li , prof .
jin - ming dong , and prof .
wei - yi zhang for their helpful discussions .
he is also deeply grateful to prof .
vladimir rittenberg for his valuable advice on how to revise the paper .
in this appendix , we shall first show that , when @xmath763 , the ideal bose gas is still well defined within the extended ensemble theory , and then discuss the frequency summation , which is very useful when we attempt to solve problems through green s functions . upon a transformation as in eq .
( [ sphi ] ) , the statistical average defined by eq .
( [ average ] ) can always be transformed into a statistical average with respect to the hamiltonian @xmath764 of eq .
( [ hibg ] ) .
it is thus sufficient for us to show that the statistical average with respect to @xmath765 is properly defined when @xmath342 .
as is well known , every statistical average can be calculated through a corresponding green s function , and every green s function can be derived from the generating functional of the system .
therefore , we need only to prove that the generating functional is properly defined when @xmath342 . according to the path - integral formalism @xcite , the generating functional @xmath766 $ ] for @xmath764
can be written as follows , @xmath767=\mathcal{n}^{-1}\int \mathcal{d}b^{+}\mathcal{d}b\text{% \thinspace } \exp \!\!\left ( \frac{1}{\hbar } \sum_{\mathbf{k},n}\left\ { b_{% \mathbf{k,}n}^{+}\left [ i\hbar \omega _ { n}-\left ( \frac{\hbar ^{2}\mathbf{k}% ^{2}}{2m}-\mu \right ) \right ] b_{\mathbf{k},n}-\phi _ { \mathbf{k},n}b_{% \mathbf{k,}n}^{+}-\phi _ { \mathbf{k,}n}^{+}b_{\mathbf{k},n}\right\ } \right ) , \label{wgenerate}\ ] ] where @xmath768 ( @xmath769 ) denotes the mutsubara frequency , and @xmath770 the normalization factor , @xmath771 b_{\mathbf{k},n}\right ) . \label{ngaussian}\ ] ] if @xmath569 , one gets immediately the familiar result , @xmath767=\exp \!\!\left ( -\frac{1}{\hbar } \sum_{\mathbf{k},n}\phi _ { \mathbf{k},n}\frac{1}{i\hbar \omega _ { n}-\left ( \frac{\hbar ^{2}\mathbf{k}% ^{2}}{2m}-\mu \right ) } \phi _ { \mathbf{k,}n}^{+}\right ) .
\label{wwg}\ ] ] since @xmath772 , the normalization factor @xmath770 converges .
when @xmath567 , the normalization factor @xmath770 will diverge .
however , that does no harm because the normalization factor can be cancelled by the numerator of eq .
( [ wgenerate ] ) , as is frequently encountered in quantum field theory @xcite . after the cancellation , the rest is still eq .
( [ wwg ] ) except @xmath567 . as mentioned in sec .
becideal , an unbounded function as integrand is permissible within lebesgue integration .
therefore , the sum in the exponent of eq .
( [ wwg ] ) is proper as a lebesgue integral .
the resulting @xmath766 $ ] is also right because a function is permitted to be unbounded in lebesgue integration . in a word ,
the generating functional for @xmath764 is always properly defined no matter how large the chemical potential @xmath102 is , which ends our proof .
now , there is no mathematical limit on the chemical potential of the ideal bose gas , as is the case for the ideal fermi gas .
that is rational and significant , any limit on chemical potential should be set by the physical theory itself rather than by the mathematical theory involved . in the extended ensemble theory ,
the chemical potential of a system is determined only by the physical requirement : the conservation of particles , whether the system obeys bose - einstein statistics or fermi - dirac statistics ; there is no limit set by the mathematical tool , i.e. , lebesgue integration . as a consequence of eq .
( [ wwg ] ) , we obtain@xmath773}{\delta\phi_{\mathbf{k,}n}^{+}\delta \phi_{% \mathbf{k},n } } \notag \\ & = \frac{1}{i\omega_{n}-\hbar^{-1}\left ( \frac{\hbar^{2}\mathbf{k}^{2}}{2m}% -\mu\right ) } , \end{aligned}\ ] ] where @xmath774 represents the temperature green s function which is defined as @xcite@xmath775 as a function of @xmath128 , @xmath774 can be unbounded if @xmath776 , but that will be all right because @xmath777 can only appear in the integrand of a lebesgue integral over @xmath128 .
those discussions indicate that , in the manipulations of @xmath778 and @xmath779 , one can benefit greatly from lebesgue integration .
that is because lebesgue integration is designed , _ ab initio _
, on the set of extended real numbers : @xmath780 where @xmath781 @xcite .
evidently , those benefits can not be provided by riemann integration .
more benefits of lebesgue integration will be seen from the following discussions .
when one evaluates the statistical average of an observable , he will encounter the summation over mutsubara frequencies . as an illustration ,
let us consider @xmath782where @xmath783 .
we shall discuss the thermodynamic limit case . here
, it is important to note that the frequency summation always goes ahead of the lebesgue integral over @xmath128 because the thermodynamic limit must be taken at last . to perform the frequency summation , it is helpful to make use of the complex function , @xmath784which has simple poles at the mutsubara frequencies ,
i.e. , @xmath785 , and transform the summation into a contour integral @xcite,@xmath786where the contour @xmath787 is depicted in fig .
it should be emphasized that the above procedure requires that the function @xmath788 must be analytic on the whole imaginary axis lest there be any poles other than @xmath789 on this axis .
sometimes , this requirement can not be met by every @xmath128 , e.g.,@xmath790each such @xmath788 that satisfies @xmath791 is singular on @xmath792 , and thus can not meet the requirement . if such singular @xmath128 s constitute just a null set @xmath793 , i.e. , @xmath794 satisfies the requirement almost everywhere on @xmath339 , which is the usual case , then those @xmath128 s can be left out of consideration for they give merely a null contribution to the lebesgue integral over @xmath128 .
therefore , we can rewrite eq .
( [ summation ] ) as @xmath795where @xmath796 . here
the set @xmath793 is removed form @xmath339 so that the function @xmath788 can meet the requirement everywhere on @xmath329 .
generally , @xmath788 are analytic on both the upper and lower half complex-@xmath295 planes @xcite , one can therefore deform the contour @xmath787 into the contours @xmath797 and @xmath798 , which results in @xmath799the contribution from the contour @xmath797 vanishes owing to the convergence factor @xmath800 @xcite .
the right - hand side can be simplified as @xmath801where @xmath802 is the spectral intensity , @xmath803 , \]]and @xmath460 represents the principal value , @xmath804 substituting eqs .
( [ sum1][sum2 ] ) into eq .
( [ sum0 ] ) , one has @xmath805
\notag \\ & = & \lim_{\alpha \rightarrow 0^{+}}\int_{e}\mathrm{d}\mathbf{k}\left [ \int_{-\infty } ^{-\alpha } \mathrm{d}\omega \,f(\omega ) \mathcal{a}(\mathbf{k}% , \omega ) + \int_{\alpha } ^{+\infty } \mathrm{d}\omega \,f(\omega ) \mathcal{a}(% \mathbf{k},\omega ) \right ] \notag \\ & = & \lim_{\alpha \rightarrow 0^{+}}\left [ \int_{-\infty } ^{-\alpha } \mathrm{d}% \omega \,f(\omega ) \int_{e}\mathrm{d}\mathbf{k}\mathcal{a}(\mathbf{k},\omega ) + \int_{\alpha } ^{+\infty } \mathrm{d}\omega \,f(\omega ) \int_{e}\mathrm{d}% \mathbf{k}\mathcal{a}(\mathbf{k},\omega ) \right ] \notag \\ & = & \mathcal{p}\int_{-\infty } ^{+\infty } \mathrm{d}\omega \,f(\omega ) \int_{e}% \mathrm{d}\mathbf{k}\mathcal{a}(\mathbf{k},\omega ) .\end{aligned}\ ] ] now , supplementing the set @xmath329 with the null set @xmath793 , one arrives at@xmath806as usual , by introducing the density of states @xmath807 , @xmath808eq .
( [ bsum ] ) can be reduced into@xmath809this is a very useful identity , as its applications , one can easily obtain the final results of eqs .
( [ lesb1 ] ) and ( [ lesb3 ] ) .
it should be stressed that eq .
( [ principal ] ) holds whether the chemical potential @xmath566 is less than , equal to , or lager than zero .
if @xmath569 , it reduces to the common result given in ref .
@xcite . it is a generalization of the common result when @xmath567 .
this generalization removes the mathematical limit on the chemical potential .
one can not reach the result of eq .
( [ principal ] ) within riemann integration .
obviously , the above analyses for the ideal bose gas are also suitable for other bose systems , e.g. , the interacting bose gas , which will be studied in sec . [ wibg ] .
* remark * : * * * * if the system obeys fermi - dirac statistics , the principal value is unnecessary . instead of eq .
( [ principal ] ) , one has@xmath810that is simply because the fermi distribution function is bounded and continuous at @xmath811 for any finite temperature @xmath812 . a more direct reason for the unnecessity of the principal value consists in the fact that , unlike the bose system , @xmath792 is not the pole of the complex function , @xmath813because the mutsubara frequencies for a fermi system are all nonzero , i.e. , @xmath814 ( @xmath769 ) . as a consequence ,
the integration contour can goes as in fig .
8 , which gives straightforwardly the final result of eq .
( [ fermi ] ) . evidently , that makes the fermi system easier to handle mathematically than the bose system .
in sec . [ wibg ] , our discussions were confined within the hartree - fock approximation of eqs .
( [ ibechf1][ibechf2 ] ) and the short - range approximation of eq .
( [ ibecsr ] ) . within those approximations
, one sees from eqs .
( [ x12ltc ] ) and ( [ x32ltc ] ) that the thermal activations at low temperatures can be expanded into even power series of @xmath24 .
we shall show that , beyond those approximations , this feature still holds as long as the interaction is weak and perturbative .
under the condition assumed , observables can be calculated via gf .
the calculations can be transformed finally into the integrals of the following form , @xmath815where @xmath816 is a real - valued function of @xmath817 and @xmath469 , which comes from a summation of gf over @xmath128 . as a function of @xmath817 , @xmath818 should be bounded at bottom or decrease fast with @xmath817 so that the integral can converge . in both cases , we can cut off the integral from below at a certain frequency @xmath819 ( @xmath820 ) , that is , @xmath821according to the definition of principal value , the right - hand side can be written as observe that @xmath823we have@xmath824integrated by substitution , it turns into@xmath825the second integral on the right - hand side vanishes exponentially as @xmath826 , it can thus be neglected .
the upper limit of the third integral can be set equal to @xmath778 at low temperatures for the integrand decreases as @xmath827 when @xmath293 becomes very large , @xmath828the second integral on the right - hand side can expanded as@xmath829substituting it into eq .
( [ pomega ] ) results in@xmath830the sum is an even power series of @xmath24 .
to the fourth order , it becomes @xmath831 it is sufficient for the discussion of the low - temperature properties of the system .
this result implies that the renormalized chemical potential @xmath490 and the internal energy @xmath329 are even functions of @xmath24 at low temperatures , which leads to the conclusion of eq .
( [ cvtv ] ) : the specific heat @xmath495 of a weakly interacting bose gas will vanish linearly as @xmath501 . the above result reminds us of the specific heat of the electron gas in a normal metal , which also vanishes linearly as @xmath501 .
this linear behavior of the electron gas can be explained analogously .
there , what is concerned is the integral,@xmath832where @xmath102 is the chemical potential .
along the same way as for the bose gas , it can be expressed as which is sufficient for the low - temperature properties of the electron gas in a normal metal because its fermi temperature is very high , in comparison to the critical temperature of the bec happening in a weakly interacting bose gas .
( [ lowfermi ] ) also implies that the chemical potential @xmath835 and the internal energy @xmath329 are even functions of @xmath24 at low temperatures , and that the specific heat at constant volume @xmath495 will vanish linearly as @xmath501 .
for example , one can easily verify , with the help of eq .
( [ lowfermi ] ) , that the @xmath102 , @xmath329 and @xmath495 for the ideal fermi gas have the low - temperature forms,@xmath836 , \\ & & \frac{e}{n}=\frac{3}{5}\mu ( 0)\left [ 1+\frac{5\pi ^{2}}{12}\left ( \frac{% k_{b}t}{\mu ( 0)}\right ) ^{2}\right ] , \\ & & \frac{c_{v}}{nk_{b}}=\frac{\pi ^{2}}{2}\frac{k_{b}t}{\mu ( 0)},\end{aligned}\]]which are familiar results .
those discussions show that there exists similarity even between so different systems
! c. n. yang , introductory note on phase transitions and critical phenomena , in _ phase transitions and critical phenomena _
by c. domb and m. s. green , academic press , london , 1972 ) , vol .
1 , pp . 15 . | in this paper , as a personal review , we suppose a possible extension of gibbs ensemble theory so that it can provide a reasonable description to phase transitions and spontaneous symmetry breaking . the extension is founded on three hypotheses , and can be regarded as a microscopic edition of the landau phenomenological theory of phase transitions . within its framework ,
the stable state of a system is determined by the evolution of order parameter with temperature according to such a principle that the entropy of the system will reach its minimum in this state .
the evolution of order parameter can cause change in representation of the system hamiltonian ; different phases will realize different representations , respectively ; a phase transition amounts to a representation transformation .
physically , it turns out that phase transitions originate from the automatic interference among matter waves as temperature is cooled down .
typical quantum many - body systems are studied with this extended ensemble theory .
we regain the bardeen - cooper - schrieffer solution for the weak - coupling superconductivity , and prove that it is stable .
we find that negative - temperature and laser phases arise from the same mechanism as phase transitions , and that they are instable .
for the ideal bose gas , we demonstrate that it will produce bose - einstein condensation ( bec ) in the thermodynamic limit , which confirms exactly einstein s deep physical insight .
in contrast , there is no bec either within the phonon gas in a black body or within the ideal photon gas in a solid body .
we prove that it is not admissible to quantize dirac field by using bose - einstein statistics .
we show that a structural phase transition belongs physically to the bec happening in configuration space , and that a double - well anharmonic system will undergo a structural phase transition at a finite temperature . for the @xmath0-symmetric vector model ,
we demonstrate that it will yield spontaneous symmetry breaking and produce goldstone bosons ; and if it is coupled with a gauge field , the gauge field will obtain a mass ( higgs mechanism ) .
also , we show that an interacting bose gas is stable only if the interaction is repulsive . for the weak interaction case , we find that the bec is a @xmath1-transition and its transition temperature can be lowered by the repulsive interaction . in connection with liquid @xmath2
, it is found that the specific heat at constant pressure @xmath3 will show a @xmath4 law at low temperatures , which is in agreement with the experiment .
if the system is further cooled down , the theory predicts that @xmath3 will vanish linearly as @xmath5 , which is anticipating experimental verifications . |
the study of the qcd phase diagram is a key issue in modern theoretical and experimental physics of dense matter .
recent heavy - ion collision experiments at rhic brookhaven @xcite have led to the insight that the quark - gluon plasma ( qgp ) at high temperatures behaves as a perfect fluid with a low viscosity to entropy ratio @xmath1 @xcite which is very close to the kss bound @xcite for this number , @xmath2 .
this strong deviation from the behavior of a gas of weakly interacting quarks and gluons is attributed to the occurence of mesonic bound states @xcite or resonances @xcite in the strongly coupled qgp ( sqgp ) .
it has been pointed out @xcite that this situation in hot and dense qcd matter bears similarities with strongly coupled plasmas in other systems where bound state dissociation or mott - anderson delocalization @xcite occurs since the effective coupling strength is modified by electronic screening and/or pauli blocking effects .
it is thus a very general effect expected to occur in a wide variety of dense fermi systems with attractive interactions @xcite .
when this mott transition from truly bound to resonantly paired states occurs under conditions of bose condensation one speaks of a bec - bcs crossover @xcite .
recently , this transition became accessible to laboratory experiments with ultracold gases of fermionic atoms coupled via feshbach resonances with a strength tunable by applying external magnetic fields @xcite .
the bec - bcs crossover transition in quark matter is of particular theoretical interest due to the additional relativistic regime it offers @xcite .
theoretical concepts explaining the appearance of nonperturbative phenomena like quantum condensates and bound states in dense fermi systems with their observable consequences shall apply here but must be formulated within a quantum field theoretic approach .
a systematic treatment of these effects is possible within the path integral formulation for finite - temperature quantum field theories .
this approach is especially suited to take into account the effects of spontaneous symmetry breaking . here
we will apply this approach on the example of a model field theory of the njl type to quark matter as a relativistic strongly interacting fermi system .
it is our aim to delineate relationships between the regions of the njl model phase diagram where @xmath0sb and color superconductivity occur and the possibility to observe quasi - bound pionic states in hic experiments where hot , dense qcd matter produced in the form of rather short - lived fireballs .
as a generic model system for the description of hot , dense fermi - systems with strong , short - range interactions we consider quark matter described by a model lagrangian with four - fermion coupling .
the key quantity is the partition function @xmath3 from which all thermodynamic quantities can be derived . in the imaginary time formalism ( @xmath4 )
it can be expressed as a path integral @xcite @xmath5 where the chemical potential @xmath6 is introduced as a lagrange multiplier for assuring conservation of baryon number .
the notation @xmath7 is shorthand for @xmath8 where @xmath9 is the inverse temperature .
the quark matter is described by a dirac lagrangian with internal degrees of freedom ( @xmath10 flavors , @xmath11 colors ) , with a current - current - type four - fermion interaction inspired by one - gluon exchange .
after fierz transformation of the interaction , we select the scalar diquark channel and the scalar - pseudoscalar meson channels so that our model lagrangian assumes the form @xmath12 ~.\end{aligned}\ ] ] here @xmath13 are the dirac matrices , @xmath14 is a color @xmath15 gell - mann matrix , @xmath16 are @xmath17 flavor matrices and @xmath18 is the charge conjugation matrix . @xmath19 and @xmath20 are coupling strengths corresponding to the different channels , see ref .
@xcite for a recent review . for the numerical analysis
we adopt parameters from ref .
@xcite and consider @xmath20 as a free parameter .
after introduction of the hubbard - stratonovich @xcite auxiliary fields @xmath21 , @xmath22 , @xmath23 , @xmath24 and the nambu - gorkov spinors @xmath25 , @xmath26 with @xmath27 , the partition function becomes a gaussian path integral in the bispinor fields which can be evaluated and yields the fermion determinant @xmath28~ , \label{detglei}\end{aligned}\ ] ] where the inverse bispinor propagator is a matrix in nambu - gorkov- , dirac- , color- and flavor space .
after fourier transformation it reads @xmath29 with @xmath30 .
so far we could derive with ( [ detglei ] ) a very compact , bosonized form of the quark matter partition function ( [ ll ] ) which is an exact transformation of ( [ ll ] ) , now formulated in terms of collective , bosonic fields . as we will demonstrate in the following ,
this form is suitable since it allows to obtain nonperturbative results already in the lowest orders with respect to an expansion around the stationary values of these fields . in performing this expansion , we may factorize the partition function into mean field ( mf ) , gaussian fluctuation ( gauss ) and residual ( res ) contributions @xmath31 in the following we will discuss the physical content of these approximations . in thermodynamical equilibrium ,
the mean field values satisfy the stationarity condition that the thermodynamical potential @xmath32 be minimal , i.e. @xmath33 this is equivalent to the fulfillment of the gap equations @xmath34 , @xmath35 and @xmath36 , together with the stability criterion that the determinant of the curvature matrix formed by the second derivatives is positive . after the evaluation of the traces and the sum over the matsubara frequencies one gets @xmath37~ , \label{thermodynamic - potential - mf}\end{aligned}\ ] ] where we have defined the particle dispersion relation @xmath38 with @xmath39 , @xmath40 .
the @xmath41 dispersion law @xmath42(@xmath43 ) is associated to the red and green quarks ( antiquarks ) , whereas the ungapped blue quarks ( antiquarks ) have the dispersion @xmath44 ( @xmath45 ) . with the aformentioned stationarity conditions applied to eq .
( [ omegamf ] ) we obtain the gap equations for the order parameters @xmath46 and @xmath47 , which have to be solved self - consistently , @xmath48\frac{\xi_{\bf p}^-}{e_{\bf p}^- } \nonumber \\ & & \hspace{1.5 cm } + \left [ 1 - 2n_f(e_{\bf p}^+ ) \right]\frac{\xi_{\bf
p}^+}{e_{\bf p}^+ } + n_f(-\xi_{\bf p}^+ ) - n_f(\xi_{\bf p}^-)\bigg\}~ , \\
\delta & = & 8g_{d } \int\frac{d^3p}{(2\pi)^3}\left [ \frac{1 - 2n_f(e_{\bf p}^-)}{e_{\bf p}^- } + \frac{1 - 2n_f(e_{\bf p}^+)}{e_{\bf p}^+ } \right]~.\end{aligned}\ ] ] the fermi distribution is @xmath49 .
solutions of the gap equations for the dynamically generated quark mass @xmath46 and for the diquark pairing gap @xmath47 at @xmath50 as a function of @xmath6 are shown in the left panel of fig .
[ order ] . from the solutions for the order parameters in dependence of the thermodynamical variables
@xmath51 and @xmath6 we have constructed the phase diagram of the present quark matter model in the @xmath52 plane , see right panel of fig . [ order ] . the two order parameters , that are indicators of phase transitions , allow to distinguish 4 phases : @xmath53 , @xmath54 : normal phase ( nqm ) ; @xmath41 , @xmath54 : color superconductor ( 2sc ) ; @xmath53 , @xmath55 : chiral symmetry broken phase ( @xmath0sb ) ; @xmath41 , @xmath56 : coexistence of @xmath0sb and 2sc phases . increasing the diquark coupling @xmath57 leads to an increase of the diquark gap and therefore a rise in the critical temperature for the second order transition to a normal quark matter phase .
it shifts also the border between color superconductivity ( 2sc ) and chiral symmetry broken phase ( @xmath0sb ) to lower values of the chemical potential . for very strong coupling @xmath58 , a coexistence region developes where
both order parameters are simultaneously nonvanishing .
since the phase border is not of first order , no critical endpoint can be identified in this case . in the @xmath0sb phase pion and diquark
exist as zero width bound states . at the chiral symmetry restoration transition
, they merge the continuum of unbound states and turn into ( resonant ) scattering states .
when this mott transition occurs within the 2sc phase we speak of a bec - bcs crossover .
now we discuss the interesting question of the quasiparticle excitations in these phases . to this end
, we will expand the action functional in the partition function up to quadratic order in the mesonic fields and arrive at a tractable approximation for the bosonized quark matter model .
let us expand now the mesonic fields around their mean field values .
in these lectures , we will focus on fluctuations in the mesonic channels , where the pion and the sigma meson will emerge as quasiparticle degrees of freedom . on the example of the pion we will explain the physics of the mott transition .
the detailed investigation of the quantized diquark fluctuations , which are also a prerequisite of the formation of baryons , will be given elsewhere @xcite .
as mentioned above , the pion mean field vanishes , but the sigma field can be separated into mean field and fluctuations as @xmath59 .
hence it is possible to decompose the inverse propagator @xmath60 into a mean field part and a fluctuation part @xmath61 , so that in the gaussian approximation the fermion determinant becomes @xmath62\big|_{\rm gauss } } { { \rm det } \left[s^{-1}_{mf}\right ] } = \exp\left\{-\frac{1}{2}\int\frac{d^{4}q}{(2\pi)^4 } { \rm tr}\left[s_{mf}(p)\sigma(q ) s_{mf}(p+q)\sigma(q)\right]\right\}~ , \label{expansion - propagator}\end{aligned}\ ] ] where the matrix @xmath63 and the propagator @xmath64 are defined as @xmath65 the matrix elements of the nambu - gorkov propagator are @xmath66 where @xmath67 , @xmath68 .
for the subsequent evaluation of traces in quark - loop diagrams , it is convenient to use this notation with projectors in color space , @xmath69 , @xmath70 and in dirac space , @xmath71\,.$ ] the summation over matsubara frequencies @xmath72 is most systematic using the above decomposition into simple poles in the @xmath73 plane .
the poles of the normal propagators @xmath74 are given by the gapped dispersion relations for the paired red - green quarks ( antiquarks ) , @xmath75 , and the ungapped dispersions @xmath76 for the blue quarks ( antiquarks ) .
the anomalous propagators @xmath77 are only nonvanishing in the 2sc phase where the pair amplitude is nonvanishing .
let us notice explicitly that this procedure has yielded an effective action that includes the fluctuation terms responsible for the excitation of scalar and pseudoscalar mesonic modes .
the evaluation of the traces ( [ expansion - propagator ] ) can be performed with the result @xmath78 where we have introduced the polarization functions @xmath79~ , \\
\pi_{\pi\pi}(q_0,{\bf q } ) & \equiv & -{\rm tr}[{\bf g}^+_p(\gamma_5{\bm\tau}){\bf g}^+_{p+q } ( \gamma_5{\bm \tau } ) + { \bf f}^-_{\bf p}(\gamma_5{\bm\tau}^t ) { \bf f}^+_{p+q}(\gamma_5{\bm \tau } ) \nonumber \\ & & + { \bf f}^+_p(\gamma_5{\bm \tau } ) { \bf f}^-_{p+q}(\gamma_5{\bm \tau}^t ) + { \bf g}^-_p(\gamma_5{\bm \tau}^t ) { \bf g}^-_{p+q}(\gamma_5{\bm \tau}^t)]~,\end{aligned}\ ] ] as the key quantities for the investigation of mesonic bound and scattering states in quark matter . indeed by using bethe salpeter equation and by evaluating the spectral functions one can obtain important information on the mesons properties . in the following we perform the further evaluation and discussion for the pionic modes , the @xmath80 mode is treated in an analogous way .
we start with the evaluation of traces and matsubara summation and obtain @xmath81 for a pionic mode at rest in the medium ( @xmath82 ) this reduces to @xmath83 \nonumber \\ & & \hspace{-5 mm } + \left[1- \frac{\xi_{\bf p}^+\xi_{\bf p}^- + \delta^2 } { e_{\bf p}^+e_{\bf p}^-}\right ] m(e_{\bf p}^+ , e_{\bf p}^- ) \left[\frac{1}{q_0 - e_{\bf p}^+ + e_{\bf p}^- } - \frac{1}{q_0 + e_{\bf p}^+ - e_{\bf p}^-}\right ] \nonumber \\ & & \hspace{-5 mm } + \left[1+\frac{\xi_{\bf p}^+\xi_{\bf p}^- + \delta^2}{e_{\bf p}^+e_{\bf p}^- } \right ] n(e_{\bf p}^+ , e_{\bf p}^- ) \left[\frac{1}{q_0 - e_{\bf p}^+ - e_{\bf p}^- } - \frac{1}{q_0 + e_{\bf p}^+ + e_{\bf p}^-}\right ] \bigg\}~ , \label{mu1}\end{aligned}\ ] ] where we have introduced the phase space occupation factors @xmath84 ( pauli blocking ) and @xmath85 .
we make use of the dirac identity @xmath86 in order to decompose the polarization function into real and imaginary parts after analytical continuation to the complex plane . the imaginary part is straightforwardly integrated after transformation from momentum to energy @xmath87 . at the pole , the integration variables transform as @xmath88 and the integration borders shift @xmath89 , where the thresholds are given by @xmath90 for @xmath53 and @xmath91 otherwise . in this way one can decompose the pion polarization function in the 2sc phase into a real and an imaginary part . from the real part we can calculate the pion mass by solving the bethe - salpeter equation @xmath92 , while a nonvanishing imaginary part corresponds to a finite pion width @xmath93 for decay into quark - antiquark pairs . in fig .
[ pion_mu000 ] , we show the temperature dependence of the masses and widths of pions and sigma - mesons for strong diquark coupling @xmath94 at vanishing chemical potential ( left panel ) and at @xmath95 mev ( right panel ) . at @xmath96 ,
the only threshold for the imaginary parts of meson decays is @xmath90 , since @xmath97 .
the @xmath80 mass is always above the threshold and therefore this state is unstable in the present model .
the pion , however , is a bound state until the critical temperature for the mott transition @xmath98 mev is reached . as can be seen from the slow rise of the decay width @xmath99 ,
the pion is still a well - identifyable , long - lived resonance .
the detailed analytic behavior of the pion at the mott transition has been discussed in the context of the njl model by hfner et al .
@xcite , see also the inset of the left panel of fig .
[ pion_mu000 ] .
it shows strong similarities with the behavior of bound states of fermionic atoms in traps when their coupling is tuned by exploiting feshbach resonances in an external magnetic field , see @xcite . in the context of rhic experiments ,
one has discussed such quasi - bound states as an explanation for the perfect liquid behavior of the sqgp @xcite .
next we want to discuss the pionic excitations in the presence of a diquark condensate in the 2sc phase , see the right panel of fig .
[ pion_mu000 ] .
we observe the remarkable fact that the 2sc condensate stabilizes the pion at @xmath50 as a true bound state , although the pion mass exceeds by far the threshold @xmath90 .
this effect is due to a compensation of gapped and ungapped quark modes and has been discussed before by ebert et al .
@xcite for @xmath50 only .
here we extend this study to the finite temperature case , where the pion obtains a finite width but is still a very good resonance .
finally , an interesting insight can come looking at fig .
[ phase_life ] . in this figure
we have reported the essential phase diagram of the two - flavor quark matter in the case of coupling @xmath100 .
the dashed line represents the chiral crossover line and the solid line refers to the phase transition towards 2sc phase .
the dot - dashed line shows the border above which the pion turns out to be a quasi bound state in the quark plasma .
the dotted line indicates the region border ( a ) below which the quasi - stable pionic states have a lifetime greater than 1 fm / c , which is a typical lifetime of fireballs . ]
thus they would be measured as bound states and could be significant in the framework of hic experiments . region ( b ) confirms our claim about the absence of stable pions in a two - flavor superconductor .
however this claim needs a comment .
first of all , it can be easily understood how the presence of a finite diquark gap ( 2sc phase ) at t=0 stabilizes the pion .
indeed as soon as the quark mass drops due to chiral symmetry restoration , the diquark gap tends to be finite and thus takes over the role of the quark mass in the dispersion relations . roughly speaking the quark start to be dressed by his interactions .
thus the pion does not `` feel '' the drop of the quark mass .
but as @xmath47 melts with increasing @xmath51 , the pion width @xmath101 raises , first slowly and then rapidly , reflecting the behaviour of @xmath47 as a function of t. this leads inevitably to a destabilization of the the pion states in the vicinity of the 2sc phase border .
the discussion of the mesonic modes in the 2sc phase points to a very rich spectrum of excitations which eventually leads to specific new observable signals of this hypothetical phase .
the cbm experiment planned at fair darmstadt and the nica project at jinr dubna could be capable of creating thermodynamical conditions for the observation of these excitations in the experiment
. one promising signal could be the scalar resonance in the pion - pion scattering at the two - pion threshold which is in principle observable , e.g. , in the two - photon decay channel @xcite .
however , the description of this state goes beyond the gaussian approximation to which we restrict ourselves in this work .
in this work we have derived and evaluated the gap equations and the scalar - pseudoscalar meson spectra within a path integral approach to the two - flavor njl type model of superconducting quark matter .
after fixing the parameters of the model to the light meson spectrum in the vacuum , the diquark coupling remains as a free parameter which has been used to extend the model beyond the traditional range of applications into the region of bec - bcs crossover .
we have presented the phase diagram of quark matter at strong and very strong coupling .
the origin of the bec - bcs crossover in superconducting quark matter is the mott transition for diquark bound states .
we explain the physics of the mott transition on the example of mesonic correlations . we have investigated the meson spectra ( bound and scattering states ) outside ( @xmath102 ) and for the first time also inside ( @xmath103 ) the color superconductivity region .
we find the thresholds for the dissociation of pionic bound states into unbound , but resonant scattering states in the quark - antiquark continuum and we have shown that outside the @xmath0sb region where the pion is a zero - width bound state , there are two regions where it can be considered as a quasi - bound state with a lifetime exceeding that of a typical heavy - ion collision fireball : ( a ) the high - temperature @xmath0sb crossover region at low densities and ( b ) the high - density color superconducting phase at temperatures below 100 mev .
d.z . and d.b .
are grateful to the organizers for providing support for their participation at the meeting .
the work of d.b . is supported in part by the polish ministry of national education ( menisw ) .
we acknowledge support from the esf research networking programme `` compstar '' for our participation at the workshop `` the complex physics of compact stars '' in ladek zdrj ( poland ) , where parts of this work have been completed .
9 b. mller and j. l. nagle , _ ann .
nucl . part .
_ , * 56 * , 93 ( 2006 ) e. shuryak , _ prog .
_ , * 53 * , 273 ( 2004 ) e. v. shuryak and i. zahed , _ phys .
c _ , * 70 * , 021901 ( 2004 ) e. v. shuryak , i. zahed , _ phys .
d _ , * 70 * , 054507 ( 2004 ) d. a. teaney , _ j. phys .
g _ , * 30 * , s1247 ( 2004 ) .
g. policastro , d.t . son and a.o .
starinets , _ phys
_ , * 87 * , 081601 ( 2001 ) ; + p. kovtun , d.t . son and a.o .
starinets , _ jhep _ , * 0310 * , 064 ( 2003 ) . d. b. blaschke and k. a. bugaev , _ fizika b _ , * 13 * , 491 ( 2004 ) [ arxiv : nucl - th/0311021 ] .
d. blaschke , _ aip conf .
_ , * 775 * , 151 ( 2005 ) .
h. abuki , _ nucl .
phys . a _ , * 791 * , 117 ( 2007 ) e. v. shuryak , _ arxiv : nucl - th/0606046 _ n. mott , _ rev . mod . phys .
_ , * 40 * , 677 ( 1968 ) a. schnell , g. rpke , p. schuck , _ phys . rev
_ , * 83 * , 1926 ( 1999 ) m. kitazawa , t. koide , t. kunihiro , y. nemoto , _ phys . rev . d _ , * 65 * , 091504 ( 2002 ) m. kitazawa , t. koide , t. kunihiro , y. nemoto , _ phys . rev .
d _ , * 70 * , 056003 ( 2004 ) d. blaschke , d. ebert , k. g. klimenko , m. k. volkov , v. l. yudichev , _ phys
. rev .
d _ , * 70 * , 014006 ( 2004 ) d. blaschke , s. fredriksson , h. grigorian , a. m. ztas , f. sandin , _ phys .
d _ , * 72 * , 065020 ( 2005 ) q. chen , j. stajic , k. levin , _ j. low temp . phys _ , * 32 * , 406 ( 2005 ) e. calzetta , b. l. hu and a. m. rey , _ phys . rev .
, * 73 * , 023610 ( 2006 ) m. w. zwierlein , c. a. stan , c. h. schunck , s. m. raupach , s. gupta , z. hadzibabic and w. ketterle , _ phys . rev . lett . _ ,
* 91 * , 250401 ( 2003 ) m. w. zwierlein , j. r. abo - shaeer , a. schirotzek , c. h. schunck and w. ketterle , _ nature _ , * 435 * , 1047 ( 2003 ) m. greiner , o. mandel , t. rom , a. altmeyer , a. widera , t. w. hnsch and i. bloch , _ physica b _ ,
* 329 * , 11 ( 2003 ) j. deng , a. schmitt , q. wang , _ phys . rev .
d _ , * 76 * , 034013 ( 2007 ) g. sun , l. he , p. zhuang , _ phys .
d _ , * 75 * , 096004 ( 2007 ) _ finite - temperature field theory _ , edited by j. kapusta , cambridge university press , cambridge , 1989 , p.26
. m. buballa , _ phys .
_ , * 407 * , 205 ( 2005 ) h. grigorian , _ phys .
_ , * 4 * , 223 ( 2007 ) h. kleinert , _ fortsch .
_ , * 26 * , 565 ( 1978 ) d. ebert , k. k. klimenko , v. l. yudichev , _ phys
. rev .
c _ , * 72 * , 015201 ( 2005 ) r. anglani , d. blaschke , d. zablocki , _ in preparation_. m. k. volkov , e. a. kuraev , d. blaschke , g. ropke and s. m. schmidt , _ phys
. lett .
b _ , * 424 * , 235 ( 1998 ) d. blaschke , yu . l. kalinovsky , a. e. radzhabov and m. k. volkov , _ phys . part .
_ , * 3 * , 327 ( 2006 ) | the relationship between mesonic correlations and quantum condensates in the quark matter phase diagram is explored within a quantum field theoretical approach of the nambu and jona - lasinio ( njl ) type .
mean - field values in the scalar meson and diquark channels are order parameters signalling the occurrence of quark condensates , entailing chiral symmetry breaking ( @xmath0sb ) and color superconductivity ( 2sc ) in quark matter .
we investigate the spectral properties of scalar and pseudoscalar meson excitations in the phase diagram in gaussian approximation and show that outside the @xmath0sb region where the pion is a zero - width bound state , there are two regions where it can be considered as a quasi - bound state with a lifetime exceeding that of a typical heavy - ion collision fireball : ( a ) the high - temperature @xmath0sb crossover region at low densities and ( b ) the high - density color superconducting phase at temperatures below 100 mev .
address = instytut fizyki teoretycznej , uniwersytet wrocawski , 50 - 204 wrocaw , poland , altaddress = institut fr physik , universitt rostock , d-18051 rostock , germany address = instytut fizyki teoretycznej , uniwersytet wrocawski , 50 - 204 wrocaw , poland , altaddress = bogoliubov laboratory for theoretical physics , jinr , 141980 dubna , russia address = dipartimento di fisica , universit di bari , i-70126 bari , italia , altaddress = i.n.f.n .
, sezione di bari , i-70126 bari , italia + e - mail : [email protected] , [email protected] , [email protected] |
symmetries have played an important role in particle physics . in quantum mechanics a symmetry is associated with a group of transformations under which a lagrangian remains invariant .
symmetries limit the possible terms in a lagrangian and are associated with conservation laws . here
we will be concerned with the role of discrete symmetries : space reflection ( parity ) @xmath5 : @xmath6 , time reversal @xmath7 : @xmath8 and charge conjugation @xmath9 : @xmath10 .
quantum electrodynamics ( qed ) and quantum chromodynamics ( qcd ) respect all these symmetries .
also , all lorentz invariant local quantum field theories are @xmath11 invariant . however , in weak interactions @xmath9 and @xmath5 are maximally violated separately but as we will see below , @xmath0 is conserved
. first indication of parity violation was revealed in the decay of a particle with spin parity @xmath12 called @xmath13-meson into two modes @xmath14 ( parity violating ) , and @xmath15 @xmath16(parity conserving ) .
lee and yang in 1956 , suggested that there is no experimental evidence for parity conservation in weak interaction .
they suggested number of experiments to test the validity of space reflection invariance in weak decays .
one way to test this is to measure the helicity of outgoing muon in the decay : @xmath17 the helicity of muon comes out to be negative , showing that parity conservation does not hold in this decay . in the rest frame of the pion ,
since @xmath18 comes out with negative helicity , the neutrino must also come out with negative helicity because of the spin conservation . thus confirming the fact that neutrino is left handed .
@xmath19 under charge conjugation , @xmath20 helicity @xmath21 under @xmath9 and @xmath5 transforms as , @xmath22 invariance under @xmath9 gives , @xmath23 experimentally , @xmath24 showing that @xmath9 is also violated in weak interactions .
however , under @xmath0 , @xmath25 which is seen experimentally .
thus , @xmath0 conservation holds in weak interaction . in the standard model
, the fermions for each generation in their left handed chirality state belong to the representation , @xmath26 of the electroweak unification group @xmath27 .
hence , the weak interaction lagrangian for the charged current in the standard model is given by , @xmath28 where @xmath29 is any of the left - handed doublet ( @xmath30 is the generation index ) .
we note that the weak eigenstates @xmath31 and @xmath32 are not equal to the mass eigenstates @xmath33 and @xmath34 .
they are related to each other by a unitarity transformation , @xmath35 where @xmath36 is called the @xmath37 matrix .
@xmath38 @xmath39 the unitarity of @xmath36 , @xmath40 gives,@xmath41 $ ] @xmath42 the second line in equation ( [ 06 ] ) expresses @xmath36 in terms of wolfenstien parametrization .
thus , @xmath43 where , @xmath44 in order to show that @xmath45 is @xmath0-invariant , we first note that under @xmath9 , @xmath5 and @xmath7 operations the dirac spinor @xmath46 transforms as follows : @xmath47 the effect of transformations @xmath9 , @xmath5 and @xmath0 on various quantities that appear in a gauge theory lagrangian are given below :
@xmath48 the vector bosons associated with the electroweak unification group @xmath49 transform under @xmath0 as : @xmath50 where , @xmath51 the lagrangian transforms as : @xmath52 thus , the weak interaction lagrangian in the standard model violates @xmath9 and @xmath5 but is @xmath0-invariant .
it is instructive to discuss the restrictions imposed by @xmath11 invariance .
@xmath53 invariance implies , @xmath54 hence , we get : @xmath55 in deriving the above result , we have put @xmath56 where @xmath57 means that momenta and spin are reversed .
since we are in the rest frame of @xmath58 , @xmath7 will reverse only magnetic quantum number and we can drop @xmath57 .
further we have used , @xmath59 @xmath60 @xmath61 where @xmath62 is the strong interaction phase shift .
if @xmath0-invariance holds , then , @xmath63 thus , the necessary condition for @xmath0-violation is that the decay amplitude @xmath64 should be complex . in view of our discussion above , under @xmath0
an operator @xmath65 is replaced by , @xmath66 the effective lagrangian has the structure ( @xmath67 ) , @xmath68 hence , @xmath0-violation requires @xmath69 .
we now discuss the implication of @xmath11 constraint with respect to @xmath0 violation of weak decays .
the weak amplitude is complex ; it contains the final state strong phase @xmath62 and in addition it may also contain a weak phase @xmath70 . taking out both these phases , @xmath71 @xmath11 ( eq . ) gives , @xmath72 we conclude that the weak interaction lagrangian in the standard model is @xmath73 invariant and since @xmath0 violation has been observed in hadronic sector ( only in @xmath74 and @xmath13 decays ) and not in leptonic sector , it is a consequence of mismatch between weak and mass eigenstates ( i.e. the phases in @xmath37 matrix ) and/or the mismatch between @xmath0-eigenstates , @xmath75 ; \text { } cp\left| x_{1,2}^{0}\right\rangle = \pm \left| x_{1,2}^{0}\right\rangle \label{11}\ ] ] and the mass eigenstates i.e. @xmath0-violation in the mass matrix .
@xmath0- violation due to mass mixing and in the decay amplitude has been experimentally observed in @xmath76 and @xmath77 . for @xmath78 decays , the @xmath73-violation in the mass matrix
is not expected in the standard model .
in fact time dependent @xmath0-violation asymmetry gives a clear way to observe direct @xmath0-violation in @xmath79 and @xmath78 decays . if @xmath0 is conserved , @xmath80 then , @xmath81 thus @xmath82 and @xmath83 are also mass eigenstates .
they form a complete set ( in units @xmath84 ) , @xmath85 the solution is , @xmath86 suppose we start with the state @xmath87 , i.e. , @xmath88 then we get , @xmath89 \notag \\ & = & \frac{1}{\sqrt{2}}\left\ { \left [ \exp \left ( -im_{1}t-\frac{1}{2}\gamma _ { 1}t\right ) \right . \right .
\notag \\ & & \left .
+ \exp \left ( -im_{2}t-\frac{1}{2}\gamma _ { 2}t\right ) \right ] \left| x^{0}\right\rangle \notag \\ & & -\left [ \exp \left ( -im_{1}t-\frac{1}{2}\gamma _ { 1}t\right ) \right .
\notag \\ & & \left .
-\exp \left ( -im_{2}t-\frac{1}{2}\gamma _ { 2}t\right ) \right ] \left| \bar{x}^{0}\right\rangle \right\ } \label{2.2}\end{aligned}\ ] ] however , in @xmath90 basis , @xmath91 the mass matrix @xmath92 is not diagonal and is given by , @xmath93 hermiticity of matrices @xmath94 and @xmath95 gives ( @xmath96 ) , @xmath97 @xmath11 invariance gives , @xmath98 @xmath99 @xmath100 diagonalization of mass matrix @xmath92 in eq .
( [ 2.3 ] ) gives , @xmath101 where , @xmath102 the eigenstates are given by , @xmath103\ ] ]
consider the process , @xmath105 thus , weak interaction can mix @xmath76 and @xmath106 , @xmath107 off diagonal matrix elements are not zero .
thus , @xmath76 and @xmath106 can not be mass eigenstates .
select the phase : @xmath108 define , @xmath109 \\ \left| k_{2}^{0}\right\rangle & = & \frac{1}{\sqrt{2}}\left [ \left| k^{0}\right\rangle + \left|
\bar{k}^{0}\right\rangle \right]\end{aligned}\ ] ] choose : @xmath110 where @xmath111 and @xmath112 are eigenstates of @xmath0 with eigenvalues @xmath113 and @xmath114 .
assuming @xmath0 conservation , @xmath115 @xmath116 where @xmath117 and @xmath118 are real .
thus , @xmath119 since , @xmath120 therefore , it is clear that , @xmath121 is allowed by @xmath0 conservation .
however , experimentally it was found that long lived @xmath112 also decay to @xmath122 but with very small probability .
small @xmath0 non conservation can be taken into account by defining , @xmath123 where @xmath124 is a small number .
thus @xmath0 non conservation manifests itself by the ratio : @xmath125 now @xmath0 non conservation implies , @xmath126 since @xmath0 violation is a small effect , therefore , @xmath127 further , if @xmath0- violation arises from mass matrix , then , @xmath128 thus , @xmath0violation can result by a small term @xmath129 in the mass matrix given in eq .
( [ 2.1 ] ) , @xmath130 diagonalization gives , @xmath131 then from eq .
( [ 2.10 ] ) up to first order , we get , @xmath132 eq . (
[ 2.2 ] ) is unchanged , replace , @xmath133 @xmath134 now , @xmath135 hence from eq . ( [ 2.2 ] ) , @xmath136 \left| k^0\right\rangle \\ -\left [ e^{\frac{-1}2\gamma _ st}e^{\frac
i2\delta mt}-e^{-\frac i2\delta mt } % \right ] \left| \bar{k}^0\right\rangle% \end{array } \right\ } \label{2.20}\ ] ] therefore , probability of finding @xmath137 at time @xmath138 ( recall that we started with @xmath139 ) , @xmath140 if kaons were stable @xmath141 , then , @xmath142 \label{2.22}\ ] ] which shows that a state produced as pure @xmath143 state at @xmath144 continuously oscillates between @xmath143 and @xmath145 state with frequency @xmath146 and period of oscillation , @xmath147 kaons , however , decay and their oscillations are damped . by measuring the period of oscillation , @xmath148 can be determined .
@xmath149 such a small number is measured as a consequence of superposition principle in quantum mechanics , @xmath150 @xmath151 can only be produced by @xmath106 in the final state .
this would give a clear indication of oscillation .
coming back to @xmath0-violation , @xmath152 so far we have considered @xmath0-violation due to mixing in the mass matrix .
it is important to detect the @xmath0-violation in the decay amplitude if any .
this is done by looking for a difference between @xmath0-violation for the final @xmath153 state and that for @xmath154 .
now due to bose statistics , the two pions can be either in @xmath155 or @xmath156 states .
using clebsch - gordon ( cg ) coefficients , @xmath157 \notag \\
a\left ( k^{0}\rightarrow \pi ^{0}\pi ^{0}\right ) & = & \frac{1}{\sqrt{3}}\left [ a_{0}e^{i\delta _ { 0}}-\sqrt{2}a_{2}e^{i\delta _ { 2}}\right ] \label{2.27}\end{aligned}\ ] ] now @xmath11-invariance gives , @xmath158 \notag \\
a\left ( \bar{k}^{0}\rightarrow \pi ^{0}\pi ^{0}\right ) & = & \frac{1}{\sqrt{3 } } % \left [ a_{0}^{\ast } e^{i\delta _ { 0}}-\sqrt{2}a_{2}^{\ast } e^{i\delta _ { 2 } } % \right ] \label{2.28}\end{aligned}\ ] ] the dominant decay amplitude is @xmath159 due to @xmath160 rule , @xmath161 . using the wu yang phase convention
, we can take @xmath159 to be real .
neglecting terms of order @xmath162 and @xmath163 , we get , @xmath164 where , @xmath165 clearly @xmath166 measures the @xmath0-violation in the decay amplitude , since @xmath0-invariance implies @xmath167 to be real .
after @xmath168 years of experiments at fermilab and cern , results have converged on a definitive non - zero result for @xmath166 , @xmath169 this is an evidence that although @xmath166 is a very small , but @xmath0-violation does occur in the decay amplitude .
further we note from eq .
( [ 2.30 ] ) , @xmath170 where numerical value is based on an analysis of @xmath171 scattering .
we now discuss the cp - asymmetry in leptonic decays of kaon .
a(k_{l}^{0 } & \rightarrow & \pi ^{+}+l^{-}+\overline{\nu } _
{ l})=\frac{1}{\sqrt{% 2}}[(1+\epsilon ) g^{*}+(1-\epsilon ) f*]\end{aligned}\ ] ] the cp - asymmetry parameter @xmath174@xmath175}{% \left| f\right| ^{2}+\left| g\right| ^{2}+(fg^{*}+f^{*}g)+o(\epsilon ^{2})}\end{aligned}\ ] ] in the standard model @xmath176 transitions are not allowed , thus @xmath177 .
hence @xmath178\ ] ] from eq .
( [ 2.226 ] ) , we get @xmath179 which gives on using expermintal values for @xmath180 and @xmath181@xmath182 in agreement with the expermimental value for @xmath183 finally we discuss cp - asymmetries for @xmath184 decays .
the decays @xmath185 are partiy conserving decays i.e. the parity of the final state is @xmath114 . now the c - partiy of @xmath16 and ( @xmath186 are given by @xmath187 and g - parity of pion is @xmath188 thus @xmath189 hence cp - conservation implies @xmath190 now g - partiy of three pions @xmath191@xmath192 only @xmath193 decays are favored as the decays for @xmath194 are highly suppressed due to centrifugal barrier . hence @xmath195 is highly suppressed .
thus we have to take into account @xmath196 amplitudes viz @xmath197 and @xmath198 .
@xmath199 contribution is expected to be suppressed as it requires @xmath200 transition .
hence cp - asymmetries of @xmath201 decays are given by @xmath202}{\text{re}a_{1}+i\epsilon \text{im}a_{1 } } \\ & \approx & \epsilon + i\frac{\text{im } a_{1}}{\text{re } a_{1 } } \\
\eta _ { + -0 } & = & \frac{a(k_{s}\rightarrow \pi ^{+}\pi ^{-}\pi ^{0})}{% a(k_{l}\rightarrow \pi ^{+}\pi ^{-}\pi ^{0})}\approx \epsilon + i\frac{\text{% i m } a_{1}}{\text{re}a_{1}}=\eta _ { 000}\end{aligned}\ ] ]
for @xmath204 ( q = d or s ) we show below that both @xmath117 and @xmath205 have the same phase .
thus , @xmath206 \simeq \left| m_{12}\right| e^{-2i\phi _ { m } } \notag \\
q^{2 } & = & e^{+2i\phi _ { m}}\left [ \left| m_{12}\right| -i\left| \gamma _ { 12}\right| \right ] \simeq \left| m_{12}\right| e^{2i\phi _ { m } } \label{3.2 } \\ 2pq & = & 2\left| m_{12}\right| = ( m_{2}-m_{1})-\frac{i}{2}\left ( \gamma _ { 2}-\gamma _ { 1}\right ) \notag \\ \rightarrow \delta m_{b}&=&\frac{\left ( m_{2}-m_{1}\right ) } { 2}=\left| m_{12}\right| \label{3.3 } \\
\delta \gamma & = & \gamma _ { 2}-\gamma _ { 1}=0 \notag\end{aligned}\ ] ] the above equations follow from the fact that , @xmath207 @xmath208 induces particle - antiparticle transition . for @xmath209 @xmath208
arises from the box diagram as shown in fig .
2 , where the dominant contribution comes out from the @xmath210quark .
thus , @xmath211 now , @xmath212 where the sum is over all the final states which contribute to both @xmath213 and @xmath214 decays .
thus , @xmath215 hence we have the result that , @xmath216 now @xmath217 transition : @xmath218 e^{2i\beta } \ ] ] hence , @xmath219 on the other hand , @xmath220 transition : @xmath221 @xmath222 @xmath223 also we have , @xmath224 } \xi\ ] ] where @xmath225 is @xmath226 breaking parameter . hence the mass eigenstates @xmath227 and @xmath228 can be written as : @xmath229 \quad cp=+1 , \phi_{m}\rightarrow 0 \label{3.4 } \\
\left| b_{h}^{0}\right\rangle & = & \frac{1}{\sqrt{2}}\left [ \left| \bar{b } ^{0}\right\rangle + e^{2i\phi _ { m}}\left| b^{0}\right\rangle \right ] \quad cp=-1 , \phi_{m}\rightarrow 0 \label{3.5}\end{aligned}\ ] ] in this case , @xmath0 violation occurs due to phase factor @xmath230 in the mass matrix .
now one gets ( from eq .
( [ 2.2 ] ) ) , using eqs.([3.3 ] ) , ( [ 3.4 ] ) and ( [ 3.5 ] ) , @xmath231 similarly we get , @xmath232 suppose we start with @xmath233 viz @xmath234 the probabilities of finding @xmath235 and @xmath236 at time @xmath138 is given by , @xmath237 these are equations of a damped harmonic oscillator , the angular frequency of which is , @xmath238 we define the mixing parameter , @xmath239 experimentally , for @xmath77 and @xmath240 , @xmath241 from eq .
( [ 3.8 ] ) and ( [ 3.9 ] ) , the decay amplitudes for , @xmath242 are given by , @xmath243 from eqs.([3.11 ] ) and ( [ 3.12 ] ) , we get for the decay rates , @xmath244 \label{i } \\
\bar{\gamma}_{\bar{f}}(t ) & = & e^{-\gamma t}\left [ \begin{array}{c } \frac{1}{2}\left ( \left| a_{\bar{f}}\right| ^{2}+\left| \bar{a}_{\bar{f } } \right| ^{2}\right ) -\frac{1}{2}\left
( \left| a_{\bar{f}}\right| ^{2}-\left| \bar{a}_{\bar{f}}\right| ^{2}\right ) \cos \delta mt
\\ + \frac{i}{2}\left ( 2i\text{im}e^{2i\phi _ { m}}a_{\bar{f}}^{\ast } \bar{a}_{%
\bar{f}}\right ) \sin \delta mt% \end{array } \right ] \label{ii}\end{aligned}\ ] ] for @xmath245 and @xmath246 change @xmath247 and @xmath248 in @xmath249 and @xmath250 respectively . as a simple application of the above equations , consider the semi - leptonic decays of @xmath233 , @xmath251 in the standard model , @xmath235 decay into @xmath252 and @xmath233 decay into @xmath253
is forbidden .
thus , @xmath254 hence , @xmath255 non zero value of @xmath256 would indicate mixing .
if , however , @xmath257 and @xmath258 due to some exotic mechanism , then @xmath259 even without mixing .
thus @xmath260 which gives , @xmath261
* case - i : * @xmath262 for this case we get , from eqs .
( [ 3.11 ] ) and ( [ 3.12 ] ) , @xmath263 where , @xmath264 this is the direct @xmath0 violation and , @xmath265 is the mixing induced @xmath0-violation .
if the decay proceeds through a single diagram ( for example tree graph ) , @xmath266 is given by ( see eqs .
( 15 ) and ( 16 ) ) , @xmath267 where @xmath268 is the weak phase in the decay amplitude . hence from eq .
( 3.13 ) , we obtain , @xmath269 in particular for the decay , @xmath270 we obtain , @xmath271 and , @xmath272 dt}{\int_{0}^{\infty } % \left [ \gamma _ { f}\left ( t\right ) + \bar{\gamma}_{f}\left ( t\right ) \right ] dt } \notag \\ \mathcal{a}_{\psi k_{s } } & = & -\sin \left ( 2\beta \right ) \,\,\frac{\left ( \delta m/\gamma \right ) } { 1+\left ( \delta m/\gamma \right ) ^{2 } } \label{3.16 } \\ \text{experiment } & : & \left ( \frac{\delta m}{\gamma } \right ) _ { b_{d}^{0}}=0.776\pm 0.008 \label{3.17}\end{aligned}\ ] ] @xmath273 has been experimentally measured .
it gives , @xmath274 corresponding to the decay @xmath275 , we have the decay @xmath276 thus for this decay @xmath277 in the standard model , @xmath278
@xmath279 this is an example of @xmath0-violation in the mass matrix .
we now discuss the direct @xmath0-violation .
direct @xmath0-violation in @xmath79 decays involves the weak phase in the decay amplitude .
the reason for this being that necessary condition for direct @xmath0 -violation is that decay amplitude should be complex as discussed in section 1 . but this is not sufficient because in the limit of no final state interactions , the direct @xmath0-violation in @xmath280 , @xmath281 decay vanishes . to illustrate this point
, we discuss the decays @xmath282 . the main contribution to this decay
is from tree graph ( see fig .
3 ) ; but this decay can also proceed via the penguin diagram ( see fig .
the contribution of penguin diagram can be written as @xmath283 where @xmath284 , @xmath285 and @xmath286 denote the contributions of @xmath287 , @xmath288 and @xmath138 quarks in the loop . now using the unitarity equation ( [ 07 ] )
, we can rewrite eq .
( [ 4.1 ] ) as , @xmath289 due to loop integration @xmath5 is suppressed relative to @xmath7 but still its contribution is not negligible .
the first part of eq .
( [ 4.2 ] ) has the same ckm matrix elements as for the tree graph , so we can absorb it in the tree graph .
hence we can write ( with @xmath290 ) , @xmath291 where @xmath292 and @xmath293 are strong interaction phases which have been taken out so that @xmath7 and @xmath5 are real .
@xmath268 is the weak phase in penguin graph .
@xmath11 invariance gives , @xmath294 hence direct @xmath0violation asymmetry is given by , @xmath295 where ( @xmath296 = ckm factor ) , @xmath297 for the time dependent @xmath0-asymmetry for @xmath298 decay we obtain from eqs .
( [ 3.13 ] ) and ( [ 4.3 ] ) , @xmath299 where the direct cp
violation parameter @xmath300 and the mixing induced parameter @xmath301 are given by , @xmath302}{1+|\lambda |^{2}}=- \frac{\sin \left ( 2\beta + 2\gamma \right ) + 2r\cos \delta \sin \left ( 2\beta + \gamma -\phi \right ) + r^{2}\sin ( 2\beta -2\phi ) } { 1 + 2r\cos \delta \cos ( \gamma + \phi ) + r^{2 } } \label{4.7b}\ ] ] @xmath303
as discussed above , we have two choices in selecting the penguin contribution . for the first choice , @xmath304 @xmath305 for this case we have , @xmath306 for the second choice , @xmath307
@xmath308 so that in this case we get , @xmath309 for @xmath310 , @xmath311 , the decay amplitudes are given by , @xmath312 \\
a_{+0 } & = & a(b^{+}\rightarrow \pi ^{+}\pi ^{0})=\frac{1}{\sqrt{2}}% te^{i\delta_{t}}e^{i\gamma } \left[1+r_{c}e^{i\delta_{ct}}\right ] \\
r_{c } & = & \frac{c}{t } , \qquad \delta_{ct}=\delta_{c } - \delta_{t}\end{aligned}\ ] ] hence for @xmath313 , the @xmath0asymmetries are given by @xmath314 for the case @xmath315 , we get , @xmath316 we end this section by considering the decays @xmath317 these decays satisfy the relations @xmath318 \\ & = & \left [ \begin{array}{c } \frac{1}{2}\left ( c - p+p_{ew}\right ) -\frac{1}{2}\left ( c+p+\frac{1}{3 } p_{ew}\right ) \\
-(-p+\frac{1}{3}p_{ew})% \end{array } \right ] = 0\end{aligned}\ ] ] where @xmath319 and @xmath320 are color suppressed , penguin and electroweak penguin amplitudes for these decay . from the above equation ,
we obtain , @xmath321 where we have neglected the terms of the order @xmath322 . the parameter @xmath323 is defined below , @xmath324 \\ r & = & \frac{\left| c\right| } { \left| p\right| } \lambda ^{2}r_{b } \\ \frac{\left| c\right| } { \left| p\right| } & \sim & \frac{a_{2}}{a_{4}}\end{aligned}\ ] ] assuming factorization for the electroweak penguin
, we get from the above equation an interesting sum rule , @xmath325 assuming @xmath326 gev@xmath327 , we get from the relation , @xmath328 which is reminiscent of current algebra and spectral function sum rules of 1960 s .
the above sum rule is very well satisfied by the experimental values , @xmath329 it is convenient to write , from eqs.([i ] ) and ( [ ii ] ) , the decay rates in the following form , @xmath330 + \left [ \gamma_{% \bar{f}}-\bar{\gamma}_{f}(t ) \right ] \notag \\ & = & e^{-\gamma t}\left\ { \cos \delta mt\left [ \left ( \left| a_{f}\right| ^{2}-\left| \bar{a}_{\bar{f}}\right| ^{2}\right ) + \left ( \left| a_{\bar{f } } \right| ^{2}-\left| \bar{a}_{f}\right| ^{2}\right ) \right ] \right .
\notag \\ & & \left .
+ 2\sin \delta mt\left [ \text{im}\left ( e^{2i\phi _ { m}}a_{f}^{\ast } \bar{a}_{f}\right ) + \text{im}\left ( e^{2i\phi _ { m}}a_{\bar{f}}^{\ast } \bar{a } _ { \bar{f}}\right ) \right ] \right\ } \notag \\ & & \label{e1 } \\ & & \left[\gamma_{f}(t)+\bar{\gamma}_{\bar{f}}(t ) \right ] - \left[\gamma_{\bar{% f}}(t)+\bar{\gamma}_{f}(t ) \right ] \notag \\ & = & e^{-\gamma t}\left\
{ \cos \delta mt\left [ \left ( \left| a_{f}\right| ^{2}+\left| \bar{a}_{\bar{f}}\right| ^{2}\right ) -\left ( \left| a_{\bar{f } } \right| ^{2}+\left| \bar{a}_{f}\right| ^{2}\right ) \right ] \right .
\notag \\ & & \left .
+ 2\sin \delta mt\left [ \text{im}\left ( e^{2i\phi _ { m}}a_{f}^{\ast } \bar{a}_{f}\right ) -\text{im}\left ( e^{2i\phi _ { m}}a_{\bar{f}}^{\ast } \bar{a } _ { \bar{f}}\right ) \right ] \right\ } \notag \\ & & \label{e2}\end{aligned}\ ] ] we now use the above equations to obtain some interesting results for the @xmath331 asymmetries for b - decays . * case - ii : * we first consider the case in which single weak amplitudes @xmath332 and @xmath333 with different weak phases describe the decays : @xmath334 @xmath11 gives , @xmath335 note @xmath62 and @xmath336 are strong phases ; @xmath337 and @xmath338 are weak phases .
the states @xmath339 and @xmath340 are c - conjugate of each other such as states @xmath341 @xmath342 @xmath343 hence , we get from eqs.([e1 ] ) , ( [ e2 ] ) , ( [ e3 ] ) and ( [ e4 ] ) , @xmath344+[\gamma _ { \bar{f}}(t)-\bar{\gamma}_{f}(t)]}{[\gamma _ { f}(t)+\bar{% \gamma}_{\bar{f}}(t)]+[\gamma _ { \bar{f}}(t)+\bar{\gamma}_{f } ] } \notag \\ & = & \frac{2\bigl|f_{f}\bigr| \bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|}{% \bigl|f_{f}\bigr|^{2}+\bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|^{2}}\sin \delta mt\sin \bigl(2\phi _ { m}-\phi -\phi ^{^{\prime } } \bigr ) \cos
\bigl(% \delta _ { f}-\overset{^{\prime } } { \delta } _ { \bar{f}}\bigr )
\label{e6 } \\ \mathcal{f}\left ( t\right ) & \equiv & \frac{\left [ \gamma _ { f}(t)+\bar{\gamma}% _ { \bar{f}}\right ] -\left [ \gamma _ { \bar{f}}(t)+\bar{\gamma}_{f}\right ] } { % \left [ \gamma _ { f}(t)+\bar{\gamma}_{\bar{f}}\right ] + \left [ \gamma _ { \bar{f}% } ( t)+\bar{\gamma}_{{f}}\right ] } \notag \\ & = & \frac{\bigl|f_{f}\bigr|^{2}-\bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|% ^{2}}{\bigl|f_{f}\bigr|^{2}+\bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|^{2}% } \cos \delta mt \notag \\ & -&\frac{2\bigl|f_{f}\bigr| \bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|}{% \bigl|f_{f}\bigr|^{2}+\bigl|\overset{^{\prime } } { f}_{\bar{f}}\bigr|^{2}}\sin \delta mt\cos \left ( 2\phi _ { m}-\phi -\phi ^{^{\prime } } \right ) \sin \bigl(% \delta _ { f}-\overset{^{\prime } } { \delta } _ { \bar{f}}\bigr ) \label{e7}\end{aligned}\ ] ] we now apply the above formula to @xmath345 and @xmath346 decays . for these decays ,
@xmath347@xmath348@xmath349 note that the effective lagrangians for decays @xmath350 are given by , @xmath351 \left [ \bar{c}\gamma _ { \mu } \left ( 1-\gamma _ { 5}\right ) b\right ] \label{e8a } \\ & & \mathcal{l_{w}}^{\prime}=v_{ub}v_{cq}^{\ast } \left [ \bar{q}\gamma ^{\mu } \left ( 1-\gamma _ { 5}\right ) c\right ] \left [ \bar{u}\gamma _ { \mu } \left ( 1-\gamma _ { 5}\right ) b\right ] \label{e8b}\end{aligned}\ ] ] respectively . in the wolfenstein parametrization of @xmath37 matrix ,
@xmath352 define , @xmath353 thus , we get from eqs . @xmath354 and @xmath355 for @xmath233 decays , ( replacing @xmath356 by @xmath323 ) , @xmath357 for the decays , @xmath358 we get ,
@xmath359 we note that for time integrated @xmath0-asymmetry , @xmath360 dt}{\int_{0}^{\infty } % \left [ \gamma _ { fs}\left ( t\right ) + \bar{\gamma}_{fs}\left ( t\right ) \right ] dt } \notag \\ & = & -\frac{2r}{1+r^{2}}\sin \left ( 2\beta _ { s } + \gamma \right ) \frac{\delta m_{b_{s}}/\gamma _ { s}}{1+\left ( \delta m_{b_{s}}/\gamma _ { s}\right ) ^{2}}% \cos ( \delta _ { f_{s}}-\overset{^{\prime } } { \delta } _ { \bar{f}_{s } } ) \label{4.35a}\end{aligned}\ ] ] the @xmath0asymmetry @xmath361 or @xmath362 involves two experimentally unknown parameters @xmath363 and @xmath364 .
both these parameters are of importance in order to test the unitarity of @xmath37 matrix viz whether @xmath37 matrix is a sole source of @xmath0violation in the processes in which @xmath0violation has been observed . within the * case ii * , we discuss @xmath79 decays into baryons and antibaryons .
so far we have discussed the @xmath0-violation in kaon and @xmath365 systems .
there is thus a need to study @xmath0-violation outside these systems .
the decays of @xmath366 mesons to baryon - antibaryon pair @xmath367 @xmath368 @xmath369 @xmath370 and subsequent decays of @xmath371 or @xmath372 to a lighter hyperon ( antihyperon ) plus a meson provide a means to study @xmath0-odd observables as for example in the process , @xmath373 the decay @xmath374 is described by the matrix element , @xmath375 \label{q1}\]]where as @xmath376 is described by the matrix elements @xmath377\ ] ] where @xmath378 is a constant containing ckm factor , @xmath268 is the weak phase .
the amplitude @xmath332 and @xmath379 are in general complex in the sense that they incorporate the final state phases @xmath380 and @xmath381 and they may also contain weak phases @xmath382 and @xmath383 note that @xmath332 is the parity violating amplitude ( @xmath384-wave ) whereas @xmath379 is parity conserving amplitude ( @xmath385-wave ) .
the @xmath11 invariance gives the matrix elements for the decay @xmath386
@xmath387 \label{q2}\ ] ] if the decays are described by a single matrix element @xmath388 .
if @xmath389 then @xmath11 and @xmath0 invariance give the same predictions viz @xmath390 the decay width for the mode @xmath374 is given by , @xmath391 \label{q4}\end{aligned}\ ] ] in order to take into account the polarization of @xmath367 and @xmath392 we give the general expression for @xmath393 , @xmath394 \label{q5}\end{aligned}\ ] ] where @xmath395 are polarization vectors of @xmath367 and @xmath396 respectively @xmath397 .
in the rest frame of @xmath79 , we get , @xmath398 \notag \\ & & \left\ { \begin{array}{c } 1+\alpha _ { f}\left ( \frac{m_{1}}{e_{1}}\mathbf{n}\cdot \mathbf{s}_{1}-\frac{% m_{2}}{e_{2}}\mathbf{n}\cdot \mathbf{s}_{2}\right )
\\ + \beta _ { f}\mathbf{n}\cdot ( \mathbf{s}_{1}\times \mathbf{s}_{2})+\gamma _ { f}% \left [ ( \mathbf{n}\cdot \mathbf{s}_{1})(\mathbf{n}\cdot \mathbf{s}_{2})-% \mathbf{s}_{1}\cdot \mathbf{s}_{2}\right ] \\ -\frac{m_{1}m_{2}}{e_{1}e_{2}}(\mathbf{n}\cdot \mathbf{s}_{1})(\mathbf{n}% \cdot \mathbf{s}_{2})% \end{array}% \right\ } \label{q6}\end{aligned}\ ] ] where , @xmath399 in the rest frame of @xmath79 , due to spin conservation , @xmath400 thus , invariants multiplying @xmath401 and @xmath402 vanish .
hence we have , @xmath403 \label{q10 } \\
\gamma _ { f } & = & \gamma _
{ f}^{++}+\gamma _ { f}^{--}=\frac{2e_{1}e_{2}}{2\pi m_{b}^{2}}\left| \vec{p}\right| f_{q}^{2}\left [ s_{f}^{2}+p_{f}^{2}\right ] = \bar{\gamma}_{\bar{f } } \label{q11 } \\
\delta \gamma _ { f } & = & \frac{\gamma
_ { f}^{++}-\gamma _ { f}^{--}}{\gamma _ { f}^{++}+\gamma _ { f}^{--}}=\alpha _ { f } , \qquad \delta \bar{\gamma}_{\bar{f } } = \bar{\alpha}_{\bar{f}}=-\alpha _ { f } \label{q12}\end{aligned}\ ] ] eqs . ( [ q11 ] ) and ( [ q12 ] ) follow from @xmath0 invariance .
it will be of interest to test these equations .
now @xmath404 @xmath214 annihilate into baryon - antibaryon pair @xmath405 through @xmath406-exchange as depicted in figs ( 5a ) and ( 5b ) . @xmath407 through annihilation diagram is shown in fig ( 6 ) .
it is clear from fig ( 5a ) and ( 5b ) , that we have the same final state configuration for @xmath404 @xmath408 thus , one would expect , @xmath409 hence we have , @xmath410 above predictions can be tested in future experiments on baryon decay modes of @xmath79-mesons . in particular @xmath411 would give direct confirmation of eqs .
( [ 24 ] ) . for the time
dependent baryon decay modes of @xmath412 , we have : @xmath413 @xmath414where we have used eqs .
( [ 26 ] ) . for @xmath415 @xmath416 [ 4 ] , @xmath417 for @xmath418 @xmath419 , @xmath420 .
eq.([46 ] ) gives a means to determine the weak phase @xmath421 or @xmath422 in the baryon decay modes of @xmath77 and @xmath240 respectively .
non - zero @xmath423 term in @xmath424 would give clear indication of @xmath0 violation especially for baryon decay modes of @xmath415 for which @xmath425 so that @xmath426 .
it may be noted that the time - dependent asymmetries arises because there are two independent amplitudes for the decays @xmath427 @xmath428 the baryon decay modes of @xmath79-mesons not only provide a means to test prediction of @xmath0 asymmetry viz @xmath429 for charmed baryons ( discussed above ) but also to test the @xmath0-asymmetry in hyperon ( antihyperon ) decays viz absence of @xmath0-odd observables @xmath430 discussed in [ 8 ] .
consider for example the decays , @xmath431@xmath432 by analyzing the final state @xmath433 one may test @xmath434 for the charmed hyperon .
we note that for @xmath435 @xmath436 m , whereas @xmath437 cm for @xmath438hyperon , so that the decays of @xmath439 and @xmath440 would not interfere with each other . by analysing the final state @xmath441 and
@xmath442 one may check @xmath0violation for hyperon decays .
one may also note that for @xmath443 complex , the competing channels viz @xmath444 @xmath445 are doubly cabibbo suppressed by @xmath446 unlike @xmath447 complex where the competing channels are suppressed by a factor of @xmath448 .
hence @xmath449 @xmath450 decays are more suitable for this type of analysis .
other decays of interest are , @xmath451 the non - leptonic hyperon ( antihyperon ) decays @xmath452 are related to each other by @xmath11 , @xmath453 hence , @xmath454 where we have selected the phase @xmath455 . here
@xmath456 is the isospin of the final state and @xmath268 is the weak phase .
thus necessary condition for non - zero @xmath0 odd observables is that the weak phase for each partial wave amplitude should be different .
for instance for the decays @xmath457 we have , @xmath458 * case iii : * here @xmath459 .
a_{\bar{f } } & = & \langle \bar{f}\left| \mathcal{l_{w}}\right| b^{0}\rangle = % \left [ e^{i\phi _ { 1}}f_{1\bar{f}}+e^{i\phi _ { 2}}f_{2\bar{f}}\right]\end{aligned}\ ] ] examples : @xmath461 @xmath462 @xmath11 gives , @xmath463\ ] ] subtracting and adding eqs .
( [ e2 ] ) and ( [ e1 ] ) , we get , we now discuss the decays listed in case ( ii ) where @xmath459 . subtracting and adding eqs . @xmath464 and @xmath465 , we get , @xmath466 where @xmath467}{% \gamma_{\bar{f},f } + \bar{\gamma}_{\bar{f},f } } \label{ccc5 } \\ a_{cp}^{\bar{f } } & = \frac{\bar{\gamma}_{f}-\gamma _ { \bar{f}}}{\gamma _ { \bar{f% } } + \bar{\gamma}_{f } } \notag \\
a_{cp}^{f } & = \frac{\bar{\gamma}_{\bar{f } } - \gamma _ { f}}{\gamma _ { f } + \bar{% \gamma}_{\bar{f } } } \label{ccc6 } \\
a_{cp}&= \frac{(\gamma_{\bar{f } } + \bar{\gamma}_{\bar{f } } ) - ( \bar{\gamma_{f}% } + \gamma_{f})}{(\gamma_{\bar{f } } - \bar{\gamma}_{\bar{f } } ) - ( \bar{% \gamma_{f } } + \gamma_{f } ) } \\ & = \frac{r_{f}a^{f}_{cp}-r_{\bar{f}}a^{\bar{f}}_{cp}}{\gamma } \label{ccc7}\end{aligned}\ ] ] where @xmath468 the following relations are also useful which can be easily derived from above equations @xmath469 \label{ccc9 } \\
\frac{r_{\bar{f}}-r_{f}}{r_{f}+r_{\bar{f } } } & = [ \delta c+a_{cp}c ] \label{ccc10 } \\ \frac{r_{\bar{f}}a_{cp}^{\bar{f}}+r_{f}a_{cp}^{f}}{r_{f}+r_{\bar{f } } } & = [ c+a_{cp}\delta c ] \label{ccc11}\end{aligned}\ ] ] for these decays , the decay amplitudes can be written in terms of tree amplitude @xmath470 and the penguin amplitude @xmath471 : @xmath472 \notag \\
a_{\bar{f } } & = e^{i\phi_{t}}e^{i\delta_{\bar{f}}^{t}}\bigl|t_{\bar{f}}\bigr| % \lbrack 1 + r_{\bar{f } } e^{i(\phi_{p}-\phi_{t})}e^{i\delta_{\bar{f } } } ] \label{ccc12}\end{aligned}\ ] ] where @xmath473 .
@xmath474 \notag \\ \bar{a}_{f } & = e^{-i\phi_{t}}e^{i\delta_{\bar{f}}^{t}}\bigl|t_{\bar{f}}% \bigr| \lbrack 1 + r_{\bar{f } } e^{-i(\phi_{p}-\phi_{t})}e^{i\delta_{\bar{f}% } } ] \label{ccc13}\end{aligned}\ ] ] @xmath475 @xmath476 hence for @xmath477 , we have @xmath478 \notag \\ a_{\bar{f } } & = \bigl|t_{\bar{f}}\bigr| e^{-i\gamma}e^{i\delta^{t}_{\bar{f } } } [ 1 - r_{\bar{f}}e^{i(\alpha + \delta_{\bar{f } } ) } ] \label{ccc16 } \\
\text{where } \qquad r_{f , \bar{f } } & = \frac{|v_{tb}| |v_{td}|}{|v_{ub}| \frac{r_{t}}{r_{b } } \frac{\bigl|p_{f,\bar{f } } \bigr|}{\bigl|t_{f,\bar{f } } % \bigr| } \label{ccc17}\end{aligned}\ ] ] and for @xmath479 , @xmath480 , we have @xmath481 \notag \\
a_{\bar{f}}^{d } & = \bigl|t_{\bar{f}}^{d}\bigr| e^{i\delta _ { \bar{f}% }
^{td}}[1-r_{\bar{f}}^{d}e^{i(-\beta + \delta _ { \bar{f}}^{d } ) } ] \label{ccc18 } \\
\text{where}\qquad r_{f,\bar{f } } & = r_{t}\frac{\bigl|p_{f,\bar{f}}^{d}\bigr|}{% \bigl|t_{f,\bar{f}}^{d}\bigr| } \notag\end{aligned}\ ] ]
as we have seen the cp asymmetries in the hadronic decays of b and k mesons involve strong final state phases
. thus strong interactions in these decays play a crucial role .
the short distance strong interactions effects at the quark level are taken care of by perturbative qcd in terms of wilson coefficients . the ckm matrix which connects the weak eigenstates
will mass eigenstates is another aspect of strong interactions at quark level . in the case of semi
leptonic decays , the long distance strong interaction effects manifest themselves in the form factors of final states after hadronization .
likewise the strong interaction final state phases are long distance effects .
these phase shifts essentially arise in terms of s - matrix which changes an in state into an out state viz .
@xmath482 in fact , the cpt invariance of weak interaction lagrangian gives for the weak decay @xmath483
@xmath484 it is difficult to reliably estimate the final state strong phase shifts .
it involves the hadronic dynamics .
using isospin , c - invariance of s - matrix and unitarity of s - matrix , we can relate these phases . in this regard ,
the decays @xmath485 described by two independent single amplitudes @xmath332 and @xmath486 discussed in section 4 case ( ii ) and the decays described by the weak amplitudes @xmath487 , described in section case ( iii ) are of interest the invariance of s - matrix viz .
@xmath488 would imply @xmath489 in the above decays , b is converted into @xmath490 . in particular , for the tree graph , the configuration is such that @xmath491 and d essentially go together into color singlet states will the third quark c(u ) recoiling ; there is a significant probability that system will hadronize as a two body final state .
thus at least for the tree amplitude @xmath492 should be equal to @xmath493 . to proceed further , we use the unitarity of s - matrix to relate the final state strong phases . the time reversal invariance gives @xmath494 where @xmath45 is the weak interaction lagrangian without the ckm factor such as @xmath495 . from eq .
@xmath496 , we have @xmath497 it is understood that the unitarity equation which follows from time reversal invaraince holds for each amplitude with the same weak phase . above equation can be written in two equivalent forms : 1 . exclusive version of unitarity + writing @xmath498 + we get from eq .
@xmath499 , @xmath500 + where @xmath501 is the scattering amplitude for @xmath502 . in this version ,
the sum is over all allowed exclusive channels .
this version is more suitable in a situation where a single exclusive channel is dominant one . to get the final result ,
one uses the dispersion relation .
2 . inclusive version of unitarity this version is more suitable for our analysis . for this case
, we write eq .
@xmath503 in the form @xmath504 parametrizing s - matrix as @xmath505 , we get after taking the absolute square of both sides of eq .
@xmath506 @xmath507 = \sum_{n , n^{\prime } \neq f } f_{n}s_{nf}^{\ast}f^{\ast}_{n^{\prime}}s_{n^{\prime}f } \label{4.73}\ ] ] the above equation is an exact equation . in the random phase approximation
, we can put @xmath508 we note that in a single channel description : @xmath509 the absorption takes care of all the inelastic channels .
similarly for the amplitude @xmath510 , we have @xmath511 the c - invariance of s - matrix gives : @xmath512 thus in particular c - invariance of s - matrix gives @xmath513 hence from eq . @xmath514 , using eqs .
( @xmath515 ) , we get @xmath516=\rho ^{2},\bar{\rho}^{2 } \label{4.78}\ ] ] where @xmath517 it is convenient to write eq . @xmath518 in the form @xmath519
\label{aa1 } \\ 0 & \leq ( \delta _ { f,\bar{f}}-\delta ) \leq \theta \label{aa2 } \\
-\theta & \leq ( \delta _ { f,\bar{f}}-\delta ) \leq 0 \label{aa3}\end{aligned}\ ] ] where @xmath520 .
the strong interaction parameters @xmath521and@xmath522can be determond by strong interaction dynamics . using @xmath523 ,
c - invarience of strong interactions and regge pole phenomonology , the scattering aplitude @xmath524 for two particle final state can be calculated.(for details see ref . [ 12 ] ) .
the s - wave scattering amplitude @xmath525 for the decay modes @xmath526 which are s - wave decay modes of @xmath233 is given by@xmath527where @xmath528using the relation @xmath529 , the phase shift @xmath530 the parameter @xmath531 and the phase angle @xmath532 can be determind .
one gets @xmath533@xmath534hence we get the following bounds@xmath535for the tree amplitude , factorization implies @xmath536 we can therefore take the point of view , the effective final state phase shift is given by @xmath537we take the lower bounds for the tree amplitude so that final state effective phase shift @xmath536 for the penguin we assume that the effective value of the final state phase shift @xmath538 is near the upper bound .
thus for @xmath539 @xmath540 ; for @xmath541 the phase shift @xmath542 where as for @xmath543 the phase shift @xmath544 these phase shifts are relavent for the direct cp - asymmetries for @xmath545 and @xmath298 decays .
the decay @xmath546 is described by two amplitudes ( for details see ref.[13])@xmath547 = \left\vert p\right\vert \left [ 1-re^{i\left ( \gamma + \delta _ { + -}\right ) } \right ] \\ p & = & -\left\vert p\right\vert e^{-i\delta _ { p}},\
\ t=\left\vert t\right\vert \\ \delta _ { + - } & = & \delta
_ { p},\text { \ } r=\frac{\left\vert t\right\vert } { % \left\vert p\right\vert } \\ a_{cp}\left ( b^{0}\rightarrow \pi ^{-}k^{+}\right ) & = & \frac{-2r\sin \gamma \sin \delta _ { + -}}{r } \\ r & = & 1 - 2r\cos \gamma \cos \delta _ { + -}+r^{2}\end{aligned}\]]neglecting the terms of order @xmath548@xmath549from the experimental values of @xmath550 and @xmath551 with @xmath552 we get @xmath553however for @xmath554 ( which corresponds to @xmath555 the value one gets for @xmath556 ) , we get @xmath557 the phase shift @xmath558 for the @xmath559 is compatible with the experimental value of the direct cp - asymmetry for @xmath546 decay mode . for @xmath560 @xmath561
is compatible with the value @xmath562 obtained by the authers of ref .
[ 13 ] . in any case
, our analysis shows that the upper limit for final state stronge phase @xmath563 is around @xmath564 .
finally , we note that the actual value of the effective final state phase shift @xmath565 depends on one free parameter @xmath566 the factorization implies @xmath567 @xmath568 for the tree amplitude ; for the penguin amplitude , @xmath538 depends on @xmath566 in any case it can not be greator than the upper bound .
* case ii : * now , we discuss the experimental tests to verify the equality ( implied by c - invariance of s - marix ) of phase shifts @xmath62 and @xmath569 for the decays @xmath570 and @xmath571 . from eqs.@xmath572
, we note that cp - asymetries : @xmath573involve dthe weak phase @xmath421 and strong phase @xmath574 these asymmetries are of interst because for @xmath575and @xmath576 hence we can verify the equality of phases @xmath62 and @xmath577 and determine the weak phase @xmath578 for @xmath418 replace @xmath579 , @xmath580 , @xmath581 and @xmath582 by @xmath583 in standared model @xmath584 the experimental results for the b decays are as follows discussed in section 4 @xmath585 to determine the parameter @xmath586 or @xmath587 , we assume factorization for the tree amplitude .
factorization gives for the decays @xmath588 : @xmath589 @xmath590 .form factors [ cols="^,^,^,^",options="header " , ] for @xmath591 , change @xmath592 , @xmath593 .
for example , for @xmath594 @xmath595 these examples have been selected keeping in view that final state phases @xmath596 are not too large . for @xmath597 ,
we have used eq . @xmath598 neglecting the second order term .
an attractive option is @xmath599 for each value of @xmath600 ; although @xmath601 is also a possibility .
@xmath602 implies @xmath603 . neglecting terms of order @xmath604
, we have @xmath605 now the second term in eq .
@xmath606 vanishes and using the value of @xmath607 given in eq .
@xmath608 , we get @xmath609 assuming @xmath610 , we obtain @xmath611 finally the cp asymmetries in the limit @xmath612 @xmath613}{\gamma ( 1+a_{cp } ) } \notag \\ & = \sqrt{1-c_{\bar{f}}^{2}}\sin ( 2\alpha _ { eff}^{\bar{f}}+\delta ) \notag \\ & = -\sqrt{1-c_{\bar{f}}^{2}}\cos \delta \label{cccc7 } \\
s_{f}=s-\delta s & = \frac{2\text{im}[e^{2i\phi _ { m}}a_{f}^{*}\bar{a}_{f}]}{% \gamma ( 1-a_{cp } ) } \notag \\ & = \sqrt{1-c_{f}^{2}}\sin ( 2\alpha _ { eff}^{\bar{f}}-\delta ) \notag \\ & = \sqrt{1-c_{f}^{2}}\cos \delta \label{cccc8}\end{aligned}\ ] ] the phase @xmath614 is defined as @xmath615 hence we have @xmath616
in weak interaction , both p and c are violated but cp is conserved by the weak interaction lagrangian . hence for @xmath617 complex @xmath618 ; the mass matrix is not diagonal in latexmath:[$% assuming @xmath0 conservation , the @xmath0 eigenstates @xmath620 and @xmath621 , the time evolution of this state can generate the state @xmath622 .
this is a source of mixing induced @xmath0 violation .
however , both in @xmath623 and @xmath203 complex , the mass eigenstates @xmath624 , @xmath625 complex , there is a small admixture of wrong @xmath0 state characterized by a small parameter @xmath626 , which gives rise to the @xmath0 violating decay @xmath627 .
this was the first @xmath0 violating decay observed experimentally . for @xmath203 complex , the mismatch between mass eigenstates and @xmath0 eigenstates @xmath628 where the phase factor is @xmath629 in the standard model viz .
one of the phases in the ckm matrix . for @xmath630
, there is no mismatch between @xmath0 eigenstates latexmath:[$% eigenstates .
there is no extra phase available in ckm matrix , with three generations of quarks to accomodate more than two independent phases @xmath582 and @xmath422 ; the unitarity of ckm matrix requires @xmath632 .
the quantum mechanical interference gives rise to non zero mass differences @xmath633 , @xmath634 and @xmath364 between mass eigenstates . the mixing induced @xmath0 violation involves these mass differences .
where @xmath332 and @xmath636 are the amplitudes for the decays @xmath637 and @xmath638 , the states @xmath639 and @xmath640 being @xmath0 conjugate of each other . for direct @xmath0 violation , at least two amplitudes with different weak phase are required : @xmath641 in section 2 , the @xmath0 violation due to mismatch between @xmath0 eigenstates @xmath650 and direct @xmath0 violation due to different weak phases bewteen the cecay amplitudes @xmath159 and @xmath167 are discussed .
it is a good illustration of @xmath0 violation due to mismatch between mass and @xmath0 eigenstates , involving the mixing parameter @xmath634 . from the experimental values of @xmath655 and @xmath656 ,
the weak phase @xmath657 is found to be @xmath658 .
corresponding to @xmath659 , we have @xmath660 and for this decay @xmath661 in this section for the case ( * i * ) , both direct and mixing induced @xmath331 violation viz .
@xmath664 , @xmath665 and @xmath666 for @xmath667 described by two amplitudes @xmath7 and @xmath668 given by tree and penguin diagrams is discussed .
we find @xmath669 and @xmath670 is essentially given by @xmath671 we consider the cdecays described by two independent decay amplitudes @xmath332 and @xmath674 with different weak phases @xmath675 and @xmath676 where the final states @xmath639 and @xmath640 are @xmath9 and @xmath0 conjugate of each other such as the states @xmath677 @xmath678 , @xmath679 @xmath680 , @xmath681 @xmath682 . 1 . no evidence that space - time symmeries are violated by fundamental laws of nature .
the translational and rotational symmetries imply that space is homogeneous and isotropic .
@xmath703if we examine the light emitted by a distant object billions of light years away , we find that atoms have been following the same laws as they are here and now .
( translational symmetry ) 2 .
discrete symmetries are not universal ; both c and p are violated in the weak interaction but repsected by electromagnetic and strong interactions .
there is no evidence for violation of time reversal invariance by any of the fundamental laws of nature .
3 . basic weak interaction lagrangian is cp conserving .
cp violation in weak interactions is a consequence of mismatch between mass eigenstates and cp eigenstates and or mismatch between weak and mass eigenstates at quark level .
there is no evidence of cp violation in lepton sector .
there is no evidence that cp invariance is violated by any of the fundamentals laws of nature as implied by cpt invariance and t - invariance .
cp violation in weak decays is an example where basic laws are cp invariant but states at quark level contain cp violating phases .
the fundamental interaction governing atoms and molecules is the electromagnetic interaction which does not violate bilateral symmetry ( left - right symmetry ) . in nature
we find organic molecules in asymmetric form , i.e. left handed or right handed .
this is another example where the basic laws governing these molecules are bilateric symmetric but states are not .
( asymmetric intial conditions ? ) 6 .
* baryon asymmetry of the universe : baryogenesis * : no evidence for existence of antibaryons in the universe .
the universe started with a complete matter antimatter symmetry in big bang picture . in subsequent evolution of the universe ,
a net baryon number is generated .
this is possible provided the following conditions of sakharov are satisfied 1 .
there exists a baryon number violating interaction .
2 . there
exist c and cp violation to induce the asymmetry between particle and antiparticle processes .
3 . departure from thermal equilibrium of x - particles which mediate the baryon number violating interactions
there seems to be no connection between cp violation required by baryogenesis and cp violation observed in weak decays .
j. d. bjorken , topics in b - physics , nucl .
11 ( proc.suppl . ) 325 ( 1989 ) ; m. beneke , g. buchalla , m. neubart and c. t. sachrajda , phys .
lett , 83 , 1914 ( 1999 ) , nucl .
b591 313 ( 2000 ) ; c. w. bauer , d. pirjol and i. w. stewart , phys .
87 , 201806 ( 2001 ) hep - ph/0107002 . 1 .
the unitarity triangle 2 .
the box diagram 3 .
the tree diagram 4 .
the penguin diagram 5 .
\(a ) @xmath406-exchange diagram for @xmath705 + \(b ) @xmath406-exchange diagram for @xmath706 6 .
annihilation diagram for @xmath707 7 .
plot of equation @xmath708 for different values of @xmath709 for @xmath710where solid curve , dashed curve , dashed doted curve , dashed bouble doted and double dashed doted curve are corresponding to @xmath711 and @xmath712 respectively . | in this review we give an overview of @xmath0-violation for @xmath1 @xmath2 @xmath3 systems .
direct @xmath4violation and mixing induced @xmath0-violation are discussed . |
neutrino experiments have revealed that neutrinos have mass and that , similar to quarks , leptons mix .
neutrino masses were discovered with the observation that neutrinos change flavor as prescribed by the mechanism of mass - induced neutrino oscillations @xcite .
three - flavor neutrino oscillations suffice to explain all neutrino data , with the exception of those from lsnd @xcite .
the lsnd result , however , still awaits confirmation from another experiment and will not be considered here .
nonetheless , even under the conservative assumption that there are no other surprises in the neutrino sector , our understanding of leptons remains incomplete .
while we currently know the absolute value of the two neutrino mass - squared differences and two out of four mixing parameters that govern neutrino@xmath4neutrino and antineutrino@xmath4antineutrino oscillations , we have little or no information regarding @xcite . ] 1 . the magnitude of the @xmath0-element of the lepton mixing matrix , or the value of @xmath1 ; 2 . the value of the complex phase @xmath5 which parameterizes cp- ( and t-)invariance violation in neutrino@xmath4neutrino and antineutrino@xmath4antineutrino oscillations ; 3 . the neutrino mass hierarchy , or the sign of @xmath6 ; 4 .
the magnitude of the neutrino masses ( we only know mass - squared differences and various upper bounds on other combinations of the neutrino masses ) ; 5 . whether neutrinos are majorana or dirac fermions .
items 1 , 2 , and 3 can be addressed by next - generation neutrino oscillation experiments . here
, we wish to concentrate on item number three and the outstanding question : how can one experimentally determine the neutrino mass hierarchy ?
the neutrino masses @xmath7 , @xmath8 , are defined in the following convenient way : @xmath9 , and @xmath10 , where @xmath11 .
a positive value of @xmath6 implies @xmath12 and a so - called normal mass hierarchy , while a negative value of @xmath6 implies @xmath13 and a so - called inverted mass hierarchy .
examples of an inverted and a normal neutrino mass hierarchy are depicted in fig .
[ normal_inverted ] .
the question of whether the neutrino mass hierarchy is normal or inverted is of considerable theoretical interest @xcite .
for example , one can argue on general grounds that grand unified theories favor a normal neutrino mass hierarchy @xcite .
this may be understood qualitatively by noting that grand unified models relate leptons to quarks , and the quark mass hierarchies are normal .
an inverted neutrino mass spectrum , on the other hand , would be quite un - quark like , and would probably reflect an underlying global leptonic symmetry ( see @xcite and references therein ) that gives rise to the near degeneracy of @xmath14 and @xmath15 .
it is easy to understand , qualitatively , why the current data are unable to distinguish positive from negative @xmath6 .
almost all information we have on @xmath6 comes from @xmath16 disappearance at atmospheric experiments and k2k .
these provide information on the @xmath16 survival probability @xmath17 , which is given ( in vacuum ) by @xmath18 modulo terms proportional to @xmath19 or @xmath3 .
these are suppressed because @xmath20 ( at the three sigma level , @xmath21 @xcite ) , which we know from solar and kamland data , and @xmath22 ( at the three sigma level , @xmath23 @xcite ) , which we know from reactor and atmospheric data .
current experiments are not sensitive to these sub - leading terms , and it is clear that eq .
( [ pmumu_today ] ) is only sensitive to @xmath24 .
hence both a normal and an inverted mass hierarchy fit the data equally well . in this paper , we explore several different oscillation probes of the neutrino mass hierarchy , and discuss under what conditions these can be experimentally accessible .
more specifically , we concentrate our efforts on probes other than `` matter - affected 13 oscillations , '' which have been heavily explored in the literature and are defined and very briefly discussed in sec .
[ sec : traditional ] .
to date , all phenomenological studies of how to determine the neutrino mass hierarchy via neutrino oscillations have , to the best of our knowledge , relied on a non - zero value for @xmath3 , and almost all appeal to matter effects in the 13 sector .
one of the goals of this paper is to argue that neither matter effects nor a nonzero value of @xmath3 are required in order to determine the neutrino mass hierarchy via neutrino oscillations at least in principle .
there are many reasons for pursuing novel oscillation probes of the neutrino mass hierarchy .
one is that we would , ultimately , like to compare distinct determinations of the mass hierarchy .
such a comparison provides a test of whether there is new physics in the lepton sector besides three flavor oscillations .
another ( and , perhaps , the most important ) reason for pursuing different hierarchy - exploring channels is that , if @xmath1 turns out to be too small , the traditional approach will not work , so new experimental probes are not only desirable but absolutely necessary ! our presentation is the following . in sec .
[ sec : traditional ] , we briefly review how the mass hierarchy can be revealed if @xmath3 is large enough and there are nontrivial matter effects at work . in sec .
[ sec : vacuum ] , we remind readers that even in the case of pure vacuum oscillations the mass hierarchy is , in principle , observable .
we then review one realistic method for determining the mass hierarchy using only vacuum disappearance channels . in sec .
[ sec : theta13=0 ] , we study in detail the case @xmath25 , and discuss various approaches one might consider in this case to determine the mass hierarchy . in sec . [ sec : end ] , we summarize our results and offer some concluding thoughts . a technical presentation of the oscillation probabilities in constant matter for vanishing @xmath1 is included as an appendix .
our approach is `` theoretical , '' in the sense that we study the behavior of oscillation probabilities as a function of the neutrino energy @xmath26 and the baseline @xmath27 instead of the expected number of observed events for a given neutrino beam , detector type , size , resolution , etc .
nonetheless , our approach allows one to estimate `` how precise '' certain measurements need to be so that the inverted and normal mass hierarchy lead to experimentally distinct results .
one very promising way of revealing the character of the neutrino mass hierarchy is to probe oscillations which are significantly affected by matter effects in the 13 sector @xcite . in this section ,
we very briefly describe one manifestation of this well known phenomenon .
readers are referred to @xcite for more details . by `` matter
affected 13 oscillations ''
we mean neutrino oscillations that take place under the following two conditions : ( i ) @xmath28 is large enough required for unveiling the mass hierarchy via matter affected 13 oscillations depends on the specific properties of the neutrino source and the detector . for detailed discussions , see @xcite .
] and ( ii ) the propagating medium contains electron number densities @xmath29 such that the matter potential , @xmath30 , where @xmath31 is the fermi constant , is comparable to the 13 frequency , @xmath32 . note that , while the latter can be arranged ( at least in principle ) by choosing an appropriate neutrino beam and propagating medium , the former requires cooperation from mother nature .
matter affected 13 oscillations can be studied , for example , in accelerator - based , long - baseline searches for @xmath33 @xcite . in constant matter densities , the @xmath33 oscillation probability @xmath34 is given by @xmath35 where @xmath36 is the matter potential .
( [ p_mue_matter ] ) is a good approximation as long as @xmath37 terms can be ignored .
@xmath38 is sensitive to the relative sign between @xmath39 and @xmath40 . since @xmath41
, a normal neutrino mass hierarchy implies that , for a fixed value of @xmath1 , @xmath42 is _ larger _ than its vacuum counterpart , @xmath43 , while @xmath44 is _
smaller_. is the @xmath45 oscillation probability . in matter
, antineutrino oscillation probabilities are given by the same expression as the neutrino ones , with @xmath46 and @xmath47 . ] the opposite behavior is expected for an inverted hierarchy . while @xmath34 is sensitive to the mass hierarchy , measuring it alone at approximately fixed values of @xmath27 and @xmath26
is often not enough to experimentally determine the mass hierarchy .
this is due to a well known @xmath1mass hierarchy degeneracy , which can be understood in the following way .
if one only measures @xmath48 , for every normal hierarchy characterized by @xmath49 , there will be an inverted one with @xmath50 that yields exactly the same value of @xmath48 .
note that , as long as @xmath51 , the degeneracy is dominated by the relative size of @xmath52 and @xmath53 and the sign of @xmath6 . qualitatively , @xmath48 can be chosen the same for a normal and an inverted mass hierarchy if one chooses @xmath52 to be less than @xmath53 .
the degeneracy described above can be broken by more experimental information .
simple ways of determining the mass hierarchy include comparing @xmath34 and @xmath44 at @xmath54 ( which can be done with one detector and both muon - neutrino and muon - antineutrino beams ) , or comparing measurements of @xmath34 at @xmath54 for different values of @xmath27 ( which requires two different detectors , but only muon - neutrino beams ) .
another way is to independently measure @xmath55 precisely enough such that one can statistically distinguish between @xmath56 and @xmath57 .
other probes of the neutrino mass hierarchy that depend on matter effects and a nonzero @xmath3 include precise atmospheric neutrino studies @xcite and measurements of the neutrino flux emitted during a nearby supernova explosion @xcite .
the procedures referred to above do not take advantage of the fact that , in the case of three neutrino flavors , neutrino oscillations are governed by two distinct oscillation frequencies . indeed , extracting the neutrino mass hierarchy via matter affected 13 oscillations works perfectly well even in the limit @xmath58 , when solar effects become utterly unobservable @xcite .
all procedures referred to above , however , rely fundamentally on a nonzero value of @xmath3 and neutrino oscillations in matter . in the next two sections
, we argue that neither the former nor the latter is required in order to determine the mass hierarchy through neutrino oscillations at least in principle .
to conclude this section , it is illustrative to mention that the `` solar mass - hierarchy '' was revealed by measuring how matter effects modify oscillations in the 12 sector . since @xmath19 is chosen positive - definite , the `` solar mass - hierarchy '' question is the following : is the @xmath59 flavor eigenstate `` predominantly @xmath60 '' or `` predominantly @xmath61 , '' _
i.e. _ , is @xmath62 larger or smaller than @xmath63 ? , this is equivalent to inquiring about the sign of @xmath64 , or whether the solar angle is in the light or the dark - side of the parameter space @xcite . ] the electron - type neutrino survival probability in vacuum , assuming that @xmath65 effects are averaged out , is incapable of addressing this question . under these circumstances
, @xmath66 is only a function of @xmath67 and hence invariant under @xmath68 .
matter effects break the degeneracy , such that , for @xmath69b solar neutrinos , @xmath70 .
current data reveal that @xmath71 ( and hence @xmath72 ) , at more than the five sigma level .
one reason we have been able to disentangle the solar mass hierarchy is that @xmath73 is `` different enough '' from @xmath74 .
had that not been the case , we would have been unable to determine whether @xmath75 ( @xmath76 ) or @xmath77 ( @xmath78 ) clearly a physical question ! via matter affected 12 oscillations .
in the case of only two neutrino flavors , the neutrino mass hierarchy @xcite is simply unobservable when it comes to studying vacuum flavor oscillations .
this is easy to see . with only two flavors , @xmath79 , and
there is simply no sensitivity to the sign of @xmath80 or the sign of @xmath81 .
this does not , of course , mean that the mass hierarchy is not physical it is just not observable if one measures only vacuum oscillation probabilities . in the case of three or more flavors ,
this is no longer the case .
pure vacuum oscillations can , in general , determine whether the neutrino mass hierarchy , as defined in sec .
[ sec : intro ] , is normal or inverted . in this section
we explicitly present how this can be done in principle , and then review how it could be done in practice by comparing accelerator @xmath16 and reactor @xmath82 disappearance searches . in vacuum
, the survival probability @xmath83 can be written as @xmath84 where @xmath85 .
note that @xmath86 .
( [ eq : p_vac ] ) is valid , of course , independent of the mass hierarchy .
we wish to probe whether normal and inverted mass hierarchies necessarily lead to different values of @xmath83 .
we will concentrate our efforts on comparing inverted and normal hierarchies which are capable of fitting the current neutrino data . more specifically
, we will only consider normal and inverted mass hierarchies that share the same value of @xmath19 and the same mixing matrix .
these constraints clearly satisfy the requirement above and substantially simplify our presentation .
relaxing these constraints would render our analyses much more opaque , and would not modify qualitatively our results .
hence , we address the following question : for a given choice of @xmath87 corresponding to a normal mass hierarchy ( @xmath88 ) , is there an inverted mass spectrum characterized by the same mixing matrix and @xmath19 but a different @xmath89 ( @xmath90 ) that yields exactly the same survival probability @xmath83 , for all values of @xmath91 ?
define @xmath92 by @xmath93 .
note that @xmath92 is restricted to be small enough so that @xmath94 is indeed a negative quantity . should be small enough so that both @xmath95 and @xmath96 agree with the currently allowed range of values for the `` atmospheric '' mass - squared difference . ] in practice , @xmath92 turns out to be of order @xmath19 , as we will see shortly . the difference between @xmath97 and @xmath98 , @xmath99 , is @xmath100\right\ } , \label{p_vac_diff}\end{aligned}\ ] ] where
@xmath101 and , for simplicity , we have dropped the ` @xmath102 ' superscript in @xmath95 .
there is , in general , no constant value of @xmath92 that renders eq .
( [ p_vac_diff ] ) exactly zero for all @xmath27 .
is restricted to be small , such that @xmath103 .
@xmath104 is guaranteed to be true for @xmath105 , independent of whether the oscillations occur in matter or in vacuum .
this is trivial to understand , since @xmath105 renders the two mass patterns in question identical such a large @xmath92 value `` inverts the inverted hierarchy . '' ] the difference in the first square - bracket requires @xmath106 , while the difference in the second one calls for @xmath107 . curiously enough , there is a special solution in the case @xmath108 : @xmath109 .
qualitatively , it is easy to understand what is happening .
the distinction between a normal and an inverted mass hierarchy is , as far as vacuum oscillations are concerned , the following . for a normal mass hierarchy , @xmath110 . for an inverted one
, the situation is reversed : @xmath111 .
one can choose , as is depicted in fig .
[ normal_inverted ] , all normal and inverted mass hierarchy frequencies ( @xmath112 ) to agree by equating @xmath113 and @xmath114 equivalent to @xmath109 . in this case , however , @xmath115 , because the @xmath116 and @xmath117 amplitudes are , in general , different ( except in the special case @xmath108 ) .
we conclude that it should be possible to distinguish between the two hierarchies by studying neutrino oscillations in vacuum .
the key requirement is the ability to tell @xmath6 from @xmath118 or , in other words , simultaneous sensitivity to @xmath19 and @xmath6 effects .
all experiments performed to date fail to have this simultaneous sensitivity for different reasons .
`` @xmath19 experiments , '' like kamland , lack sensitivity to @xmath6 effects , which `` average out . ''
turns out to be too small . ] `` @xmath6 experiments , '' like k2k , lack sensitivity to @xmath19 effects , which are suppressed by a small @xmath119 . from the discussion above , it seems that an experiment capable of measuring @xmath6 with an uncertainty @xmath120 should be able to tell a normal from an inverted mass hierarchy .
however , it turns out that this is not automatically the case .
assume one is probing @xmath83 with neutrino energies and baselines such that @xmath121 , as is the case for proposed next - generation accelerator and reactor neutrino experiments @xcite .
under these circumstances , @xmath122 and @xmath123 + o\left((\delta_{12 } l)^2,(xl)^2\right ) .
\label{p_small_diff}\ ] ] we are interested in values of @xmath92 such that @xmath124 , and have made use of the fact that @xmath125 one can easily find @xmath92 such that , neglecting @xmath126 terms , @xmath127 : @xmath128 for @xmath129 and ignoring the possibility that @xmath130 .
this means that if one probes , say , @xmath17 in vacuum at values of @xmath27 and @xmath26 such that @xmath122 and measures , assuming a normal mass - hierarchy , @xmath131 ev@xmath132 ( a 2% measurement of the `` atmospheric '' mass - squared difference ) , there will be a statistically equivalent solution for an inverted mass hierarchy and @xmath133 ev@xmath132 if @xmath134 ev@xmath132 .
, @xmath135 as the value of @xmath92 required to render the @xmath136 survival probability for an inverted and normal hierarchy indistinguishable in a range of values of @xmath27 and @xmath26 .
similarly , we define @xmath137 as the equivalent quantity for @xmath138 transitions ( @xmath139 ) . ] today , @xmath140 is constrained to be @xmath141\times 10^{-3}~\rm ev^2 , \label{eq : xmu}\ ] ] taking three sigma upper and lower bounds for the oscillation parameters @xcite .
appearance channels also contain a similar normal inverted degeneracy .
if one were to probe @xmath138 in a setup where @xmath122 , a normal and inverted mass - hierarchy would yield the same oscillation probability @xmath142 ( up to @xmath143 effects ) if @xmath144 where @xmath145 $ ] .
one way to eliminate the degeneracy presented above is to compare two distinct oscillation channels , since the value of @xmath92 that renders inverted and normal hierarchies indistinguishable depends on the parent and daughter neutrino flavors . for concreteness
, we will discuss under what conditions one can determine the mass - hierarchy by studying @xmath82 and @xmath16 disappearance ( at , say , a reactor neutrino experiment and a long - baseline conventional neutrino beam setup , respectively ) .
one variant of this procedure has already been explored ( using a different `` language '' ) in the literature @xcite and is , to the best of our knowledge , the only oscillation probe of the neutrino mass hierarchy identified so far that does not rely significantly on matter effects . in the case of a reactor neutrino experiment , the vacuum expression for
the survival probability is a very good approximation . for accelerator - based experiments ,
this is only true for low - energy neutrinos ( @xmath148 few hundred mev ) assuming that the neutrinos are propagating inside the earth s crust .
some consequences of nontrivial matter effects are discussed in secs .
[ sec : traditional ] ( for nonzero @xmath3 ) and [ sec : theta13=0 ] ( for @xmath149 ) . in the @xmath16 disappearance experiment
, the two hierarchies can be made compatible by choosing @xmath150 , as defined in eq .
( [ eq : xmu ] ) . on the other hand , for the reactor experiment , the two hierarchies will yield identical results if @xmath151 , given by @xmath152\times 10^{-3}~\rm ev^2,\ ] ] taking current three sigma upper and lower bounds for @xmath153 and @xmath19 . as long as @xmath154
, the `` wrong hierarchy '' measurements of @xmath6 obtained via the electron - type antineutrino and muon - type neutrino disappearance analyses will disagree , thus breaking the degeneracy .
figure [ xmu_xx](top ) depicts the currently allowed values of @xmath140 and @xmath155 , in units of @xmath19 , if all oscillating parameters are allowed to vary within their two , three , and four sigma allowed ranges @xcite . in order to realistically determine the mass hierarchy ,
both the reactor and accelerator experiments are required to measure the atmospheric mass - squared difference with great precision : @xmath156 .
given what is currently known of the oscillation parameters , these could conspire in such a way that @xmath157 , as can be clearly seen in fig .
[ xmu_xx](top ) . improved knowledge of the oscillation parameters ( especially @xmath73 and @xmath1 ) will help determine how challenging this proposition really is . while a 2% measurement of @xmath6 in an accelerator @xmath16 disappearance experiment does not seem too far - fetched , as precise a measurement in a reactor experiment seems too optimistic .
`` beta beams , '' for example , may serve as a more advantageous alternative capable of precisely studying @xmath158 transitions in vacuum @xcite .
we are aware of an ongoing in - depth analysis of how to determine the neutrino mass hierarchy via a comparison of @xmath159 and @xmath160 ( @xmath161 ) , to appear shortly @xcite .
similar options which are potentially useful include comparing @xmath159 to @xmath162 or @xmath33 .
the currently allowed values of @xmath140 and @xmath163 are depicted in fig .
[ xmu_xx](bottom ) .
needless to say , precisely measuring @xmath6 in an electron or tau appearance channel seems far from realistic .
furthermore , as far as tau appearance experiments are concerned , neutrino energies are required to be significantly larger than 1 gev , and the pure vacuum approximations we have been discussing so far simply do not apply in the case of an earth - based accelerator experiment .
we will briefly comment on the @xmath162 channel in the next section . comparing the inverted normal hierarchy degeneracy in @xmath159 and @xmath164 oscillations
requires more care . the value of @xmath165 obtained from eq .
( [ eq : xab ] ) diverges as @xmath166 as @xmath167 , and one may be tempted to conclude that @xmath165 could end up much larger than @xmath140 , rendering this comparison very sensitive to the mass hierarchy .
this is not necessarily the correct conclusion for a few reasons .
one is that , for small values of @xmath3 , the @xmath6 dependency of the @xmath33 oscillation probability disappears , and at some point all sensitivity to `` atmospheric '' oscillations goes away @xmath165 may be large , but not experimentally accessible !
another is that , in the limit @xmath168 , eq . ( [ eq : xab ] ) need not apply , since it was derived under the assumption that @xmath124 .
we emphasized in sec . [
sec : traditional ] that when @xmath167 the traditional approach to determining the mass hierarchy through matter effects in the 13 sector fails , and new exploration channels are not only desirable but necessary .
furthermore , the approach studied in the previous section also fails in the limit @xmath167 .
one reason is simply that , in this case , oscillations involving electron neutrinos do not depend on @xmath6 ( including its sign ) .. ] hence , in the limit @xmath169 , we are forced to explore the neutrino mass hierarchy through @xmath17 , @xmath170 and @xmath171 .
for vanishing @xmath1 , neutrino oscillation experiments can still distinguish , in principle , an inverted mass hierarchy from a normal one .
for example , in vacuum , @xmath172 ( see eq .
( [ p_vac_diff ] ) ) still does not vanish for all @xmath91 for any choice of @xmath92 if @xmath25 , unless @xmath173 ( maximal solar mixing , experimentally excluded ) .
the same is true of @xmath174 .
it is , however , curious that no new information is obtained if one compares the muon disappearance , tau disappearance , and tau appearance channels ( for the same value of @xmath27 and @xmath26 ) .
this can be seen in a variety of ways and boils down to the fact that , for @xmath175 , all three oscillation channels change `` in the same way '' when one compares a normal to an inverted mass hierarchy .
for example , in the limit @xmath149 and @xmath122 , @xmath176 using eqs .
( [ eq : xaa],[eq : xab ] ) . the same behavior is observed in matter or outside the limit @xmath122 .
for details , see appendix [ app : tech ] .
it suffices , therefore , to concentrate one s efforts on the muon disappearance channel , which is what we do in this section . throughout
, we will assume @xmath149 , unless otherwise noted . in sec .
[ sec : vacuum ] , we argued that as long as experiments are performed under conditions such that @xmath122 , for every normal hierarchy choice of mass - squared differences there is an inverted hierarchy one that yields the same oscillation probability ( up to @xmath177 corrections ) : @xmath93 , where @xmath92 is given by either eq .
( [ eq : xaa ] ) or eq .
( [ eq : xab ] ) . in order to break this degeneracy ,
it is clear that one needs to explore setups where @xmath178 .
values of @xmath179 imply very long baselines or very small neutrino energies . given
that we are considering experiments in which a muon is created , we need to steer clear from very low energies , and are forced to look at long baselines . under these circumstances , it is quite likely that we will have to deal with relevant matter effects ( short of considering experiments that shoot neutrino beams into space ) .
nonetheless , we first warm - up by discussing how one should go about disentangling the mass - hierarchy with muon disappearance in pure vacuum oscillations , and turn to matter effects in the next subsection .
we start by making the following point : for a fixed @xmath27 and @xmath26 there are always values of @xmath92 that renders @xmath104 .
explicitly , @xmath180 translates ( after a long series of trigonometric manipulations ) into @xmath181=0.\ ] ] the values of @xmath92 that satisfy the trigonometric equation above are @xmath182 , integer @xmath183 , and @xmath184 for integer values of @xmath183 and the appropriate definition of the inverse tangent function .
the first set of solutions violates the restrictions we impose on the allowed values of @xmath92 ( as already pointed out , @xmath185 transforms an inverted hierarchy into a normal one ) and will be ignored .
values of @xmath92 that satisfy eq .
( [ eq : x(l / e ) ] ) are depicted in fig .
[ fig : xvac ] as a function of @xmath186 , for @xmath187 .
as expected ( see eq .
( [ eq : mumu = tautau ] ) ) , for @xmath122 , @xmath188 for @xmath189 .
figure [ fig : xvac ] reveals that , for a fixed @xmath91 , there are values of @xmath92 of order @xmath19 that render @xmath190 , even when @xmath191 .
indeed , for large enough values of @xmath119 , there are several ( discreet ) values of @xmath92 of order @xmath19 that do the trick . ,
all @xmath92 values that obey eq .
( [ eq : x(l / e ) ] ) for @xmath192 diverge , and only the @xmath189 solution is acceptable .
] on the other hand , there is no constant value of @xmath92 capable of solving @xmath190 for all @xmath91 , as already extensively discussed .
it is , therefore , clear that while a normal and an inverted mass hierarchy lead to distinct @xmath17 , if @xmath159 processes are studied around a fixed value of @xmath91 , it is not possible to tell one mass hierarchy from the other . in order to break the degeneracy ,
more experimental information is required .
for example , if one were to perform a very precise `` short baseline '' experiment characterized by small values of @xmath119 , one would find two degenerate solutions for the `` atmospheric '' mass - squared difference : @xmath95 and @xmath193 . on the other hand , a `` long baseline '' narrow - band - beam experiment with typical energy and baseline such that @xmath194 , would measure , assuming that the neutrino mass hierarchy is indeed normal , @xmath95 and @xmath195 .
the mass hierarchy would be determined by combining both data sets as long as the precision @xmath196 with which _ both _ experiments measure @xmath6 is better than @xmath197 . as an example , if @xmath198 at the `` long - baseline '' site , @xmath199 .
[ fig : xvac ] can also be used to estimate how precise a broad - band - beam experiment in vacuum needs to be in order to establish the mass hierarchy . if one takes this experiment s data and `` bins '' it in @xmath91 , each @xmath91 bin ( or @xmath119 bin ) has to be capable of measuring @xmath6 with @xmath200 .
bin will contain several @xmath6 `` wavelengths , '' which allows one to , in principle , perform a measurement of @xmath6 .
we will return to this issue with more details in the next subsection .
] the same is true , of course , of experiments with variable baselines , most notably atmospheric neutrino experiments .
we return to these issues and make a more realistic assessment of the situation in the next subsection . when @xmath149 , the importance of matter effects is estimated by comparing @xmath37 with the matter potential @xmath40 .
numerically @xmath201 we are interested in neutrino energies larger than 100 mev , so that @xmath202 charged current processes are kinematically allowed , and typical earth densities , which are around 25 g/@xmath203 .
as shown in appendix [ app : tech ] , @xmath17 in matter can be written as @xmath17 in vacuum with @xmath204 and @xmath73 replaced by effective quantities .
matter effects will modify neutrino oscillations ( and our ability to determine the mass hierarchy ) in two ways : * @xmath205 , so that the difference between the effective 13 and 23 oscillation frequencies will be different than @xmath37 . in the case of vacuum oscillations , this difference was recognized as the quantity that determines how precisely one needs to measure @xmath116 effects in order to establish the mass hierarchy ; * @xmath206 , so that the effective solar angle is different from its vacuum counterpart . in the case of vacuum oscillations
, @xmath207 measures the ratio of the amplitudes associated with 13 and 23 oscillations , and plays an important role in determining the value of @xmath92 which renders @xmath208 for a fixed @xmath91 .
two limiting cases are particularly easy to describe .
one is the limit of strong matter effects : @xmath209 . under these circumstances , @xmath210 , and @xmath211 . hence , @xmath212 in which case @xmath213 renders @xmath208
this can be understood in the following way . in the case of very strong matter effects and neutrino ( antineutrino ) oscillations , @xmath214 ( @xmath215 ) , so that @xmath216 ( @xmath217 ) oscillations are governed by only one oscillation frequency @xcite .. hence , high energy ( @xmath218 few gev ) @xmath216 oscillations inside the earth probe neither @xmath6 nor @xmath118 , but the solar - mixing weighted average of the two .
this agrees with intuition obtained from the trivial limits @xmath219 or @xmath220 , when @xmath216 oscillations can depend only on @xmath118 or @xmath6 , respectively . ] under theses circumstances , it is impossible to tell one mass hierarchy from the other .
another easy - to - describe limiting case is @xmath221 .
under these circumstances , we can use the results of sec .
[ sec : vacuum ] to determine the value of @xmath92 that renders @xmath208 without too much effort .
we start by computing the auxiliary quantity @xmath222 , defined by @xmath223 ( @xmath224 ) , which renders @xmath225 .
the definition of @xmath226 is identical to the one for @xmath92 ( see sec . [
sec : vacuum ] ) after all parameters are replaced by effective ones .
while @xmath226 has no simple physical interpretation , it can be related to @xmath92 in a simple way and we can easily solve for it . because the expression for @xmath17 in matter is the same as the one in vacuum with the vacuum mixing parameters replaced by effective ones , we can simply read off the solution from eq .
( [ eq : xaa ] ) : @xmath227 . we can relate @xmath226 to @xmath92 : @xmath228 where in the last step we simply repeat the definition of @xmath92 .
hence , @xmath229 and , since @xmath230 , @xmath213 .
this result agrees with the vacuum value of @xmath92 obtained in the case @xmath122 , as expected . as is well known , for two flavor oscillations @xmath231 in the limit @xmath232 and it also turns out that here , in the limit @xmath221 , @xmath231 for all @xmath139 . the two limiting cases discussed above reveal that , in order to break the normal inverted degeneracy in @xmath216 oscillations , it is necessary to probe @xmath17 for @xmath27 and @xmath26 values such that @xmath194 , independent of whether there are matter effects .
numerically , @xmath233 furthermore , one needs to steer clear of very strong matter effects , which translates into @xmath148 few gev for earth - based sources and detectors and very long baselines .
similar neutrino energies and baselines have , of course , also been identified as the optimal conditions to study matter effects in the 12 sector @xcite .
we proceed to discuss how to determine the mass hierarchy following the strategy discussed in the previous subsection .
we compute the value of @xmath92 that renders @xmath208 at a fixed value of @xmath91 . following our previous discussion , it is straightforward to obtain @xmath234 using eqs .
( [ eq : x(l / e ) ] ) and ( [ xm_to_x ] ) .
figure [ fig : xmatter ] depicts @xmath92 values that render @xmath208 in matter as a function of @xmath91 for different baselines , assuming @xmath235 ev@xmath132 and @xmath236 .
we choose matter densities equal to the average matter density value along the neutrino path , assuming that the neutrinos are produced and detected near the earth s surface , and using the preliminary earth model @xcite .
we extend all curves to very large values of @xmath91 , even if these correspond to experimentally unexaminable values of the neutrino energy ( @xmath237 mev ) . in order to avoid an overly busy plot
, we only display a few values of @xmath183 ( see eq .
( [ eq : x(l , e ) ] ) ) for each @xmath27 and @xmath26 .
as advertised , significant effects are expected for large values of @xmath27 ( @xmath238 km ) and small energy values ( @xmath239 km / gev ) . in this region , not only is @xmath240 , but also @xmath241 , as long as @xmath91 is not too large .
very large @xmath91 values correspond to very small energy values and @xmath242 , in which case matter effects become unobservable .
this behavior can be seen in fig .
[ fig : xmatter ] for @xmath243 km and @xmath244 km / gev .
this regime can not be observed in practice , as @xmath242 implies @xmath245 mev for typical earth matter densities .
[ fig : xmatter ] reveals several distinct ways of determining the mass hierarchy . as briefly discussed in the previous subsection , a `` broad - band '' experiment with , say ,
a 6000 km baseline capable of precisely measuring the atmospheric mass - squared difference for energy values around , say , @xmath246 gev ( @xmath247 km / gev ) and @xmath248 mev ( @xmath249 km / gev ) should be able to determine the mass hierarchy as long as @xmath250 . by `` measuring the atmospheric mass - squared difference at some particular energy value , '' we mean the following . select a narrow window around a particular @xmath91 value ( say @xmath251 ) .
since we need values of @xmath252 , we should find at least a few atmospheric oscillations inside our @xmath91 window .
numerically , @xmath253 and it is clear that a very big challenge for `` broad - band '' neutrino experiments is to make sure that atmospheric effects do not `` average out '' one needs to be able to `` see '' not the first and second oscillation maxima , but the tenth and the eleventh ! this implies , among other things , excellent energy resolution ( and angular resolution in the case of atmospheric neutrinos ) , and precise knowledge of the neutrino beam .
a detailed quantitative study of the experimental requirements is clearly outside the scope and philosophy of this paper .
one can also determine the mass hierarchy by performing two different experiments at the same value of @xmath91 but different @xmath27 .
for example , a 3000 km baseline experiment performed at @xmath91 around @xmath254 would find @xmath255 ( among other values ) , while a different experiment performed around the same value of @xmath91 but , say , @xmath256 km , would encounter @xmath257 . if both experiments can determine the atmospheric mass - squared difference with an uncertainty smaller than @xmath258 , they should be able to unveil whether the neutrino mass hierarchy is normal or inverted . in the previous subsections , we described how different @xmath16 disappearance searches can reveal the neutrino mass ordering , including the fact that neutrinos will often traverse a significant amount of matter on their way to the detector .
when this is the case , it is well known that @xmath17 will differ from @xmath260 because matter effects violate cp - invariance , there are no intrinsic cp - invariance violating effects in neutrino@xmath4neutrino oscillations . ] ( and cpt - invariance ) .
it seems , therefore , plausible that one should be able to determine the mass hierarchy by comparing muon neutrino and muon antineutrino disappearance .
we expect @xmath261 only for @xmath194 and @xmath262 given that ( i ) in the limit @xmath242 or @xmath122 , @xmath263 , ( ii ) in the limit @xmath264 , @xmath265 , both given by eq .
( [ pmumu_largea ] ) .
all of these features can be observed in fig .
[ fig : nu_anti_nu](top ) , which depicts the values of @xmath92 that render @xmath266 and those that render @xmath267 as a function of @xmath91 , for @xmath268 km , @xmath235 ev@xmath132 , @xmath236 , and @xmath269 g/@xmath203 . fig .
[ fig : nu_anti_nu](bottom ) reveals that the comparison of results obtained with a muon neutrino and antineutrino beam performed at @xmath268 km and , say , @xmath270 mev ( @xmath271 km / gev ) will determine the neutrino mass - hierarchy as long as @xmath272 .
one of the challenges of next - generation neutrino experiments is to establish the neutrino mass hierarchy : is the relatively isolated @xmath273 state ( much ) heavier than @xmath60 and @xmath61 , or is it ( much ) lighter ( see fig .
[ normal_inverted ] ) ? while we currently do not know the answer to this question , we do know that these two possibilities a normal or an inverted mass hierarchy are dramatically different . in the case of a normal hierarchy ,
the neutrino masses can be ordered just like the quark and charged lepton masses @xmath274 . in the case of an inverted mass hierarchy ,
the neutrino masses are ordered in a most unusual way @xmath275 and @xmath276 , _ i.e. _ , two of the neutrino mass eigenstates are quasi - degenerate .
we are yet to encounter two fundamental particles with quasi - degenerate masses .
it is not surprising , therefore , that the determination of the neutrino mass hierarchy has been identified as one of the key discriminators of neutrino mass models @xcite . in the literature , studies of matter
affected 13 oscillations have been singled out as the most realistic means of establishing the character of the neutrino mass hierarchy .
it is interesting that these studies do not rely on properties unique to three ( or more ) neutrino flavor oscillations , and significantly mirror the ( successful ) path taken in order to determine the `` solar '' neutrino mass hierarchy .
such studies , briefly described in sec . [ sec : traditional ] , require @xmath3 to be large , and utterly fail if the limit @xmath169 applies . in this paper
, we have concentrated on probes of the neutrino mass hierarchy that do not rely on neutrino oscillations in matter or on @xmath3 being `` large . '' instead , they rely heavily on the fact that , given that there are at least three neutrinos , @xmath142 will depend on at least two different oscillation frequencies . first , we show that , in principle , vacuum oscillations are capable of distinguishing an inverted from a normal mass hierarchy for a given normal mass hierarchy , there is no inverted mass hierarchy capable of producing the same @xmath277 .
we also argue that only neutrino oscillation experiments that are simultaneously sensitive to @xmath6 and @xmath19 effects can `` see '' the mass hierarchy .
hence , we need experiments capable of measuring @xmath24 with precision @xmath196 smaller than @xmath19 .
second , we argue that experiments performed at @xmath91 values such that @xmath278 suffer from a normal inverted degeneracy , which can only be broken if more experimental information is available .
the same is true of `` narrow - band - beam '' experiments which probe `` around '' a fixed value of @xmath27 and @xmath26 , regardless of the magnitude of @xmath119 .
if @xmath279 , the neutrino mass hierarchy can be unveiled by combining searches for @xmath16 disappearance with searches for @xmath82 ( or @xmath59 ) disappearance , as long as both experiments are capable of measuring @xmath24 with precision @xmath196 smaller than @xmath19 .
it seems unlikely that this precise a measurement of @xmath6 via @xmath82 disappearance can be performed by reactor neutrino experiments .
the kamland reactor experiment has measured @xmath19 with precision @xmath280 , which we take as an optimistic estimate of the precision with which a next generation reactor experiment can measure @xmath6 if @xmath281 . in the limit @xmath169 , any information regarding the neutrino mass hierarchy ( if one sticks to neutrino oscillation probes ) has to come from searches of @xmath16 or @xmath282 disappearance at distinct baselines and neutrino energies .
if these experiments are to be performed on the earth , at least one of them will require a very long baseline ( @xmath239 km / gev ) and relatively small neutrino energies ( @xmath283 gev ) in order to guarantee that @xmath194 and to avoid very strong matter effects ( @xmath284 ) , respectively .
different potential scenarios include comparing @xmath16 disappearance searches at short @xmath285 and long ( @xmath179 ) baselines , comparing searches for @xmath16 disappearance at different @xmath27 but the same value of @xmath91 , or comparing @xmath16 and @xmath282 disappearance in the presence of nontrivial matter effects ( @xmath286 ) .
note that `` broad - band '' accelerator experiments or studies of atmospheric neutrinos are potentially capable of performing all of these comparisons without the aid of other experimental setups one can interpret a broad - band / atmospheric experiment as a collection of multiple experiments at different values of @xmath91 . instead of concentrating on the feasibility of any of the setups summarized above , we chose to identify the conditions that need to be met in order for one to successfully distinguish an inverted mass hierarchy from a normal one .
we find that , irrespective of the scenario , we require multiple experiments capable of measuring @xmath24 at the few percent level ( @xmath287 , taking three sigma bounds for mass - squared differences @xcite ) . whether this can be achieved in practice will be left for detailed experimental studies .
we would , however , like to summarize the progress that can be expected in the foreseeable future . today , @xmath288 @xcite .
in the near future , the minos experiment expects to measure the atmospheric mass - squared difference with @xmath289 @xcite , while next - generation accelerator experiments aim at @xmath290 @xcite . in order to achieve such precise numbers ,
it is clear that improved knowledge of the neutrino beams and the neutrino
target cross sections are needed .
we wish to conclude by reemphasizing that neutrino oscillation experiments can , in principle , address a most relevant and fundamental question in high energy physics what is the character of the neutrino mass hierarchy ? this capability does not depend on whether the oscillations occur in matter or on whether @xmath3 is nonzero .
if , however , @xmath3 is sufficiently small , our study here shows that unveiling the mass hierarchy through oscillations may prove very hard .
detailed studies are required in order to determine how hard .
adg and bk are grateful to hamish robertson for a conversation that ultimately lead to this paper , and to stephen parke for enlightening discussions .
bk would like to thank gabriela barenboim for the hospitality of the universitat de valncia , where part of this work was completed .
fermilab is supported by the us department of energy contract de - ac02 - 76cho3000 .
the work of adg is sponsored in part by the us department of energy contract de - fg02 - 91er40684 .
simple analytic expressions for constant matter density neutrino oscillation probabilities can be obtained in the case @xmath149 .
these are presented in this appendix .
other very useful approximate expressions , including @xmath291 corrections to the results derived below , can be found , for example , in @xcite .
neutrino oscillations in matter are described by the following set of differential equations : @xmath292 where @xmath293 is a vector of neutrino flavor eigenstates and @xmath294 the @xmath295 parameters contain all information regarding the neutrino mass - squared differences , and describe the energy dependence of the system .
the forward scattering of electron neutrinos in matter is accounted for by @xmath296 , where @xmath297 is the fermi coupling constant , and @xmath298 is the electron number density of the medium . as a function of the mass density @xmath299 , @xmath40 can be expressed numerically as @xmath300 after we fix the electron to baryon fraction @xmath301 .
@xmath302 is a unitary mixing matrix . in the standard parametrization @xcite with @xmath1
set to zero , and ignoring potential majorana phases , which are not relevant here , @xmath303 where @xmath304 are column vectors . since @xmath302 is unitary , @xmath305 , which means that @xmath304 form an orthonormal basis for the space of three component vectors . the solutions to the three coupled differential equations in eq .
( [ diff_equations ] ) can be expressed in terms of the eigenvalues @xmath306 and eigenvectors @xmath307 of * m * , @xmath8 .
even though * m * is a three by three matrix , @xmath306 and @xmath308 may be found very easily . * m * , when @xmath302 is given by eq .
( [ eq : mns ] ) , may be expressed in the following convenient form : @xmath309 it is clear that one of the eigenvectors of @xmath310 is @xmath311 : @xmath312 .
furthermore , since @xmath310 is hermitian , the other two eigenvectors are orthogonal to @xmath311 , which means they can be expressed as orthogonal linear combinations of @xmath313 and @xmath314 .
the problem of finding the three eigenvalues and eigenvectors of a three by three matrix is now reduced to finding two eigenvectors in a two by two effective matrix , which is very easily solved . indeed , the problem reduces to the well known case of two - flavor neutrino oscillations in constant matter , and its solution is very well known @xcite .
hence , the resulting oscillation probabilities are conveniently expressed in terms of effective mixing angles and frequencies , which are simply related to their vacuum counterparts .
@xmath315 @xmath316 , and @xmath317 reduces to @xmath37 in the vacuum limit @xmath242 .
the oscillation probabilities coincide exactly with the corresponding vacuum expressions once the vacuum oscillation parameters are replaced by the corresponding effective ones .
we explicitly list all oscillation probabilities in constant matter below , keeping in mind that , due to t - invariance , , there are no t - invariance violating effects in neutrino oscillations in vacuum . furthermore , a constant matter potential does not induce t - invariance violating effects in neutrino oscillations in matter .
matter - induced cp - invariance violation is , of course , present . ]
@xmath318 , @xmath139 .
the oscillation probabilities of antineutrinos are given by the same expressions , with @xmath319 .
@xmath320 as expected , @xmath321 ( @xmath135 ) does not depend on @xmath6 , and hence does not care whether the neutrino mass hierarchy is normal or inverted .
this is not the case of @xmath322 , which are given by @xmath323 the terms of @xmath322 that depend on @xmath6 are identical , up to an overall sign .
hence , @xmath324 is the same for all non - electron oscillation probabilities .
any normal inverted degeneracy observed in a study of @xmath159 will be present in exactly the same form in a similar study of @xmath162 or @xmath325 .
a. aguilar _ et al .
_ [ lsnd collaboration ] , phys .
d * 64 * , 112007 ( 2001 ) .
s. eidelman _ et al . _
[ particle data group collaboration ] , phys .
b * 592 * , 1 ( 2004 ) . for recent general overviews of theoretical developments in neutrino physics
see , for example , r.n .
mohapatra _ et al .
_ , hep - ph/0412099 ; a. de gouva , mod . phys . lett . a * 19 * ( 2004 ) 2799 .
albright , phys .
b * 599 * , 285 ( 2004 ) .
see , for example , updated version of m. maltoni , t. schwetz , m.a .
tortola and j.w.f .
valle , new j. phys .
* 6 * , 122 ( 2004 ) , hep - ph/0405172 , and many references therein . for recent studies
see , for example , r. gandhi , p. ghoshal , s. goswami , p. mehta and s. u. sankar , hep - ph/0411252 ; o. mena and s. parke , phys .
d * 70 * , 093011 ( 2004 ) ; p. huber , m. lindner , m. rolinec , t. schwetz and w. winter , phys .
d * 70 * , 073014 ( 2004 ) ; h. minakata , h. nunokawa and s.j .
parke , phys .
d * 68 * , 013010 ( 2003 ) .
m. apollonio _ et al .
_ , hep - ph/0210192 ; and many references therein . for recent studies see , for example , p. huber , m. maltoni and t. schwetz , hep - ph/0501037 ; d. indumathi and m. v. n. murthy , phys . rev .
d * 71 * , 013001 ( 2005 ) ; j. bernabeu , s. palomares ruiz and s.t .
petcov , nucl .
b * 669 * , 255 ( 2003 ) ; and references therein . for recent studies
see , for example , v. barger , p. huber and d. marfatia , hep - ph/0501184 ; m. kachelrie and r. toms , hep - ph/0412100 ; a.s .
dighe , m.t . keil and g.g .
raffelt , jcap * 0306 * , 005 ( 2003 ) ; c. lunardini and a. yu . smirnov , jcap * 0306 * , 009 ( 2003 ) ; and many references therein .
g. barenboim and a. de gouva , hep - ph/0209117 .
a. de gouva , a. friedland and h. murayama , phys .
b * 490 * , 125 ( 2000 ) .
i. ambats _ et al . _
[ no@xmath326a collaboration ] , fermilab - proposal-0929 ( see also ` http://www-nova.fnal.gov/ ` ) ; d. ayres _
[ no@xmath326a collaboration ] , hep - ex/0210005 ; m. v. diwan _ et al .
_ , phys .
d * 68 * , 012002 ( 2003 ) ( see also ` http://www.neutrino.bnl.gov/ ` ) ; y. itow _ et al .
_ , hep - ex/0106019 ( see also ` http://neutrino.kek.jp/jhfnu/ ` ) ; k. anderson _
et al . _ , hep - ex/0402041 .
( see also ` http://www.hep.anl.gov/ndk/hypertext/reactor_neutrino.html ` ) ; s.t .
petcov and m. piai , phys .
b * 533 * , 94 ( 2002 ) .
s. choubey , s.t .
petcov and m. piai , phys .
d * 68 * , 113006 ( 2003 ) .
p. zucchelli , phys .
b * 532 * ( 2002 ) 166 ; m. mezzetto , j. phys .
g * 29 * , 1771 ( 2003 ) ; j. burguet - castell , d. casper , j.j .
gomez - cadenas , p. hernandez and f. sanchez , nucl .
b * 695 * , 217 ( 2004 ) .
h. nunokawa , s.j .
parke , and r. zukanovich - funchal , in preparation . | is the relatively isolated member of the neutrino mass spectrum heavier or lighter than the two closely - spaced members ?
this question the character of the neutrino mass hierarchy is of great theoretical interest .
all previously identified experiments for addressing it via neutrino oscillations require that the currently unknown size of the @xmath0 element of the leptonic mixing matrix ( parameterized by the unknown @xmath1 mixing angle ) be sufficiently large , and will utterly fail in the limit @xmath2 .
for this reason , we explore alternative oscillation approaches that would still succeed even if @xmath1 vanishes .
we identify several alternatives that require neither a nonzero @xmath3 nor the presence of significant matter effects .
all include multiple percent - level neutrino oscillation measurements , usually involving muon - neutrino ( or antineutrino ) disappearance and very long baselines .
we comment on the degree of promise that these alternative approaches show . |
black holes ( bhs ) in einstein - maxwell theory are characterized by three parameters @xcite : mass @xmath2 , electric charge @xmath3 and angular momentum @xmath4 .
bhs are thought to be abundant objects in the universe .
their mass is estimated to vary between @xmath5 and @xmath6 or higher @xcite , their electrical charge is negligible because of the effect of surrounding plasma @xcite and their angular momentum is expected to be close to the extremal limit because of accretion and merger events @xcite .
a non - comprehensive list of some astrophysical bh candidates @xcite is shown in table [ tab : data ] . despite the wealth of circumstantial evidence
, there is no definite observational proof of the existence of astrophysical bhs due to the difficulty to detect an event horizon in astrophysical bh candidates @xcite .
thus astrophysical objects without event horizon , yet observationally indistinguishable from bhs , can not be excluded a priori .
some of the most viable alternative models describing an ultra - compact astrophysical object include gravastars , boson stars , wormholes and superspinars .
+ dark energy stars or _ gravastars _ are compact objects with de sitter interior and schwarzschild exterior @xcite .
these two regions are glued together by a model - dependent intermediate region . in the original model
@xcite the intermediate region is an ultra - stiff thin shell .
models without shells or discontinuities have also been investigated @xcite .
+ _ boson stars _ are macroscopic quantum states which are prevented from undergoing complete gravitational collapse by heisenberg uncertainty principle @xcite .
their models differ in the scalar self - interaction potential which also set the allowed maximum compactness for a boson star .
+ an exhaustive description of _ wormholes _ can be found in the monograph @xcite ( see also ref .
@xcite ) . in this work
we shall consider particular wormholes which are infinitesimal variations of bh spacetimes .
these wormholes may be indistinguishable from ordinary bhs @xcite .
+ _ superspinars _ are solutions of the gravitational field equations that violate the kerr bound . these geometries could be created by high energy corrections to einstein gravity such as those present in string - inspired models @xcite .
.mass , @xmath2 , radius , @xmath7 , angular momentum , @xmath8 , and compactness , @xmath9 , for some bh candidates ( from @xcite ) .
mass and radius are in solar units . [ cols="^,^,^,^,^",options="header " , ] objects with @xmath10 could potentially describe superspinars .
several arguments suggest that objects rotating above the kerr bound are unstable .
firstly , extremal kerr bhs are marginally stable .
thus the addition of extra rotation should lead to instability .
secondly , fast - spinning objects usually take a pancake - like form @xcite and are subject to the gregory - laflamme instability @xcite .
finally , kerr - like geometries , like naked singularities , seem to be unstable against a certain class of gravitational perturbations @xcite called algebraically special perturbations @xcite . for objects with
@xmath10 the surface or mirror can be placed anywhere outside @xmath11 . in general
the instability is as strong as in the @xmath12 regime .
an example in shown in fig .
[ fig : super ] for the surface at @xmath13 .
this result confirms other investigations suggesting that ultra - compact objects rotating above the kerr bound are unstable @xcite .
we investigated the ergoregion instability of some ultra - compact , horizonless objects which can mimick the spacetime of a rotating black hole .
we studied some of the most viable bh mimickers : gravastars , boson stars , wormhole and superspinars .
if rotating , boson stars and gravastars may develop ergoregion instabilities .
analytical and numerical results indicate that these objects are unstable against scalar field perturbations for a large range of the parameters .
slowly rotating gravastars can develop an ergoregion depending on their angular momentum , their compactness and the thickness of their intermediate region .
in a recent work @xcite it has pointed out that slowly rotating gravastars may not develop an ergoregion . in the formation of the ergoregion for rotating gravastars an important role
is played by the thickness ( see figure [ fig : delta ] ) which is not easily detectable .
thus further investigations are needed to better understand the ergoregion formation in physical resonable gravastar models .
the instability timescale for both boson stars and gravastars can be many orders of magnitude stronger than the instability timescale for ordinary stars with uniform density . in the large @xmath14 approximation , suitable for a wkb treatment , gravitational and scalar perturbations have similar instability timescales . in the low-@xmath15 regime
gravitational perturbations are expected to have even shorter instability timescales than scalar perturbations .
instability timescales can be as low as @xmath16 seconds for a @xmath17 objects and about a week for supermassive bhs , @xmath18 , monotonically decreasing for larger rotations and a larger compactness .
the essential features of wormholes and superspinars have been captured by a simple model whose physical properties are largely independent from the dynamical details of the gravitational system .
numerical and analytic results show that the ergoregion instability of these objects is extremely strong for any value of their angular momentum , with timescales of order @xmath19 seconds for a @xmath20 object and @xmath21 seconds for a @xmath18 object .
therefore , high rotation is an indirect evidence for horizons .
although further studies are needed , the above investigation suggests that exotic objects without event horizon are likely to be ruled out as viable candidates for astrophysical ultra - compact objects .
this strengthens the role of bhs as candidates for astrophysical observations of rapidly spinning compact objects .
the authors warmly thank matteo losito for interesting discussions and for sharing some of his results .
this work is supported by fundao para a cincia e tecnologia ( fct ) - portugal through project ptdc / fis/64175/2006 and by the national science foundation through ligo research support grant nsf phy-0757937 .
r. d. blandford and r. l. znajek , mon . not .
soc . * 179 * , 433 ( 1977 ) . c. f. gammie , s. l. shapiro and j. c. mckinney , astrophys .
j. * 602 * , 312 ( 2004 ) .
j. m. miller , a. c. fabian , m. a. nowak and w. h. g. lewin , arxiv : astro - ph/0402101 .
r. narayan , j. e. mcclintock and r. shafee , aip conf .
* 968 * ( 2008 ) 265 j. m. wang , y. m. chen , l. c. ho and r. j. mclure , astrophys .
j. * 642 * , l111 ( 2006 ) . m. a. abramowicz , w. kluzniak and j. p. lasota , astron .
astrophys . * 396 * , l31 ( 2002 ) .
o. mazur and e. mottola , `` gravitational condensate stars , '' arxiv : gr - qc/0109035 . c. b. m. chirenti and l. rezzolla , class .
* 24 * , 4191 ( 2007 ) . c. cattoen ,
t. faber and m. visser , class .
* 22 * , 4189 ( 2005 ) .
d. j. kaup , phys .
rev . * 172 * , 1331 ( 1968 ) ; r. ruffini and s. bonazzola , phys . rev .
* 187 * , 1767 ( 1969 ) . j. p. s. lemos , f. s. n. lobo and s. quinet de oliveira , phys .
d * 68 * , 064004 ( 2003 ) .
t. damour and s. n. solodukhin , phys .
d * 76 * , 024016 ( 2007 ) .
e. g. gimon and p. horava , arxiv:0706.2873 [ hep - th ] . s. w. hawking , phys .
d * 72 * ( 2005 ) 084013 [ arxiv : hep - th/0507171 ] .
r. narayan , i. yi and r. mahadevan , nature * 374 * , 623 ( 1995 ) .
a. e. broderick and r. narayan , class .
* 24 * , 659 ( 2007 ) .
m. vallisneri , phys .
lett . * 84 * , 3519 ( 2000 ) .
j. p. s. lemos and o. b. zaslavskii , phys .
d * 76 * ( 2007 ) 084030 [ arxiv:0707.1094 [ gr - qc ] ] .
j. p. s. lemos and o. b. zaslavskii , phys .
d * 78 * ( 2008 ) 024040 [ arxiv:0806.0845 [ gr - qc ] ] .
v. cardoso , p. pani , m. cadoni and m. cavaglia , phys .
d * 77 * ( 2008 ) 124044 v. cardoso , p. pani , m. cadoni and m. cavaglia , class .
* 25 * ( 2008 ) 195010 j. l. friedman , commun .
phys . * 63 * , 243 ( 1978 ) .
v. cardoso , o. j. c. dias , j. p. s. lemos and s. yoshida , phys .
d * 70 * , 044039 ( 2004 ) [ erratum - ibid .
d * 70 * , 049903 ( 2004 ) ] ; v. cardoso and o. j. c. dias , rev .
d * 70 * , 084011 ( 2004 ) ; e. berti , v. cardoso and j. p. s. lemos , phys .
d * 70 * , 124006 ( 2004 ) .
b. f. whiting , j. math .
phys . * 30 * ( 1989 ) 1301 .
n. comins and b. f. schutz , proc .
r. soc .
a * 364 * , 211 ( 1978 ) .
b. kleihaus , j. kunz and m. list , phys . rev .
d * 72 * , 064002 ( 2005 ) ; b. kleihaus , j. kunz , m. list and i. schaffer , phys .
d * 77 * ( 2008 ) 064025 [ arxiv:0712.3742 [ gr - qc ] ] .
j. b. hartle , astrophys .
j. * 150 * , 1005 ( 1967 ) ; j. b. hartle and k. s. thorne , astrophys .
j. * 153 * , 807 ( 1968 ) . b. f. schutz and n. comins , mnras * 182 * , 69 ( 1978 ) .
c. b. m. chirenti and l. rezzolla , phys .
d * 78 * ( 2008 ) 084011 [ arxiv:0808.4080 [ gr - qc ] ] .
f. e. schunck and e. w. mielke , in _ relativity and scientific computing _
, edited by f. w. hehl , r. a. puntigam and h. ruder ( springer , berlin , 1996 ) , p. 138 .
v. cardoso , o. j. c. dias , j. l. hovdebo and r. c. myers , phys .
d * 73 * , 064031 ( 2006 ) .
s. chandrasekhar , proc .
r. soc .
london , ser .
a * 392 * , 1 ( 1984 ) ; _ the mathematical theory of black holes _ , ( oxford university press , new york , 1983 ) .
s. a. teukolsky , phys .
lett * 29 * , 1114 ( 1972 ) ; s. a. teukolsky , astrophys .
j. * 185 * , 635 ( 1973 ) .
e. berti , v. cardoso and m. casals , phys .
d * 73 * , 024013 ( 2006 ) [ erratum - ibid .
d * 73 * , 109902 ( 2006 ) .
r. emparan and r. c. myers , jhep * 0309 * , 025 ( 2003 ) .
r. gregory and r. laflamme , phys .
lett . * 70 * , 2837 ( 1993 ) . v. cardoso and o. j. c. dias , phys
. rev .
lett . * 96 * , 181601 ( 2006 ) ; v. cardoso and l. gualtieri , class .
* 23 * , 7151 ( 2006 ) ; v. cardoso , o. j. c. dias and l. gualtieri , arxiv:0705.2777 [ hep - th ] .
v. cardoso and m. cavagli , phys .
d * 74 * , 024027 ( 2006 ) .
g. dotti , r. gleiser and j. pullin , phys .
b * 644 * , 289 ( 2007 ) .
g. dotti , r. j. gleiser , i. f. ranea - sandoval and h. vucetich , arxiv:0805.4306 [ gr - qc ] . | ultra - compact , horizonless objects such as gravastars , boson stars , wormholes and superspinars can mimick most of the properties of black holes . here
we show that these `` black hole mimickers '' will most likely develop a strong ergoregion instability when rapidly spinning .
instability timescales range between @xmath0s and @xmath1 weeks depending on the object , its mass and its angular momentum . for a wide range of parameters
the instability is truly effective .
this provides a strong indication that astrophysical ultra - compact objects with large rotation are black holes . |
layered transition metal oxides present a paradigm for strongly correlated electronic systems subject to interesting correlated states and the competition of multiple order parameters .
the high temperature oxide superconductors are a much - studied case in point .
moreover the discovery of superconductivity in sodium cobalt oxide @xcite confirms this general approach of exploring novel layered metal oxides .
here we consider layered tungstates as another new model system capable of displaying a complex phase behaviour dependent upon interlayer coupling and doping state .
tungsten oxide has been the subject of much research in the past few decades due to its interesting physical and electronic properties and doping capability .
tungsten trioxide can form a variety of stable structures at room temperature , such as pyrochlore , hexagonal , or distorted cubic @xcite , all of which are comprised of corner- and/or edge - shared wo@xmath5 octahedra .
these structures can be doped by inserting mono- or di - valent atomic species into the interstitial vacancy sites within the oxide structure @xcite , or by removing oxygen @xcite .
this results in an increase in electrical conductivity @xcite and dramatic colour changes @xcite , which has led to tungsten trioxide systems being used in electrochromic applications @xcite .
tungsten trioxide can be hydrated to form layered structures , with a general formula of wo@xmath1@xmath6@xmath7h@xmath3o . in the mono - hydrate , wo@xmath1@xmath6h@xmath3o ( or h@xmath3wo@xmath4 )
, two - dimensional layers of corner - shared wo@xmath5 octahedra are formed , with a water molecule ( @xmath0oh@xmath3 ) substituted for one apical oxygen of the tungsten atom , and a terminal oxygen completing the structure @xcite . in the di - hydrate ( wo@xmath1@xmath62h@xmath3o )
, the second water molecule is inserted between the layers , as in moo@xmath1@xmath62h@xmath3o @xcite .
an increase in the interlayer spacing is observed @xcite and it thus follows that one could perhaps substitute organic species between the tungsten oxide layers .
indeed this has proven successful , with various groups succeeding in intercalating organic molecules such as pyrazine @xcite , 4,4-bipyridine @xcite , pyridine @xcite , tert - alkylammonium species @xcite , mono - aminoalkanes @xcite and diamino - alkanes @xcite .
such materials provide an interesting template for doping and substitutional ( e.g. magnetic ions ) studies in low - dimensional systems . here
we present infrared and raman results on a selection of wo@xmath1-based hybrid materials in order to gain a more extensive understanding of the overall structure ; in particular , the bonding nature of the organic to the inorganic layer , and the impact of the organic intercalant on the structure of the inorganic layer .
diaminoalkane - tungsten oxide samples were synthesised as described in ref . .
in brief , h@xmath3wo@xmath4 was dissolved in hot aqueous ammonia solution ( 33 wt .
@xmath8 ) and then a solution of the organic amines ( @xmath9,@xmath10-diamine , h@xmath3n(ch@xmath3)@xmath11nh@xmath3 , hereby abbreviated to dan ; or phenethylamine , c@xmath5h@xmath12(ch@xmath3)@xmath3nh@xmath3 , abbreviated to phen ) dissolved in ammonia was added in a 2:1 molar ratio .
the excess solvent was then evaporated off and the hybrid materials were obtained as white / cream - coloured powders .
the entire synthesis was carried out under flowing n@xmath3 gas .
powder x - ray diffraction spectra were recorded using a philips pw1700 series powder diffractometer with co k@xmath9 radiation .
infrared spectra were collected on a bomem da8 ft spectrometer over the mid - ir range ( 450 - 4000 @xmath13 ) using the kbr disc method with a resolution of 2 @xmath13 .
raman spectra were collected using a jobin - yvon labram hr spectrometer with an excitation wavelength of 633 nm .
the x - ray diffraction spectra shown in figure [ fig : xrd]a display a strong peak at low 2@xmath14 angles . as the length of the intercalated organic molecule is increased , the corresponding d - spacing also increases in a linear fashion ( shown in figure [ fig : xrd]b ) . from the slope and intercept of this line we propose a model by which single molecular layers of 2-dimensional corner - shared wo@xmath5 octahedra are separated by the organic molecules which are aligned almost perpendicular to the tungsten oxide layers .
one would expect such a layered compound to display properties approaching those of a purely 2-dimensional system as the distance between the oxide layers is increased ( by increasing the length of the organic spacer molecule ) .
the physical properties of this system will be the topic of future publications .
( a ) x - ray diffraction spectra of tungsten oxide - based hybrid materials : ( i ) w - da2 ; ( ii ) w - da4 ; ( iii ) w - da6 ; ( iv ) w - da8 ; ( v ) w - da12 .
( b ) linear progression of first major peak versus alkane chain length.,width=321,height=415 ] the normalised infrared and raman spectra of the solid samples studied are shown in figures [ fig : raman]-[fig : irzoom ] , and the peak positions tabulated in tables [ table1]-[table2 ] .
the diaminoalkane hybrids wo@xmath4@xmath6dan ( hereby abbreviated to w - dan ) with alkyl lengths longer than two carbons have virtually identical spectra @xcite and so are summarised in the tables as w - dan , in comparison with the differing w - da2 .
the infrared spectra of wo@xmath1 and its hydrate , h@xmath3wo@xmath4 , have much fewer and broader peaks than their hybrid counterparts .
both the infrared and raman spectra of these two compounds correspond well with those presented by daniel et al .
these results are summarised in three sections : the effect on the inorganic layer , the effect on the organic species , and the organic - inorganic bonding nature .
raman spectra of tungsten oxide - related materials : ( a ) wo@xmath1 ; ( b ) h@xmath3wo@xmath4 ; ( c ) w - da2 ; ( d ) w - da6 ; ( e ) w - da12 ; ( f ) w - phen .
the high frequency region is displayed on a different scale to enable the important features to be seen.,width=321,height=226 ] infrared spectra of tungsten oxide - related materials : ( a ) wo@xmath1 ; ( b ) h@xmath3wo@xmath4 ; ( c ) w - da2 ; ( d ) w - da6 ; ( e ) w - da12 ; ( f ) w - phen.,width=321,height=226 ] low - frequency region of the infrared spectra of tungsten oxide - related materials : ( a ) wo@xmath1 ; ( b ) h@xmath3wo@xmath4 ; ( c ) w - da2 ; ( d ) w - da6 ; ( e ) w - da12 ; ( f ) w - phen.,width=321,height=226 ] [ cols="<,^,^,^,^,^,^,^,^,^,^ " , ] the presence of co - ordinated water molecules in h@xmath3wo@xmath4 can be seen by the broad o@xmath0h stretching peak at 3410 ( 3370 ) @xmath13 ( ir and raman respectively ) , the h@xmath3o bending peak at 1614 @xmath13 ( ir only ) , and the w@xmath0oh@xmath3 co - ordinated water peak in the raman spectrum at 377 @xmath13 .
these peaks do not occur in the wo@xmath1 spectra , as expected . as mentioned earlier , the structure of h@xmath3wo@xmath4 consists of layers of corner - shared wo@xmath5 octahedra with alternate apical arrangements of w@xmath0oh@xmath3 and w = o . in the raman spectrum
the w = o bonding is shown clearly by a sharp band centred at about 950 @xmath13 .
this is also present in the ir spectrum .
surprisingly , the wo@xmath1 sample also exhibits a small peak at this position in the raman and ir spectra .
this is not expected as the structure of wo@xmath1 consists only of single w@xmath0o bonds .
however it can be explained by the presence of disorder in the sample and loss of oxygen @xcite powder is a pale yellow colour , which is reduced over 1 - 2 days in air to a pale green colour .
this is indicative of a loss of oxygen but can be regained by storing in an oxygen atmosphere or heating in oxygen for a few hours . ] , which results in the formation of a small fraction of w = o bonds .
there are also w = o terminations on the surfaces of the powder particles . in the diaminoalkane hybrid compounds
the characteristic w = o peak , formerly at 950 @xmath13 in h@xmath3wo@xmath4 , shifts to a lower frequency of 890 - 900 @xmath13 . while in several of these compounds there are peaks at 950 @xmath13 they are not as intense as the w = o peak .
it is also of interest to note that the peak at 888 @xmath13 in the w - da2 raman spectrum is a doublet , suggesting perhaps the presence of two different w = o bonds . in w - phen , the peak structure in the range 900 - 100
@xmath13 is quite different from the other hybrid spectra .
this may be due to the organic molecule being mono - dentate ( as ir spectra of other mono - aminoalkane hybrids show that the 950 @xmath13 w = o peak is virtually unaltered @xcite ) or more complicated due to the presence of the aromatic ring . in any case , the sharp raman peak at 1004 @xmath13 is likely to be a w = o peak due to the similarity in shape between it and the known w = o peak in h@xmath3wo@xmath4 , for example
. there may be more than one variation of the w = o bond within the structure , which may also account for the multiple peaks observed in this region , as is the case in wo@xmath1@xmath62h@xmath3o @xcite .
there is a relationship between bond length ( an indication of bond strength ) and the frequency of the vibrational modes(s ) for the terminal w@xmath0o bond , namely that as the bond length decreases , bond strength and frequency increases @xcite .
it is not therefore impossible to observe more than one terminal w@xmath0o vibration .
the remaining bands can be assigned as follows : 580 - 860 @xmath13 o@xmath0w@xmath0o stretching , 430 - 470 @xmath13 inorganic raman mode ( w@xmath0o ) , 260 - 350 @xmath13 o@xmath0w@xmath0o bending , 220 - 250 @xmath13 w@xmath0o@xmath0w stretching , 90 - 200 @xmath13 lattice modes .
these bands occur at similar positions in the hybrid compounds .
perhaps the most telling difference between w - da2 and w - dan ( n@xmath152 ) is the presence or absence of the broad peak centred at 2100 @xmath13 in the ir spectra .
this feature is due to a combination of the asymmetrical @xmath0nh@xmath1@xmath2 bending vibration and the torsional oscillation of the @xmath0nh@xmath1@xmath2 group @xcite .
both @xmath0nh@xmath3 scissoring and @xmath0nh@xmath1@xmath2 bending occurs in the region 1580 - 1630 @xmath13 , however the latter displays three peaks ( as seen in w - dan ) while the former yields only two ( as in w - da2 ) .
lastly , the n@xmath0h stretching bands occur at slightly higher energies in w - dan compared with w - da2 ( 3240 - 3300 @xmath13 , c.f . 3135 - 3250 @xmath13 ) and these peaks are also much sharper .
however , an increase in alkyl chain length should have little effect on the n@xmath0h stretching frequency @xcite .
the observed differences are discussed in section iii .
the presence of broad vibrational bands in the w - dan spectra in the range 2430 - 2790 @xmath13 can be ascribed to nh@xmath1@xmath2 symmetrical stretching @xcite and often appear as a broad , ill - defined band under the c@xmath0h stretching modes near 2800 @xmath13 @xcite . in h@xmath3wo@xmath4
a small peak is observed in the ir spectrum at 1614 @xmath13 , which corresponds to h@xmath3o bending .
while the hybrid samples also exhibit peaks near this value ( 1600 - 1630 @xmath13 ) , these peaks are sharp and well defined , and the vibrational band at 3400 @xmath13 ( indicative of o@xmath0h stretching ) is absent .
thus we conclude that most of the co - ordinated water molecules ( w@xmath0oh@xmath3 ) have been dehydrogenated during the organic intercalation process ( a small feature at 377 @xmath13 in the raman spectra may correspond to w@xmath0o@xmath0h ) .
the peaks in the hybrid spectra in the range 1600 - 1630 @xmath13 correspond instead to nh@xmath3 or nh@xmath1@xmath2 bending . as only the terminal amino / ammonium group
is involved in interactions with the inorganic layer , it is no surprise that the bands corresponding to c@xmath0h , c@xmath0n and c@xmath0c vibrations remain virtually unchanged .
the peaks in the hybrid spectra corresponding to these vibrations can be assigned as follows : 2850 - 2960 @xmath13 c@xmath0h stretching , 1300 - 1500 @xmath13 ch@xmath3 wagging , bending , and scissoring ; also c = c aromatic stretching for w - phen ( 1450 - 1500 @xmath13 ) , 950 - 1100 @xmath13 c@xmath0n and c@xmath0c stretching . from 460 - 550 @xmath13
there are a small number of unassigned organic bands ( including a sharp peak in the ir at 537 @xmath13 for w - da2 , corresponding to the bending vibration of the nccn backbone @xcite ) .
the most noticeable difference among the hybrid samples is that w - da2 shows quite a different bonding nature to the hybrids with longer organic chains : there is no evidence for the presence of ammonium ( @xmath0nh@xmath1@xmath2 ) terminal groups in w - da2 .
as mentioned previously , w - da2 lacks a feature at 2100 @xmath13 found in the other hybrid spectra that corresponds to deformations of terminal @xmath0nh@xmath1@xmath2 .
combined with the presence of the @xmath0nh@xmath1@xmath2 rocking mode found at 770 @xmath13 @xcite , this implies that some of the amine molecules in w - dan appear as r@xmath0nh@xmath1@xmath2@xmath16@xmath17o@xmath0w , while all those of w - da2 appear as neutral @xmath0nh@xmath3 species .
this raises the question of how charge balance is achieved in w - da2 and also whether hydrogen bonding is the exclusive mechanism of bonding , whereas in the longer organic molecule hybrids there is clearly an electrostatic component as well due to the presence of terminal ammonium groups .
there are also differences in the n@xmath0h stretch region , suggesting changes in the bonding chemistry as one progresses from da2 to longer dan alkyl chains .
as mentioned in section ii , the n@xmath0h stretching bands for w - da2 appear at lower wavenumbers than for longer - chain w - dan species . hydrogen bonding ( which holds the da2 molecules in place between the layers ) will cause these bands to shift to lower frequencies @xcite .
da2 is the only case where the inductive effect on the terminal amine groups is known to be a perturbing factor to the molecular vibrations due to the shorter alkyl chain length @xcite .
plot of pk@xmath18 values from ref .
versus the number of carbon atoms of @xmath9,@xmath10-diamines in aqueous solution.,width=321,height=226 ] one possible explanation for the observed difference in structure between da2 and longer dan molecules intercalated into an acidic metal oxide , arises from examining the trend in basicity of @xmath9,@xmath10-diamines in solution .
although the trend in basicity only changes slightly with increasing alkyl length for the monoamines ( due to combinations of solution effect , steric hindrance to solvation , and number of available h - bonds @xcite ) , a substantial increase in basicity is observed in going from da2 to da8 ( fig .
[ fig : pka ] ) .
this increase in base strength for the diaminoalkanes is due to an increase of the inductive effect as the number of methylene groups ( @xmath0ch@xmath3@xmath0 ) increases .
the formation of alkylammonium ( @xmath0nh@xmath1@xmath2 ) end groups will therefore be more favourable for the more basic diaminoalkanes ( i.e. diaminobutane and above ) .
this apparent difference is also manifested in the tga profile of these hybrids , with w - da2 dissociating at a higher temperature than any of the other w - dan compounds @xcite .
the strong raman peak at 950 @xmath13 in h@xmath3wo@xmath4 corresponds to an apical w = o bond , which is characteristic of a layered structure .
this peak is present in all of the hybrid compounds although it is shifted to lower wavenumbers of 890 - 900 @xmath13 , indicating that the apical oxygen is not as tightly bound to the tungsten as in h@xmath3wo@xmath4 .
this too can be expected as charge balance in h@xmath3wo@xmath4 is achieved by co - ordinated water molecules binding alternately upper and lower of the tungsten layer , leaving the remaining apical oxygen atoms to bind relatively strongly . in the hybrid materials one might expect the influence of organic molecules on each apical oxygen to be more uniform
, hence this interaction will weaken the w = o bond , and this is indeed evident from the raman shift .
there is evidence to suggest that in w - phen , a monodentate amine hybrid , there are several variations of the w = o bond , which in itself warrants further study .
the structure and electronic density of states of these layered systems are currently being investigated by @xmath19 computations .
the question of the relocation of protons in the w - da2 system is a particular issue of study .
infrared and raman spectroscopy have been utilised to gain an understanding of the bonding nature between the organic and inorganic components of tungsten oxide - hybrid systems with different length diaminoalkanes and how the inclusion of organic molecules affects the structure of the inorganic layers .
tungsten oxide - ethylenediamine displays a different bonding nature from the other hybrid materials in that it appears to rely on hydrogen - bonding only , with no terminal @xmath0nh@xmath1@xmath2 ammonium groups to form electrostatic bonds .
this is thought to be related to the lower basicity of ethylenediamine as opposed to the longer diaminoalkane species .
all of the tungsten oxide - organic hybrid materials exhibit a decrease in the strength of the apical w = o bond of the wo@xmath4@xmath20 layers compared with the layered tungsten oxide monohydrate ( tungstic acid ) , h@xmath3wo@xmath4 .
the authors would like to acknowledge the financial assistance from the new zealand foundation of research science and technology ( contract number : irlx0201 ) , the royal society of new zealand marsden fund , and the macdiarmid institute for advanced materials and nanotechnology ( victoria university , new zealand ) . | tungsten oxide - organic layered hybrid materials have been studied by infrared and raman spectroscopy , and demonstrate a difference in bonding nature as the length of the interlayer organic ` spacer ' molecule is increased . ethylenediamine - tungsten oxide clearly displays a lack of terminal @xmath0nh@xmath1@xmath2 ammonium groups which appear in hybrids with longer alkane molecules , thus indicating that the longer chains are bound by electrostatic interactions as well as or in place of the hydrogen bonding that must be present in the shorter chain ethylenediamine hybrids .
the presence of organic molecules between the tungsten oxide layers , compared with the layered tungstic acid h@xmath3wo@xmath4 , shows a decrease in the apical w = o bond strength , as might be expected from the aforementioned electrostatic interaction . |
to determine the value of @xmath11 , one need to first evaluate the energy of a pair of nearest fe - fe moments in parallel ( @xmath42 ) and anti - parallel ( @xmath43 ) alignments with respect to a non - magnetic reference state , respectively .
then from their difference , one can determine the value of @xmath11 by the following formula @xmath44 where @xmath45 is the spin of each fe ion .
it should be emphasized that @xmath42 is not necessary to be equal to @xmath46 since the energy of the reference state may not be located exactly at the middle of the energy between the ferromagnetic and antiferromagnetic states .
this energy lineup between the nonmagnetic and any other magnetic state needs self - consistent total energy calculations to determine , that is what we have done .
thus @xmath42 and @xmath43 should be determined independent .
similarly , @xmath12 can be determined from the difference between the energy of a pair of next - nearest fe - fe moments in the parallel ( @xmath47 ) and anti - parallel ( @xmath48 ) alignments : @xmath49 to determine the values of @xmath50 , @xmath51 , we have calculated the total energies of the ferromagnetic ( @xmath52 ) , square antiferromagnetic ( @xmath53 ) , and collinear antiferromagnetic ( @xmath54 ) states , respectively .
the spin configurations of these three states are shown in fig .
[ config ] .
the corresponding energy differences with respect to the nonmagnetic state ( @xmath55 ) are ( 0.0905 , -0.10875 , -0.21475 ) ev / fe . if these energy differences result mainly from the exchange interactions between the nearest or next - nearest fe moments , then we obtain the following equations @xmath56 from them , we further find that @xmath57 thus the values of @xmath11 and @xmath12 are @xmath58 the energy of the current ferromagnetic state is less accurately determined since this state is not a stable state .
the error in @xmath59 will give rise to the error in @xmath11 and @xmath12 .
[ j - fm ] shows how @xmath11 and @xmath12 change with @xmath59 .
as we see , we find that @xmath60 even we assume the deviation of @xmath59 from our calculated value is as big as 0.1 ev . thus we believe that the collinear antiferromagnetic order observed in laofeas is indeed due to the competition of superexchange interactions .
the above estimation indicates that @xmath61 within the error of calculation for laofeas . in this parameter range ,
as shown by yao and carlson@xcite with the spin wave approximation , the @xmath11 superexchange interaction term will compete strongly with the @xmath12-term .
this results in a strong reduction of the magnetic moment of fe , which would naturally explain why the observed magnetic moment ( @xmath62 ) is significantly smaller than that obtained from the density functional calculations for laofeas . | from first - principles calculations , we have studied the electronic and magnetic structures of the ground state of laofeas .
the fe spins are found to be collinear antiferromagnetic ordered , resulting from the interplay between the strong nearest and next - nearest neighbor superexchange antiferromagnetic interactions . the structure transition observed by neutron scattering is shown to be magnetically driven .
our study suggests that the antiferromagnetic fluctuation plays an important role in the fe - based superconductors .
this sheds light on the understanding of the pairing mechanism in these materials .
recently an iron - based material laofeas was reported to show superconductivity with a transition temperature @xmath0 by partial substitution of o with f atoms@xcite . soon after , other families of fe - as oxyarsenides with la replaced by sm@xcite , ce@xcite , pr@xcite and other rare earth elements were found superconducting with @xmath1 more than 50k . like cuprates ,
these iron arsenides have a layered structure .
the superconducting pairing is believed to happen in the iron - based feas layers .
the high transition temperature and the preliminary band structure calculation suggests that the superconductivity in these fe - arsenide superconductors is not mediated by electron - phonon interaction .
it is commonly believed that the understanding of electronic structures of the parent compound laofeas is the key to determine the underlying mechanism to make it superconducting upon doping .
the early band structure calculations suggested that the pure laofeas compound is a nonmagnetic metal but with strong ferromagnetic or antiferromangtic ( afm ) instability @xcite .
later , it was found that the antiferromagnetically ordered state@xcite has a lower energy than the nonmagnetic one , probably due to the fermi surface nesting@xcite .
dong et al.@xcite predicted that the afm state should form a collinear - striped structure by breaking the rotational symmetry .
this collinear ordered afm state has indeed been observed by the neutron scattering experiment @xcite .
furthermore , the neutron scattering measurement found that there is a structure transition with a monoclinic lattice distortion at @xmath2k and the collinear order is formed about 15@xmath320k below this transition temperature . without this distortion ,
the square afm order induced purely by the fermi surface nesting is expected to be more stable since there are two orthogonal but equivalent nesting directions ( @xmath4 ) and ( @xmath5 ) , which can lower the energy of the ground state by keeping its rotational symmetry@xcite . in this paper
, we report the theoretical result on the electronic and magnetic structures of the ground state of laofeas obtained from first - principles band structure calculations .
we find that there are strong nearest and next - nearest neighbor superexchange interactions in this material ( similar conclusion was obtained by yildirim@xcite ) .
the nearest and next nearest neighbor superexchange interactions have almost the same amplitude within error of calculation .
their competition affects strongly the electronic structure of the ground state .
this drives a small monoclinic lattice distortion and a collinear ordering of fe spins , as observed by neutron scattering .
in our calculations the plane wave basis method was used@xcite .
we adopted the local ( spin ) density approximation and the generalized gradient approximation of perdew - burke - ernzerhof@xcite for the exchange - correlation potentials .
the ultrasoft pseudopotentials @xcite were used to model the electron - ion interactions . after the full convergence test , the kinetic energy cut - off and the charge density cut - off of the plane wave basis were chosen to be 600ev and 4800ev , respectively .
the gaussian broadening technique was used and a mesh of @xmath6 k - points were sampled for the brillouin - zone integration .
laofeas has a tetragonal layered structure with @xmath7 symmetry .
a crystal unit cell consists of eight atoms with alternating feas and lao layers along the c axis . in the calculation ,
the internal atomic coordinates within the cell were determined by the energy minimization .
unit cell while the large dashed square is a @xmath8 unit cell .
the collinear ordered fe spins in the ground state are shown by red arrows.,width=226 ] the previous band structure calculations @xcite used an @xmath9 crystal unit cell as the working cell , in which two fe , as , la , and o atoms were included . to explore the magnetic structure , in particular the collinear afm state of laofeas
, here we use a @xmath10 unit cell ( fig .
1 ) . in order to determine the values of the nearest and next - nearest neighbor coupling constants of spin - spin interaction , @xmath11 and @xmath12 ( see fig .
1 ) , we have evaluated the minimal energies of four different magnetic states of fe ions with constraints imposed if not stable .
these four states have nonmagnetic , ferromagnetic , square afm , and collinear afm orders , respectively . if the energy of the nonmagnetic state of laofeas is set to zero , we find that the energies of the ferromagnetic , square afm , and colliear afm states are ( 0.0905 , -0.010875 , -0.21475 ) ev / fe , respectively .
thus the ground state is a collinear - ordered afm state , in agreement with the experimental observation@xcite .
the magnetic moment around each fe atom is found to be about @xmath13 , varying weakly in the above three magnetically ordered states .
this suggests that the spin of fe ions is between 1 and 3/2 .
the magnetic moment obtained from the neutron scattering model is about 6 times smaller than this result .
this is probably because the correlated effect , especially the strong competition between different afm states , has not been fully included in the density functional theory calculation . to quantify the afm interactions in this material , we assume that these energy differences are predominantly contributed from the interactions between the fe spins which can be modeled by the following frustrated heisenberg model with the nearest and next - nearest neighbor couplings @xmath11 and @xmath12 @xmath14 whereas @xmath15 and @xmath16 denote the summation over the nearest and next - nearest neighbors , respectively . from the calculated energy data ,
we find that @xmath17 and @xmath18 ( the detail how these parameters are determined is given in the appendix ) . if the spin of each fe ion @xmath19 , then @xmath20 and @xmath21 .
the above result indicates that there are competing afm interactions between fe spins .
in particular , the afm coupling between two next - nearest neighboring fe spins is very strong .
this will frustrate the spin nel structure and give rise to a collinear ordered afm ground state . to explore the origin of these afm interactions ,
we have calculated the charge distribution around fe and as ions . the result ( see fig . [
fig : fig2]a and [ fig : fig2]b ) shows that there is almost no charge distribution between two diagonal fe atoms , but there is a strong bonding between fe and as ions .
this indicates that the afm coupling @xmath12 is induced by the superexchange bridged by as ions .
this superexchange is afm because the intermediated state associated with the virtual hopping bridged by as ions is a spin singlet .
the charge distribution between two nearest fe ions is finite .
thus there is a direct exchange interaction between two neighboring fe spins .
since there is a strong hund s coupling between the spins of 3d electrons within each fe ion , the direct exchange interaction is found to be ferromagnetic when the distance of two fe atoms is between 2.4and 2.85 .
however , the overall magnetic coupling @xmath11 between the two nearest fe spins in laofeas is antiferromagnetic .
thus @xmath11 is also dominated by the superexchange interaction bridged by as 4p orbitals . in the collinear afm phase , we find that a small structural relaxation , for which the lattice is slightly expanded along the antiferromagnic ordering direction ( y - axis in fig . 1 ) and shrunk along the ferromagnetic ordering direction ( x - axis in fig .
1 ) , can further lower the ground state energy .
this changes the angle between two principal axes in ab - plane , @xmath22 , from @xmath23 to @xmath24 .
the corresponding energy gain is 7mev .
however , we find that this small lattice distortion affects weakly the band structure and the fe moments .
fig .
[ fig:3 ] shows the total and projected density of states of laofeas at the collinear afm phase . in comparison with the nonmagnetic phase
, we find that most of the states around the fermi level are gapped by the collinear afm order .
this suppresses the total carrier density by more than two orders of magnitude .
the strong suppression is consistent with the hall coefficient measurement which shows that the absolute value of the hall coefficient is enhanced by more than 150 times in the afm phase at 4k versus the nonmagnetic phase above 150k .
furthermore , we find that the density of states at the fermi level is also suppressed compared with the nonmagnetic state@xcite .
however , it is not suppressed as strongly as for the total carrier density .
the corresponding electronic specific heats are evaluated as 0.65mj/(k@xmath25mol ) ( stripe - ordered afm ) and 4.28mj/(k@xmath25mol ) ( nonmagnetic ) , respectively .
this is also consistent with the specific heat measurement .
we have also calculated the band structure of laofeas with 5% f- or 5% sr - doping by taking the virtual crystal approximation in the collinear afm phase .
we find that the overall band structure is hardly changed by 5% electron or hole doping .
the fe moment is also unchanged . only the fermi energy moves up or down with electron or hole doping .
however , as shown in fig .
[ fig:5 ] the fermi surface changes dramatically .
orbitals around one fe atom .
the fermi energy is set to zero.,width=321 ] by projecting the density of states onto the five 3d orbitals of fe ( fig .
[ fig:4 ] ) , we find that the five up - spin orbitals are almost completely filled and the five down - spin orbitals are only partially filled .
however , the down - spin electrons are nearly uniformly distributed in these five 3d orbitals .
this indicates that the crystal field splitting imposed by as atoms is relatively small and the fe 3d - orbitals hybridize strongly with each other .
as the hund rule coupling is strong , this would lead to a large magnetic moment formed around each fe atom , as found in our calculations .
the frustration between the @xmath11 and @xmath12 terms will suppress strongly the afm ordering at the two fe - sublattices , each connected only by the @xmath12 terms .
this , together with the quantum fluctuation , will reduce strongly the average magnetic moment around each fe measured by experiments .
( see fig . 1 for the definition of @xmath22 ) .
the fermi energy is set to zero .
@xmath26 and @xmath27 correspond to 5% f- and 5% sr - doping cases , respectively .
( b ) the brillouin zone .
( c ) the fermi surface of the undoped compound : a hole - type cylinder along @xmath28z and two electron - type pockets between @xmath28 and @xmath29 .
@xmath30 corresponds to the parallel - aligned moment line and @xmath31 corresponds to the antiparallel - aligned moment line .
( d ) the fermi surface of the 5% f - doped compound .
( e ) the fermi surface of the 5% sr - doped compound.,width=321 ] fig .
[ fig:5 ] shows the band structure and the fermi surfaces of electrons in the collinear afm state . unlike in the nonmagnetic state
, there are now only three fermi surface sheets in undoped case , one small hole cylinder along @xmath28z and two small electron pockets formed between @xmath28 and @xmath29 , crossing the fermi level . from the volumes enclosed by these fermi surfaces , we find that the hole carrier density is about @xmath32 and the electron carrier density is about @xmath33
. both decrease by more than two orders of magnitude in comparison with the nonmagnetic or square - afm states@xcite . upon f(sr)-doping , the electron ( hole ) fermi surface sheets
expand while the hole(electron ) fermi surface sheets shrink . with 5% f(sr)-doping ,
the whole fermi surface becomes electron(hole)-typed and the corresponding electron(hole ) carrier density is @xmath34 ( @xmath35 ) , increasing by about 25 times compared with total carrier density in the undoped case .
the above discussion shows that there are strong nearest and next nearest neighbor superexchange interactions in laofeas .
the interplay between these afm interactions can affect strongly the magnetic structure of the ground state .
upon doping , the afm ordering will be suppressed . however , we believe that the remanent afm fluctuation will survive , similar as in cuprate superconductors .
thus the effective low energy model for describing these fe - based superconductors , no matter whether it is a single- or multi - band hamiltonian , should include the frustrated heisenberg terms defined by eq .
( [ eq : heisenberg ] ) . in the fe - based superconductors ,
the magnetic fluctuation , induced by either the afm superexchange interactions or the on - site hund s rule coupling , can be responsible for the superconducting pairing .
the former interaction favors a spin singlet pairing , while the latter favors a spin triplet pairing .
however , the superconductivity induced by the hund s rule coupling would generally involve the interband pairing , which is limited by the available phase space if the total momentum of cooper pair is zero .
this would suggest that the spin triplet pairing is not energetically favorable in a system with strong afm fluctuations .
moreover , from the study of high - t@xmath36 cuprate superconductivity , we know that the strong next - nearest neighbor afm interaction favors an extended @xmath37 or @xmath38-wave pairing . therefore , we believe that the leading pairing instability will be in spin singlet channel , if the superconductivity is driven by the afm fluctuation
. however , the competing nearest neighbor afm interaction may introduce a small new component with different symmetry , for example a @xmath39-wave gap , to the pairing function .
thus the resulting gap parameter will generally be a superposition of two components with different symmetries , for example an extended s plus @xmath39 pairing state .
in conclusion , we have presented first - principles calculations of the electronic structure of laofeas .
we find that there are strong antiferromagnetic nearest and next - nearest neighbor superexchange interactions , bridged by as 4p orbitals .
the next nearest neighbor antiferromagnetic coupling is comparable to the nearest neighbor one .
this gives rise to the collinear afm ordering of fe spins in the ground state as observed by neutron scattering .
the existence of strong antiferromagnetic fluctuations in the fe - based superconductors bears a strong analogy to the high - t@xmath36 cuprates .
this suggests that the superconductivity in these two different kinds of high - t@xmath36 materials may have a common origin .
we wish to thank n.l .
wang , j.l .
luo , p. dai , y.h .
su and h.g .
luo for fruitful discussions .
this work is partially supported by national natural science foundation of china and by national program for basic research of most , china .
while this paper was being finalized , we learnt of the work of s. ishibashi and coworkers ( arxiv:0804.2963 ) .
they obtained a similar band structure .
note added in revision : in a @xmath40 fe - fe lattice of the @xmath11-@xmath12 model with periodic boundary conditions , the @xmath11 and @xmath12 terms are overcounted by a factor of 2 and 4 , respectively .
this overcounting has not been considered in our previous estimation of @xmath11 and @xmath12 , which leads to an overestimation of the values of these parameters as well as the ratio @xmath41 .
this error has been corrected in this revised version . |
retrospective estimates of influenza activity ( ili activity level , as reported by the cdc ) were produced using our model , argo , for the time period of 2009 - 03 - 29 to 2015 - 07 - 11 , assuming we had access only to the historical cdc s ili reports up to the previous week of estimation .
we compared argo s estimates with the ground truth : the cdc - reported weighted ili activity level , published typically with one or two weeks delay , by calculating a collection of accuracy metrics described in the materials section .
these metrics include the root mean squared error ( rmse ) , mean absolute error ( mae ) , mean absolute percentage error ( mape ) , correlation with estimation target , and correlation of increment with estimation target . for comparison , we calculated these accuracy metrics for ( a ) gft estimates ( accessed on 2015 - 07 - 11 ) , ( b ) estimates produced using the method of santillana et al .
2014 @xcite , ( c ) estimates produced by combining gft with an ar(3 ) autoregressive model @xcite , ( d ) estimates produced with an ar(3 ) autoregressive model @xcite , and ( e ) a naive method that simply uses the value of the prior week s cdc s ili activity level as the estimate for the current one . for fair comparison , all benchmark models ( b d )
are dynamically trained with a two - year moving window .
table 1 summarizes these accuracy metrics for all estimation methods for multiple time periods .
the first column shows that argo s estimates outperform all other alternatives , in every accuracy metric for the whole time period .
the other columns of table 1 show the performance of all the methods for the 2009 off - season h1n1 flu outbreak , and each regular flu season since 2010 .
the panels of figure [ fig : all_pred ] display the estimates against the observed cdc - reported ili activity level .
close inspection shows that , in the post-2009 regular flu seasons , argo uniformly outperformed all other alternative estimation methods in terms of root mean squared error , mean absolute error , mean absolute percentage error , and correlation .
argo avoids the notorious over - shooting problem of gft , as seen in figure [ fig : all_pred ] . during the 2009 off - season
h1n1 flu outbreak , argo had the smallest mean absolute percentage error . in terms of root
mean squared error and mean absolute error , argo ( relative rmse = 0.640 , relative mae = 0.584 ) had the second best performance , under - performing slightly only to gft+ar(3 ) model ( relative rmse = 0.580 , relative mae = 0.570 ) . in terms of correlation ,
argo ( r=98.5% ) had similar performance to ( the potentially in - sample data of ) gft ( r=98.9% ) @xcite and gft+ar(3 ) model ( r=98.6% ) , while outperforming all the other alternatives . to assess the statistical significance of the improved prediction power of argo
, we constructed a 95% confidence interval for the relative efficiency of argo compared to other benchmark methods .
the relative efficiency of method 1 to method 2 is the ratio of the true mean squared error of method 2 to that of method 1 @xcite , which can be estimated by its observed value ( see eq ) ; its confidence interval can be constructed by stationary bootstrap of the error residual time series @xcite .
table [ tab : relative_efficiency ] shows that argo is estimated to be at least twice as efficient as any other alternative and the improvement in accuracy is highly statistically significant .
it is well - known that cdc reports undergo revisions , weeks after their initial publication , that respond to internal consistency checks and lead to more accurate estimates of patients with ili symptoms seeking medical attention .
thus , the available historical cdc information , in a given week , is not necessarily as accurate as it will be .
we tested the effect of using ( potentially inaccurate ) unrevised information by obtaining the historical unrevised and revised reports , and the dates when the reports were revised , from the cdc website for the time period of our study .
we used only the information that would have been available to us , at the time of estimation , and produced a time series of estimates for the whole time period described before .
we compared our estimates to all other methods and found that argo still outperformed them all .
moreover , the values of all five accuracy metrics for argo essentially did not change , suggesting a desirable robustness to revisions in cdc s ili activity reports .
the results are shown in table s1 in the supporting information .
we faced an additional challenge in producing real - time estimates for the latest portion of the 2014 - 2015 flu season . at the time of writing
this article , the only data available to us for the week of march 28 , 2015 and later came from the _ google trends _ website .
the information from _ google trends _ has even lower quality than from _ google correlate _ and changes every week .
these undesired changes affected the quality of our estimates . in order to assess the stability of argo in the presence of these variations in the data
, we obtained the search frequencies of the same query terms from _ google trends _
website on 25 different days during the month of april 2015 , and produced a set of 25 historical estimates using argo .
the results of the accuracy metrics associated to these estimates are shown in table s2 in the supporting information .
this table shows that , despite the observed variation in the _ google trends _ data , argo is threefold more stable than the method of @xcite , and still outperforms on average any other method .
.estimate of relative efficiency of argo compared to other models with 95% confidence interval ( ci ) .
relative efficiency being larger than one suggests increased predictive power of argo compared to the alternative method .
[ cols=">,>,>",options="header " , ]
the results presented here demonstrate the superiority of our approach both in terms of accuracy and robustness , when compared to all existing flu tracking models based on google searches .
the value of these results is even higher given the fact that they were produced with low quality input variables .
it is highly likely that our methodology would lead to even more accurate results if we were given access to the input variables that google uses to calculate their estimates .
the combination of seasonal flu information with dynamic reweighting of search information , appears to be a key factor in the enhanced accuracy of argo .
the level of ili activity last week typically has a significant effect on the current level of ili activity , and ili activity half a year ago and/or one year ago could provide further information , as shown in figure s1 of the supporting information , which reflects a strong temporal auto - correlation .
the integration of time series information leads to a smooth and continuous estimation curve and prevents undesired spikes .
however , simply adding gft to an autoregressive model is suboptimal compared to argo , because simply treating gft as an individual variable is incapable to adjust for time series information at the resolution of individual query terms , and many terms included in gft may no longer provide extra information once time series information is incorporated . in fact , once the time series information is included , fewer google search query terms remain significant .
for example , among 100 _ google correlate _ query terms , argo selected 14 terms on average each week , whereas the method of @xcite and gft @xcite selected 38 and 45 terms each week on average , respectively .
the combination of argo s smoothness and sparsity lead to a substantial reduction on the estimation error , as observed in tables 1 and 2 , where argo shows improved performance in all evaluation metrics over the whole time period and is twice as efficient as gft+ar(3 ) .
our methodology allows us to transparently understand how google search information and historical flu information complement one another .
time series models tend to be slow in response to sudden observed changes in cdc s ili activity level .
the ar(3 ) model shows this `` delaying '' effect , despite its seemingly good correlation .
google searches , on the other hand , are better at detecting sudden ili activity changes , but are also very sensitive to public s over - reaction . to investigate further the responsiveness ( co - movement ) of argo towards the change in ili activity , we calculated the correlation of increment between each estimation model and cdc s ili activity level .
the correlation of increment between two time series @xmath2 and @xmath3 is defined as @xmath4 , which measures how well @xmath2 captures the changes in @xmath3 .
table 1 shows that argo has similar capability in capturing the changes in ili level to that of gft and the method of @xcite , while outperforming the time series model ar(3 ) uniformly .
time series information ( seasonality ) tends to pull argo s estimate towards the historical level .
this was evident at the onset of the off - season h1n1 flu outbreak ( week ending at 05/02/2009 ) , which resulted in argo s under - estimation .
argo self - corrected its performance the following week by shifting a portion of model weights from the time series domain to the google searches domain .
inversely , at the height of 2012 - 13 season , argo , gft and the method of @xcite all missed the peak due to an unprecedented surge of search activity .
argo achieved the fastest self - correction by redistributing the weights not only across google terms but also across time series terms , missing the peak by only 1 week , as opposed to 2 weeks for @xcite and about 4 weeks for gft .
it is important to note that while we have used cdc s ili as our gold standard for influenza activity in the us population , and data from google correlate / trends as our independent variables , our methodology can be immediately adapted to any other suitable ili gold standard and/or set of independent variables .
while argo displays a clear superiority over previous methods , it is not fail - proof .
since it relies on the public s search behavior , any abrupt changes to the inner works of the search engine or any changes in the way health - related search information is displayed to users will affect the accuracy of our methodology @xcite .
we expect that argo will be fast at correcting itself if any such change takes place in the future . as in any predictive method ,
the quality of past performance does not guarantee the quality of future performance . in this article
, we fixed the search query terms after 2010 so as to directly compare our results with gft , which kept the same query terms since 2010 ; future application of argo may update search terms more frequently .
argo can be easily generalized to any temporal and spatial scales for a variety of diseases or social events amenable to be tracked by internet searches or services @xcite .
further improvements in influenza prediction may come from combining multiple predictors constructed from disparate data sources @xcite .
after the submission of this article , google announced that gft would be discontinued and that their raw data would be made accessible to selected scientific teams .
this announcement happened soon after the gft team published a manuscript that proposed a new time - series based method for the ( now discontinued ) gft engine @xcite .
this new development makes our contribution timely and useful in providing a transparent method for disease tracking in the future .
all data used in this article are publicly available .
therefore , irb approval is not needed . to avoid forward - looking information in our out - of - sample predictions , and to make the search term selection in our approach consistent with the main revision to gft @xcite immediately after the h1n1 pandemic
, we obtained the highest correlated terms to the cdc s ili using google correlate ( www.google.com/trends/correlate ) for two different time periods .
for the first time period ( pre - h1n1 period ) , we inserted only cdc s ili data from jan 2004 to march 28 , 2009 into google correlate , and used the resulting most highly correlated search terms as independent variables for our out - of - sample predictions for the time period april 4 , 2009 - may 22 , 2010 .
for the second time period ( post - h1n1 ) , we inserted only cdc s ili data from jan 2004 to may 22 , 2010 into google correlate to select new search terms as done in @xcite .
these last search terms were used as independent variables for all subsequent predictions presented in this work .
tables s4 and s5 in the supporting information show all query terms identified . for the pre - h1n1 period ( the first time period ) ,
the terms from google correlate include spurious ( or over - fitted ) terms like `` march vacation '' or `` basketball standings '' , as discussed in @xcite .
however , figure s1 in the supporting information shows that these spurious terms were often not selected by argo , i.e. , argo would give them zero weights , demonstrating its robustness . for the post - h1n1 time period ,
the updated query terms from google correlate include mostly flu - related terms ( see table s5 in supporting information ) .
this suggests that spurious terms were `` filtered - out '' by including off - season flu data . for the time period of march 28 , 2015 up to the date of submission of this article ,
we acquired search frequencies for this set of query terms from google trends ( www.google.com/trends . date of access : july 11 , 2015 ) as google correlate only provides data up to march 28 , 2015 at the time of writing this article .
google correlate standardizes the search volume of each query to have mean zero and standard deviation one across time and contains data only from 2004 to mar 2015 . to make google correlate data compatible with google trends data , we linearly transformed the google correlate data to the same scale of 0 to 100 in our analysis .
we used google correlate data up to its last available date , and then switched to google trends data afterwards .
this is indicated in figure [ fig : all_pred ] by different shades of the background .
we used the latest version of google flu trends ( 4th version , revised in oct 2014 ) weekly estimates of ili activity level as one of our comparison methods .
gft is available at www.google.org/flutrends/us/data.txt ( date of access : 2015 - 07 - 11 ) .
we use the weighted version of cdc s ili activity level as the estimation target ( available at gis.cdc.gov/grasp/fluview/fluportaldashboard.html .
date of access : 2015 - 07 - 11 ) .
the weekly revisions of cdc s ili are available at the cdc website for all recorded seasons ( from week 40 of a given year to week 20 of the subsequent year ) .
for example , ili report revision at week 50 of season 2012 - 13 is available at www.cdc.gov/flu/weekly/weeklyarchives2012-2013/data/senallregt50.htm ; ili report revision at week 9 of season 2014 - 15 is available at www.cdc.gov/flu/weekly/weeklyarchives2014-2015/data/senallregt09.html .
our model argo is motivated by a hidden markov model .
the _ logit_-transformed cdc - reported ili activity level @xmath5 is the intrinsic time series of interest .
we impose an autoregressive ( ar ) model with lag @xmath6 on it , which implies that the collection of vectors @xmath7 is a markov chain ( this captures the clinical fact that flu lasts for a period , but not indefinitely ) .
the vector of _
log_-transformed normalized volume of google search queries at time @xmath8 , @xmath9 , depends only on the ili activity at the same time , @xmath10 ( this follows the intuition that flu occurrence causes people to search flu related information online ) . the markovian property on block @xmath11 leads to the ( vector ) hidden markov model structure . @xmath12 our formal mathematical assumptions are : + ( 1 ) @xmath13 + ( 2 ) @xmath14 + ( 3 ) conditional on @xmath10 , @xmath15 is independent of @xmath16 + where @xmath17 , and @xmath18 is the covariance matrix . to make the variables more normal
, we transform the original ili activity level @xmath19 from @xmath20 $ ] to @xmath21 using the logit function , obtaining the @xmath10 , and transform the google search volumes from @xmath22 $ ] to @xmath21 using the log function , obtaining the @xmath15 .
the log function is appropriate because google search frequencies usually have exponential growth rate near peaks and are artificially scaled to @xmath22 $ ] by dividing the running maximum . since _
google trends _ is in integer scale from 0 to 100 , we add a small number @xmath23 before the transformation to avoid taking the log of @xmath24 .
the predictive distribution @xmath25 is normal with mean linear in @xmath26 and @xmath15 and constant variance ( see the supporting information ) .
this observation leads to equation below , which defines the argo model .
let @xmath27 be the _
logit_-transformed cdc s ( weighted ) ili activity level @xmath19 at time @xmath8 , and @xmath28 the _ log_-transfomred google search frequency of term @xmath29 at time @xmath8 .
our argo model is given by @xmath30 where @xmath15 can be thought as the exogenous variables to time series @xmath5 .
we chose @xmath31 ( weeks ) to capture the within - year seasonality in ili activity , and @xmath32 ( google search terms ) following the data availability from _
google correlate_. since we have more independent variables than the number of observations , the usual maximum likelihood estimate ( ordinary least squares ) method will fail .
therefore , we impose regularities for parameter estimation . in general
we have three kinds of penalties , @xmath0 penalty @xcite , @xmath1 penalty @xcite , and a linear combination of @xmath0 and @xmath1 penalties @xcite .
all parameters are dynamically trained every week with a 2-year ( 104 weeks ) rolling window . in a given week , the goal is to find parameters @xmath33 , @xmath34 , and @xmath35 that minimize @xmath36 where @xmath37 are hyper - parameters .
ideally , we would like to use cross - validation to select all 4 hyper - parameters .
however , since we have only 104 training data points at a given week due to the two - year moving window , the cross - validation result is highly noisy .
thus , we need to pre - specify some of the hyper - parameters . for model simplicity and sparsity , combining with the evidence seen from cross - validation , we set @xmath38 , leading to @xmath0 penalization on both autoregressive and google search terms . with the remaining @xmath39 and @xmath40 , the cross - validation results still have considerable variance . by the same sparsity and simplicity consideration , we further constrained @xmath41 .
therefore , the argo model we finally propose is equation with constraint @xmath42 and @xmath43 .
a detailed discussion of our specification of the hyper - parameters is provided in the supporting information .
the root mean squared error ( rmse ) , mean absolute error ( mae ) , and mean absolute percentage error ( mape ) of estimator @xmath44 to the target ili activity level @xmath45 are defined , respectively , as @xmath46 , @xmath47 , @xmath48 .
the correlation of estimator @xmath44 to the target ili activity level @xmath45 is their sample correlation coefficient .
the correlation of increment between @xmath49 and @xmath19 is defined as + @xmath50 .
the relative efficiency of estimator @xmath51 to estimator @xmath52 is @xmath53 , where @xmath54 $ ] , which can be estimated by @xmath55 the 95% confidence interval can be constructed by time series stationary bootstrap method @xcite , where the replicated time series of the error residual is generated using geometrically distributed random blocks with mean length 52 ( which corresponds to one year ) .
we obtain the basic bootstrap confidence interval for @xmath56 and then recover the original scale by exponentiation .
the non - parametric bootstrap confidence interval takes the autocorrelation and cross - correlation of the errors into account , and is insensitive to the mean block length .
s. c. kou s research is supported in part by nsf grant dms-1510446 . 10 ginsberg j et al .
( 2009 ) detecting influenza epidemics using search engine query data .
457:10121014 .
polgreen pm , chen y , pennock dm , nelson fd , weinstein ra ( 2008 ) using internet searches for influenza surveillance .
47(11):14431448 .
yuan q et al .
( 2013 ) monitoring influenza epidemics in china with search query from baidu . 8(5):e64323 .
paul mj , dredze m , broniatowski d ( 2014 ) twitter improves influenza forecasting . .
mciver dj , brownstein js ( 2014 ) wikipedia usage estimates prevalence of influenza - like illness in the united states in near real - time .
10(4):e1003581 .
santillana m , nsoesie eo , mekaru sr , scales d , brownstein js ( 2014 ) using clinicians search query data to monitor influenza epidemics .
59(10):14461450 .
wesolowski a et al .
( 2014 ) commentary : containing the ebola outbreak the potential and challenge of mobile network data . .
chan eh , sahai v , conrad c , brownstein js ( 2011 ) using web search query data to monitor dengue epidemics : a new model for neglected tropical disease surveillance . 5(5):e1206 .
preis t , moat hs , stanley he ( 2013 ) quantifying trading behavior in financial markets using google trends .
bollen j , mao h , zeng x ( 2011 ) twitter mood predicts the stock market .
2(1):18 .
wu l , brynjolfsson e ( 2015 ) _ the future of prediction : how google searches foreshadow housing prices and sales_. in _ economic analysis of the digital economy _
avi goldfarb sg , tucker c. ( university of chicago press ) , pp .
89118 .
helft m ( 2008 ) google uses searches to track flu s spread ( the new york times ) .
butler d ( 2013 ) when google got flu wrong . 494(7436):155 .
cook s , conrad c , fowlkes al , mohebbi mh ( 2011 ) assessing google flu trends performance in the united states during the 2009 influenza virus a ( h1n1 ) pandemic . 6(8):e23610 .
lazer d , kennedy r , king g , vespignani a ( 2014 ) the parable of google flu : traps in big data analysis
. 343(6176):12031205 .
santillana m , zhang dw , althouse bm , ayers jw ( 2014 ) what can digital disease detection learn from ( an external revision to ) google flu trends ? 47(3):341347 .
stefansen c ( 2014 ) google flu trends gets a brand new engine .
googleresearch.blogspot.com/2014/10/google-flu-trends-gets-brand-new-engine.html .
\(2014 ) influenza ( seasonal ) , fact sheet number 211 . accessed april , 2015 .
shaman j , karspeck a ( 2012 ) forecasting seasonal outbreaks of influenza .
109(50):2042520430 .
lipsitch m , finelli l , heffernan rt , leung gm , redd sc ( 2011 ) improving the evidence base for decision making during a pandemic : the example of 2009 influenza a / h1n1 .
9(2):89115 .
nsoesie eo , brownstein js , ramakrishnan n , marathe mv ( 2014 ) a systematic review of studies on forecasting the dynamics of influenza outbreaks .
8(3):309316 .
chretien jp , george d , shaman j , chitale ra , mckenzie fe ( 2014 ) influenza forecasting in human populations : a scoping review . 9(4):e94130 .
nsoesie e , mararthe m , brownstein j ( 2013 ) forecasting peaks of seasonal influenza epidemics .
soebiyanto rp , adimi f , kiang rk ( 2010 ) modeling and predicting seasonal influenza transmission in warm regions using climatological parameters .
5(3):e9450 .
shaman j , karspeck a , yang w , tamerius j , lipsitch m ( 2013 ) real - time influenza forecasts during the 20122013 season .
yang w , lipsitch m , shaman j ( 2015 ) inference of seasonal and pandemic influenza transmission dynamics .
112(9):27232728 .
paolotti d et al .
( 2014 ) web - based participatory surveillance of infectious diseases : the influenzanet participatory surveillance experience .
20(1):1721 .
dalton c et al .
( 2009 ) flutracking : a weekly australian community online survey of influenza - like illness in 2006 , 2007 and 2008 .
33(3):31622 .
smolinski ms et al .
( 2015 ) flu near you : crowdsourced symptom reporting spanning two influenza seasons .
( 0 ) e1-e7 .
althouse bm , ng yy , cummings da ( 2011 ) prediction of dengue incidence using search query surveillance .
5(8):e1258 .
ocampo aj , chunara r , brownstein js ( 2013 ) using search queries for malaria surveillance , thailand .
12(1):390 .
scarpino sv , dimitrov nb , meyers la ( 2012 ) optimizing provider recruitment for influenza surveillance networks .
8(4):e1002472 .
davidson mw , haim da , radin jm ( 2015 ) using networks to combine `` big data '' and traditional surveillance to improve influenza predictions . 5 .
burkom hs , murphy sp , shmueli g ( 2007 ) automated time series forecasting for biosurveillance .
26(22):42024218 .
everitt bs , skrondal a ( 2002 ) the cambridge dictionary of statistics . .
politis dn , romano jp ( 1994 ) the stationary bootstrap .
89(428):13031313 .
tsukayama h ( 2014 ) google is testing live - video medical advice ( the washington post ) .
accessed on april 20 , 2015 .
gianatasio d ( 2014 ) how this agency cleverly stopped people from googling their medical symptoms : the right ads at the right time ( adweek , online ) .
accessed on april 20 , 2015 .
yang ac , tsai sj , huang ne , peng ck ( 2011 ) association of internet search trends with suicide death in taipei city , taiwan , 20042009 . 132(1):179184
. cavazos - rehg pa et al .
( 2014 ) monitoring of non - cigarette tobacco use using google trends . .
tibshirani r ( 1996 ) regression shrinkage and selection via the lasso .
58(1):267288 .
hoerl ae , kennard rw ( 1970 ) ridge regression : biased estimation for nonorthogonal problems .
12(1):5567 .
zou h , hastie t ( 2005 ) regularization and variable selection via the elastic net .
67(2):301320 .
lampos v , miller ac , crossan s , stefansen c ( 2015 ) advances in nowcasting influenza - like illness rates using search query logs . .
5:12760 santillana m , nguyen at , dredze m , paul mj , nsoesie e , brownstein js ( 2015 ) combining search , social media , and traditional data sources to improve influenza surveillance , 11(10):e1004513 * supporting information *
details of our methodology are presented as follows .
first , the predictive distribution in the formulation of the argo model and the corresponding assumptions are described ; second , the statistical strategy to determine the hyper parameters of the argo model is explained ; third , the results of two sensitivity analysis aimed at testing the robustness of the argo methodology(a ) with respect to subsequent revisions of cdc s ili activity reports , and ( b ) with respect to observed variation of the input variables coming from _ google trends _ data are presented ; fourth , the exact search query terms identified by google correlate with different data access dates are presented ; fifth , a heatmap showing the coefficients for the time series and google search terms dynamically trained by argo is included .
to improve normality for both the input variables and the dependent variables , the cdc - reported ili activity level was _
logit_transformed , and the linearly normalized volume of google search queries were _ log_transformed . to avoid taking the log of 0 , we add a small number @xmath57 before the log - transformation .
these transformations led to two sets of variables , the intrinsic ( influenza epidemics activity ) time series of interest @xmath58 , and the ( google search ) variable vector @xmath9 at time @xmath8 ( that depends only on @xmath10 ) , respectively .
our formal mathematical assumptions are : 1 .
@xmath13 2 .
@xmath14 3 .
conditional on @xmath10 , @xmath15 is independent of @xmath16 where @xmath59 , and @xmath18 is the covariance matrix . the predictive distribution @xmath60 is given by @xmath61 which is a normal distribution , whose mean is a linear combination of @xmath26 and @xmath15 , and whose variance is a constant .
the optimized parameters of the argo model , @xmath33 , @xmath62 , @xmath63 are obtained by @xmath64 the training period consists of a two year ( @xmath65 weeks ) rolling window that immediately precedes the desired date of estimation .
the hyper
parameters are @xmath37 .
we tested the performance of argo with the following specifications of hyper parameters : 1 .
restrict @xmath38 and @xmath41 , cross validate on @xmath39 .
this is our proposed argo with the same @xmath0 penalty for google search terms and autoregressive lags.[item : same_l1 ] 2 .
restrict @xmath38 , cross validate on @xmath66 .
this is argo with separate @xmath0 penalties for google search terms and autoregressive lags.[item : sep_l1 ] 3 .
restrict @xmath67 and @xmath68 , cross validate on @xmath69 .
this is argo with the same @xmath1 penalty for google search terms and autoregressive lags.[item : same_l2 ] 4 .
restrict @xmath68 , cross validate on @xmath70 .
this is argo with separate @xmath1 penalties for google search terms and autoregressive lags.[item : sep_l2 ] 5 .
restrict @xmath71 , cross validate on @xmath72 .
this is argo with the same elastic net ( both @xmath0 and @xmath1 ) penalty for google search terms and autoregressive lags.[item : enet ] table [ tab : hyperpar ] summarizes the in - sample estimation performance for our proposed argo , together with the other specifications of hyper parameters .
it is apparent from the table that the @xmath0 penalty generally outperforms @xmath1 penalty .
the @xmath0 penalty tends to shrink the coefficients of unnecessary independent variables to be exactly zero , and thus eliminates redundant information ; on the other hand , the @xmath1 penalty can only shrink the coefficients to be close to zero . as a result
, @xmath1 penalized coefficients are not as sparse as their @xmath0 counterparts .
furthermore , from table [ tab : hyperpar ] , we see that argo with separate @xmath0 penalties ( specification [ item : sep_l1 ] ) outperforms argo with separate @xmath1 penalties ( specification [ item : sep_l2 ] ) , in terms of both root mean squared error and mean absolute error .
similarly , argo with the same @xmath0 penalty ( specification [ item : same_l1 ] ) outperforms argo with the same @xmath1 penalty ( specification [ item : same_l2 ] ) , in terms of both root mean squared error and mean absolute error .
the elastic net model , which combines @xmath0 penalty and @xmath1 penalty , does not provide any error reduction . in the cross - validation process of setting @xmath72 for the elastic net model
, 70 weeks out of 116 in - sample weeks showed that the smallest cross - validation mean error when restricting @xmath73 ( i.e. zero @xmath1 penalty ) is within one standard deviation of the global smallest cross - validation mean error , suggesting that restricting @xmath1 penalty term to be zero ( i.e. @xmath73 ) will introduce little bias .
therefore , for the simplicity and sparsity of the model , we drop the @xmath1 penalty terms and use only @xmath0 penalty .
next we want to decide between the remaining two specifications , argo with separate @xmath0 penalties ( specification [ item : sep_l1 ] ) , and argo with the same @xmath0 penalty ( specification [ item : same_l1 ] ) .
one might argue that google search terms and autoregressive lags are different sources of information and thus should have different @xmath0 penalties .
however , empirical evidence in table [ tab : hyperpar ] shows that , again , giving extra flexibility to @xmath66 does not generate improvement compared to fixing @xmath43 . in the cross - validation process of setting @xmath66 for separate @xmath0 penalties ,
99 weeks out of 116 in - sample weeks showed that the smallest cross - validation mean error when restricting @xmath43 ( i.e. same @xmath0 penalty ) is within one standard deviation of the global smallest cross - validation mean error .
this may well be due to the gain from variance reduction when imposing the restriction @xmath43 . based on the same simplicity and sparsity consideration
, we finally decided to restrict @xmath38 and @xmath43 in the setting of hyper
parameters for argo .
within a flu season , cdc reports are constantly revised to improve their accuracy as new information is incorporated .
thus , cdc s weighted ili figures displayed in previously published reports may change in subsequent weeks . as a consequence , in a given week the available cdc ili information from the most recent weeks may be inaccurate . to test the robustness of argo in the presence of these revisions and
mimic the real - time tracking in our retrospective predictions , we trained argo and all other alternative models based on the following schedule .
suppose @xmath74 is the cdc - reported ili activity level of week @xmath29 accessed at week @xmath75 . since cdc s ili activity report is typically delayed for one week , on week @xmath75 the historical ili activity level data we have is @xmath76 .
due to revisions , ili activity level of week @xmath29 accessed at different weeks @xmath77 may be different but will converge to a finalized value @xmath78 eventually .
hence , to avoid using forward looking information , in week @xmath75 , we train all models with the ili activity level accessed at that week @xmath76 . in this sense , any future revision beyond week @xmath75 will not be incorporated in the training at week @xmath75 . yet for the accuracy metrics , the estimation target remains the finalized the ili activity level ( @xmath79 ) .
table [ tab : results ] shows the estimation results when using the aforementioned schedule .
note that argo still outperforms all other alternative models .
moreover , the absolute values of all four accuracy metrics for argo trained this way essentially do not change compared to argo trained with finalized ili activity level in the main text , indicating the robustness of argo . the weekly revisions of cdc s ili activity reports are available at cdc website from week 40 of the year to week 20 of the subsequent year for all seasons studied in this article .
for example , ili activity level revisions at week 50 of season 2012 - 2013 are available at http://www.cdc.gov/flu/weekly/ weeklyarchives2012- 2013/data / senallregt50.htm ; ili activity report revision at week 9 of season 2014 - 2015 is available at http://www.cdc.gov /flu / weekly / weeklyarchives2014 - 2015/data / senallregt09.html ( the webpage has suffix `` htm '' for seasons before 2014 - 2015 and suffix `` html '' for 2014 - 2015 season ) . in this retrospective case study , when the revisions of ili activity level were not available for a particular week during off - season period , the finalized ili activity level was used instead .
_ google trends _ historical data constantly change as a consequence of re - normalizations and algorithm updates .
to study the robustness of argo to _ google trends _ data revisions , we obtained the search frequencies of the search query terms identified by _ google correlate _ on may 22 , 2010 ( see figure 2 in the main text and table [ tab : phrases ] below ) from the _ google trends _ website ( http://www.google.com/trends ) on 25 different days in april 2015 .
we studied the variability of argo s performance when using these 25 different versions of _ google trends _ data as input variables for the common time period of sep 28 , 2014 to mar 29 , 2015 .
we studied the 2014 - 15 flu season only partially ( up to march 2015 ) because this is the longest study period covered by all the obtained versions of _ google trends _ data , at the time ( may 1 , 2015 ) of the first submission of this article .
we want to emphasize that google correlate data were only available up to feb 2014 when accessed in april 2015 . despite the inevitable variation to the revision of the low - quality data from _ google trends _ ,
argo still achieves considerable stability compared to the method of santillana et al .
@xcite during this time period .
table [ tab : sensi_gt ] suggests that argo is threefold more robust than the method of @xcite .
the incorporation of time series information helps argo achieve the stability . as an extreme example , ar(3 )
model focuses entirely on the time series information and is thus independent of _ google trends _ data revisions .
gft , formulated with the original search variables as inputs , is by construction insensitive to the changes in _ google trends _ data . for this portion of the study
, we included the signal from gft for context only and we treat it as exogenous in our analysis .
based on the results from previous time periods , it is highly likely that if we had access to google s internal raw data ( i.e. , historical search volume for disease - related phrases ) we would have achieved the same stability as well . yet even with these low - quality data , argo outperforms gft uniformly on all versions of data in terms of both root mean squared error and mean absolute error .
tables [ tab : phrases09 ] and [ tab : phrases ] list the search query phrases identified by google correlate as of march 28 , 2009 and of may 22 , 2010 , respectively .
the march 2009 version included spurious terms such as `` college.basketball.standings '' , `` march.vacation '' , `` aloha.ski '' , `` virginia.wrestling '' , etc .
these spurious terms did not appear in the may 2010 version .
figure [ fig : coef ] shows the coefficients for the time series and google search terms dynamically trained by argo via a heatmap .
the level of ili activity last week is seen to have a significant effect on the current level of ili activity , and ili activity half a year ago and/or one year ago could provide further information as the figure shows . among _
google correlate _ query terms , argo selected 14 terms out of 100 on average each week . | accurate real - time tracking of influenza outbreaks helps public health officials make timely and meaningful decisions that could save lives .
we propose an influenza tracking model , argo ( autoregression with google search data ) , that uses publicly available online search data .
in addition to having a rigorous statistical foundation , argo outperforms all previously available google - search based tracking models , including the latest version of google flu trends , even though it uses only low - quality search data as input from publicly available google trends and google correlate websites .
argo not only incorporates the seasonality in influenza epidemics but also captures changes in people s online search behavior over time .
argo is also flexible , self - correcting , robust , and scalable , making it a potentially powerful tool that can be used for real - time tracking of other social events at multiple temporal and spatial resolutions .
this is the preprint of the paper published at pnas : dx.doi.org/10.1073/pnas.1515373112 .
there are some minor differences between this preprint and the published paper .
big data sets are constantly generated nowadays as the activities of millions of users are collected from internet - based services .
numerous studies have suggested great potential of these big data sets to detect / manage epidemic outbreaks ( influenza @xcite , ebola @xcite , dengue @xcite ) , predict changes in stock prices @xcite and housing prices @xcite , etc . in 2009 , google flu trends ( gft ) , a digital disease detection system that uses the volume of selected google search terms to estimate current influenza - like illnesses ( ili ) activity , was identified by many as a good example of how big data would transform traditional statistical predictive analysis @xcite .
however , significant discrepancies between gft s flu estimates and those measured by the centers for disease control ( cdc ) in subsequent years led to considerable doubt about the value of digital disease detection systems @xcite . while multiple articles have identified methodological flaws in gft s original algorithm @xcite and
have led to incremental improvements @xcite , a statistical framework that is theoretically sound and capable of accurate estimation is still lacking . here
we present such a framework that culminates in a new method that outperforms all existing methodologies for tracking influenza activity using internet search data .
influenza outbreaks cause up to 500,000 deaths a year worldwide , and an estimated 3,000 to 50,000 deaths a year in the usa @xcite .
our ability to effectively prepare for and respond to these outbreaks heavily relies on the availability of accurate real - time estimates of their activity .
existing methods to predict the timing , duration and magnitude of flu outbreaks remain limited @xcite .
well - established clinical methods to track flu activity , such as the cdc s ilinet , report the percentage of patients seeking medical attention with ili symptoms ( www.cdc.gov/flu/ ) .
while cdc s % ili is only a proxy of the flu activity in the population , it can help officials allocate resources in preparation for potential surges of patient visits to hospital facilities .
see @xcite for further discussion .
cdc s ili reports have a delay of one to three weeks due to the time for processing and aggregating clinical information .
this time lag is far from optimal for decision - making purposes . in order to alleviate this information gap , multiple methods combining climate ,
demographic and epidemiological data with mathematical models have been proposed for real - time estimation of flu activity @xcite . in recent years
, methods that harness internet - based information have also been proposed , such as google @xcite , yahoo @xcite , and baidu @xcite internet searches , twitter posts @xcite , wikipedia article views @xcite , clinicians queries @xcite , and crowd sourced self - reporting mobile apps such as influenzanet ( europe ) @xcite , flutracking ( australia ) @xcite , and flu near you ( usa ) @xcite . among them , gft has received most attention and has inspired subsequent digital disease detection systems @xcite .
interestingly , google has never made their raw data public , thus , making it impossible to reproduce the exact results of gft .
we highlight three limitations of the original gft algorithm , previously identified in @xcite .
first , it was shown that a static approach , which does not take advantage of newly available cdc s ili activity reports as the flu season evolves , produced model drift , leading to inaccurate estimates .
second , the idea of aggregating the multiple query terms ( the independent variables in the gft model ) into a single variable did not allow for changes in people s internet search behavior over time ( and thus changes in query terms abilities to track flu ) to be appropriately captured .
third , gft ignored the intrinsic time series properties , such as seasonality of the historical ili activity , thus overlooking potentially crucial information that could help produce accurate real time ili activity estimates .
the new methodology presented here produces robust and highly accurate ili activity level estimates by addressing the three aforementioned shortcomings of the multiple gft engines .
in addition , we provide a theoretical framework that , for the first time , justifies the prevailing usage of linear models in the digital disease detection literature by incorporating causality arguments through a hidden markov model .
this theoretical framework contains as a special case the model developed in @xcite .
our new model not only achieves the goal of ( a ) dynamically incorporating new information from cdc reports as they become available and ( b ) automatically selecting the most useful google search queries for estimation as in @xcite , but also largely improves estimation by ( c ) including the long - term cyclic information ( seasonality ) from past flu seasons on record as input variables , and ( d ) using a two - year moving window ( which immediately precedes the desired date of estimation ) for the training period to capture the most recent changes in people s search patterns and time series behavior @xcite .
our methodology efficiently builds a prediction model from individual search frequency as well as the past records of ili activity .
it utilizes both sources of information more efficiently than simply combining gft with autoregressive terms as suggested in @xcite , since gft is not optimally aggregated to provide additional information on top of time series information .
furthermore , we provide a quantitative efficiency metric that measures the statistical significance of the improvement of our methodology over other alternatives . for example
, our method is twice as accurate as the method that combines gft with autoregressive terms ( see table [ tab : relative_efficiency ] ) .
finally , even though we use as input only the publicly available , low - quality data from the _ google correlate _ and _ google trends _ websites , our method has significant improvement over the latest version of gft .
we name our model argo , which stands for autoregression with google search data . statistically speaking
, argo is an autoregressive model with google search queries as exogenous variables ; argo also employs @xmath0 ( and potentially @xmath1 ) regularization in order to achieve automatic selection of the most relevant information . |
spatial solitons are nonlinear localized states that keep their form during propagation due to the balance between diffraction and self - induced nonlinear potential @xcite .
recently , there has been a lot of interest in a generalized type of spatial solitons , the so - called azimuthons .
these are azimuthally modulated beams , that exhibit steady angular rotation upon propagation @xcite .
they can be considered as azimuthally perturbed optical vortices , i.e. beams with singular phase structure @xcite .
theoretical studies demonstrated both , stable and unstable propagation of azimuthons @xcite , and the first experimental observation of optical azimuthons was recently achieved in rubidium vapors @xcite .
it appears that higher order solitonic structures and optical vortices are generally unstable in typical nonlinear media with local ( kerr - like ) response @xcite . on the other hand
, it has been shown that a spatially nonlocal nonlinear response provides stabilization of various complex solitonic structures including vortices @xcite .
consequently , azimuthons and their dynamics have been studied almost exclusively in the context of spatially nonlocal nonlinear media @xcite . in spite of the fact that there are various physical settings exhibiting nonlocality such as nematic liquid crystals @xcite , bose - einstein condensates @xcite , plasmas @xcite , thermo - optical materials @xcite etc .
, experimental realization of such systems is always quite involved .
moreover , from the theoretical point of view , nonlocal media are quite challenging for numerical modeling and analytical treatment . in this paper
we propose a much simpler and experimentally accessible optical system to study the propagation of azimuthons : a weakly nonlinear optical multi - mode waveguide . here , weakly nonlinear means that the nonlinear induced index change , which is proportional to the intensity of the optical beam , is small compared to the index profile ( or trapping potential ) of the waveguide .
following @xcite , we can expect that weakly nonlinear azimuthons are stable in multi - mode waveguides .
the paper is organized as follows : in sec .
[ modelling ] , we briefly introduce the general model equation for beam propagation in weakly nonlinear waveguides , and then we discuss in detail the properties of dipole azimuthons in circular and square waveguides in sec .
[ dipoles ] , respectively . in sec .
[ higherorder ] , higher order azimuthons are investigated , and in sec .
[ conclusion ] we conclude .
we consider the evolution of a continuous wave ( cw ) optical beam with amplitude @xmath0 , where @xmath1 and @xmath2 denote the transverse and longitudinal coordinates , respectively .
then the propagation of this beam in a weakly - guiding waveguide with kerr nonlinearity in the scalar , slowly varying envelope approximation is described by the following equation @xcite : @xmath3 where @xmath4 refers to the carrier central wave number in the medium ; @xmath5 is the kerr nonlinear coefficient , @xmath6 the linear refractive index distribution , @xmath7 the background index , and @xmath8 . we consider a weakly - guiding waveguide , so both the linear and nonlinear induced index change @xmath9 and @xmath10 are small compared to the mean index @xmath7 , and at the same time @xmath11 , to guarantee a weak nonlinearity . from the mathematical point of view , eq .
( [ physicaleq ] ) is a ( 2 + 1)-dimensional nonlinear schrdinger ( nls ) equation with linear potential representing the waveguide profile . for technical convenience ,
we rescale eq .
( [ physicaleq ] ) to dimensionless quantities with @xmath12 , @xmath13 , @xmath14 , and @xmath15 , where @xmath16 represents the transverse spatial extent of the waveguide .
then , the two - dimensional ( 2d ) nls equation for the scaled wave function @xmath17 reads @xmath18 with an `` attractive '' bounded potential @xmath19 , given by the spatial profile of the refractive index of the waveguide . here
we will consider propagation in step index waveguides with parameter @xmath20 given as follows @xmath21 @xmath22 is the height of the potential and determines the number of linear waveguide modes .
considering the specific case of silica - made waveguides , typical values for the parameters are @xmath23 , @xmath24 , @xmath25 at a vacuum wavelength of @xmath26 nm , and those values do not vary much when we choose any @xmath27 in the range from visible to near - infrared .
the laser - induced - damage - threshold ( lidt ) for synthetic fused silica is about @xmath28 for nanosecond pulses ( according to eq .
( 1 ) and table 2 in ref .
@xcite ) corresponding to @xmath29 , up to which the model should be valid . for shorter , femtosecond pulses
this damage threshold is @xmath30 times higher , but since we consider `` cw '' beams the lower value for nanosecond pulses is relevant .
it is justified to neglect temporal effects on beam propagation , because the dispersion length is already of the order of kilometers for pulse durations of a few tens of picoseconds , much longer than propagation distances considered in this work . throughout this paper , we choose @xmath31 , and find that @xmath32 through @xmath33\right|}$ ] , which nicely corresponds to the radius of a standard circular multi - mode fiber @xcite .
in addition , in order to guarantee intensities smaller than the lidt the condition @xmath34 should be fulfilled .
note that in eq .
( [ partial ] ) @xmath35 is the sign of nonlinearity : @xmath36 ( @xmath37 ) represents focusing ( defocusing ) nonlinearity .
the main difference between weakly nonlinear waveguides with focusing and defocusing nonlinearity is that defocusing nonlinearity supports higher amplitudes of the wave function , and in general leads to more stable and robust configurations . in this paper , however , we consider the experimentally relevant case of a focusing nonlinearity .
azimuthons are a straightforward generalization of the usual ansatz for stationary solutions @xcite .
they represent spatially rotating structures and hence involve an additional parameter , the rotation frequency @xmath38 ( see also @xcite ) , so we seek approximate solutions of the form @xmath39 where @xmath40 and @xmath41 the azimuthal angle in the transverse plane @xmath42 , @xmath43 is the stationary profile , @xmath38 the rotation frequency , and @xmath44 the propagation constant . for @xmath45 ,
azimuthons become ordinary ( non - rotating ) solitons .
the simplest family of azimuthons is the one connecting the dipole soliton with the single charged vortex soliton @xcite .
a single charged vortex consists of two equal - amplitude dipole - shaped structures representing real and imaginary part of @xmath43 .
if these two components differ in amplitude , the resulting structure forms a `` rotating dipole '' azimuthon .
if one of the components is zero we deal with the dipole soliton , which consists of two out - of - phase humps and does not rotate for symmetry reasons . in a first attempt ,
let us assume we know the radial shape of the linear vortex mode @xmath46 which is normalized according to @xmath47 . then , using separation of variables , we consider the simplest so - called `` rotating dipole '' azimuthon with ansatz @xcite @xmath48 , \label{complexamplitude}\ ] ] where @xmath49 is an amplitude factor , and @xmath50 the azimuthal modulation depth of the resulting ring - like structure . because we are operating in the weakly nonlinear regime , using linear waveguide modes as initial conditions for nonlinear ( soliton ) solutions is a quite good approximation .
after plugging eq .
( [ ansatz ] ) into eq .
( [ partial ] ) , multiplying by @xmath51 and @xmath52 respectively , and integrating over the transverse coordinates we end up with @xcite @xmath53 @xmath54 this system relates the propagation constant @xmath44 and the rotation frequency @xmath38 of the azimuthons to integrals over their stationary amplitude profiles : @xmath55 @xmath56 @xmath57 @xmath58 |u(r)|^2 ~\mathrm{d}r~\mathrm{d}\phi,\ ] ] @xmath59 @xmath60 @xmath61 \frac{\partial u^*(r)}{\partial \phi } u(r ) ~\mathrm{d}r~\mathrm{d}\phi.\ ] ] the first two quantities @xmath62 and @xmath63 have straightforward physical meanings , namely power and angular momentum . the integrals @xmath64 and @xmath65 are related to the diffraction mechanism of the system , whereas @xmath66 and @xmath67 account for waveguide and nonlinearity .
thus we can formally solve for the rotation frequency with these quantities and obtain @xmath68 after inserting eq .
( [ complexamplitude ] ) into eq .
( [ angularfrequency ] ) , it turns out that only the nonlinear term contributes to the rotation frequency @xmath38 ( see also eq .
( 10 ) in ref .
@xcite ) . in order to give an estimate for the rotation frequency
, we use the linear stationary modal profile @xmath46 of a circular waveguide expressed in terms of bessel functions of first kind @xmath69 and modified bessel function of second kind @xmath70 : @xmath71 where @xmath72 is a normalization factor such that @xmath73 .
thus , the analytically predicted rotation frequency is @xmath74 from the above equation , one can see that sense of rotation is opposite for focusing and defocusing nonlinearities .
moreover , as @xmath75 , one can find from eq .
( [ analyticalangular ] ) an expression for the physical distance @xmath76 over which the azimuthon in a circular waveguide performs one full rotation : @xmath77 for example , assuming propagation in a silica fiber with the parameters discussed earlier ( see sec . [ modelling ] ) and taking @xmath78 , @xmath79
, we obtain @xmath80 .
figure [ circularrelations ] shows the dependence of the azimuthon rotation frequency as a function of its amplitude @xmath49 ( left panel ) and the modulation parameter @xmath81 ( right panel ) .
blue solid lines represent the analytical predictions and , in excellent agreement , red dots are obtained from numerical simulations .
an exemplary propagation of the dipole azimuthon in circular waveguide rotating with @xmath82 is illustrated in fig .
[ circularevolution ] .
the top iso - surface plot displays 3d rigid and continuous rotation of the azimuthon during propagation .
the row below depicts the transverse intensity ( left ) and phase ( right ) distribution of the dipole azimuthon during propagation after rotating by @xmath83 .
versus amplitude factor @xmath49 ( left panel ) and amplitude ratio @xmath81 ( right panel ) .
blue solid lines show analytical predictions from eq .
( [ analyticalangular ] ) .
red dots denote results obtained from numerical simulations of eq .
( [ partial ] ) .
the values of @xmath49 and @xmath81 are shown next to the lines.,title="fig:",scaledwidth=40.0% ] versus amplitude factor @xmath49 ( left panel ) and amplitude ratio @xmath81 ( right panel ) .
blue solid lines show analytical predictions from eq .
( [ analyticalangular ] ) .
red dots denote results obtained from numerical simulations of eq .
( [ partial ] ) .
the values of @xmath49 and @xmath81 are shown next to the lines.,title="fig:",scaledwidth=40.0% ] in the circular waveguide .
the iso - surface plot on the top clearly displays the spiraling of the azimuthon during propagation .
figures in the row below depict the azimuthon s transverse intensity distribution at input , and after rotating @xmath83 , respectively .
the white line indicates the waveguide boundaries.,title="fig:",scaledwidth=55.0% ] + in the circular waveguide .
the iso - surface plot on the top clearly displays the spiraling of the azimuthon during propagation .
figures in the row below depict the azimuthon s transverse intensity distribution at input , and after rotating @xmath83 , respectively .
the white line indicates the waveguide boundaries.,title="fig:",scaledwidth=35.0% ] in the circular waveguide .
the iso - surface plot on the top clearly displays the spiraling of the azimuthon during propagation .
figures in the row below depict the azimuthon s transverse intensity distribution at input , and after rotating @xmath83 , respectively .
the white line indicates the waveguide boundaries.,title="fig:",scaledwidth=35.0% ] because of the lack of circular symmetry , azimuthons in the strict sense of eq .
( [ complexamplitude ] ) ( preservation of shape during rotation with constant frequency ) can not exist in a square waveguide .
however , as we will show below , the azimuthon - like behavior is still possible even though the beam propagation is accompanied by variation of the beam transverse intensity distribution . to set the initial condition , we use two linear degenerated orthogonal dipole modes @xmath84 , @xmath85 ( as in the case of the circular waveguide ) , which are normalized according to @xmath86 , and superpose them as before to form the azimuthon - like object @xmath87 the field eq .
( [ squareansatz ] ) is then used as an initial condition to the nonlinear schrdinger equation eq .
( [ partial ] ) .
however , in contrast to the circular waveguide , the orientation of the dipoles @xmath84 and @xmath85 is important in nonlinear regime . if the two orthogonal dipoles @xmath84 and @xmath85 are oriented along the diagonals of the waveguide cross - section ( see first subplot of fig .
[ square_r_evolution ] ) , for a given amplitude @xmath49 rotation occurs only if the modulation parameter @xmath81 exceeds a critical value @xmath88 .
moreover , the rotation is no longer constant as in the case of cylindrical waveguide but fluctuates and hence in what follows we use its average value ( termed `` average frequency '' ) @xmath89 with @xmath90 being the propagation distance corresponding to one full @xmath91 rotation .
this threshold value @xmath92 decreases if we choose different initial dipole orientations , ) by superposition . ] and appears to be zero ( within our numerical accuracy ) for parallel orientation with respect to the waveguide boundaries . , respectively .
the white line indicates the waveguide boundaries .
the iso - surface plot at the bottom displays the rotation and deformation during propagation .
the initial amplitude and modulation parameters are @xmath93.,title="fig:",scaledwidth=35.0% ] , respectively .
the white line indicates the waveguide boundaries .
the iso - surface plot at the bottom displays the rotation and deformation during propagation .
the initial amplitude and modulation parameters are @xmath93.,title="fig:",scaledwidth=35.0% ] + , respectively .
the white line indicates the waveguide boundaries .
the iso - surface plot at the bottom displays the rotation and deformation during propagation .
the initial amplitude and modulation parameters are @xmath93.,title="fig:",scaledwidth=55.0% ] let us focus on the case where the initial dipole - like field structure is oriented along the diagonal of the waveguide cross - section , and from now on the notation @xmath84 and @xmath85 stands for this orientation .
figure [ square_r_evolution ] shows an exemplary propagation of the resulting azimuthon - like solution with @xmath94 , and an average rotation frequency of @xmath95 .
the top row depicts intensity and phase distribution at different propagation distance while the bottom plots illustrate the full 3d evolution of the `` soliton '' . the rotation ( counter - rotating w.r.t .
phase ) accompanied by beam deformation is clearly visible .
for an amplitude ratio @xmath81 less than the critical value ( @xmath96 ) the azimuthon no longer rotates but swings back and forth and hence its average rotation frequency is zero .
the propagation of this wobbling dipole with an amplitude ratio @xmath81 smaller than the critical value is displayed in fig .
[ square_t_evolution ] .
the right panels show the maximum angle which the dipole attains during propagation .
the 3d surface plot in the bottom row clearly illustrates the swinging movement of the `` soliton '' upon propagation .
.,title="fig:",scaledwidth=35.0% ] .,title="fig:",scaledwidth=35.0% ] + .,title="fig:",scaledwidth=55.0% ] due to the azimuthon profile deformation , it is not possible to find an analytical expression of @xmath97 as a function of @xmath49 and @xmath81 as in the case of circular waveguide ( i.e. eq .
( [ analyticalangular ] ) ) .
therefore , we need to resort to numerical simulations . in fig .
[ squarerelations ] we show the numerically determined relation between @xmath97 and the azimuthon parameters @xmath49 ( left panel ) and @xmath81 ( right panel ) . the threshold - like behavior is evident in the right graph of this figure : the region @xmath96 , in which the soliton wobbles , is depicted by a green line . versus amplitude factor @xmath49 ( left panel ) and amplitude ratio @xmath81 ( right panel ) in a square waveguide .
red dots are numerical results obtained from eq .
( [ partial ] ) , blue curves ( @xmath98 ) and blue lines ( @xmath99 ) are the fitting results to the numerical simulations .
the green line represents the superposed dipoles twist during propagation .
the values of @xmath49 and @xmath81 are shown next to the curves .
the black square represents the analytical estimate for @xmath92.,title="fig:",scaledwidth=40.0% ] versus amplitude factor @xmath49 ( left panel ) and amplitude ratio @xmath81 ( right panel ) in a square waveguide .
red dots are numerical results obtained from eq .
( [ partial ] ) , blue curves ( @xmath98 ) and blue lines ( @xmath99 ) are the fitting results to the numerical simulations .
the green line represents the superposed dipoles twist during propagation .
the values of @xmath49 and @xmath81 are shown next to the curves .
the black square represents the analytical estimate for @xmath92.,title="fig:",scaledwidth=40.0% ] let us have a closer look at the boundary between domains of azimuthon rotation ( @xmath94 ) and wobbling ( @xmath96 ) .
we find numerically that the critical value @xmath92 , which separates those two domains , depends very weakly on the azimuthon amplitude @xmath49 : we are in weakly nonlinear limit , and may use linear waveguide modes to elucidate nonlinear propagation properties of the azimuthons .
the appearance of a threshold value @xmath92 , above which the azimuthon rotates , can be explained by considering the hamiltonian associated with propagation equation eq .
( [ partial ] ) , @xmath100 which is a conserved quantity upon propagation . depending on their orientation , stationary dipoles ( @xmath101 ) have a slightly different value of @xmath102 in nonlinear regime . if we take our diagonal ( w.r.t .
waveguide cross - section ) dipoles @xmath84 and @xmath85 from above , we can construct a parallel dipole @xmath103 in the following way : @xmath104 the intensity distribution of this dipole @xmath103 is very close to that shown in the snapshot of the rotating azimuthon shown in the right panel in fig .
[ square_r_evolution ] .
one can show using the definition of the hamiltonian eq .
( [ hamiltonian ] ) that @xmath105 .
in fact , it turns out that those two dipole orientations ( parallel and diagonal ) correspond to the extremal values of the hamiltonian . now , if the azimuthon rotates it has to pass through all possible orientations , i.e. , its value of @xmath102 has to be larger than @xmath106 .
we can use this reasoning to estimate the value of @xmath92 : the left panel of fig .
[ angularmomentumcurve ] depicts the dependence of the hamiltonian of superposed diagonal dipoles as a function of @xmath81 ( red line ) for constant power @xmath107 .
the blue line represents the value of the hamiltonian of the parallel dipole @xmath108 at the same power level .
as the red curve monotonically increases with modulation parameter @xmath81 , the intersection of the two curves gives an estimate of the critical value of the azimuthon modulation @xmath92 , which is shown in the right panel in fig .
[ squarerelations ] by a black square .
in other words , in order to rotate from the diagonal to vertical position the azimuthon must overcome some kind of energy barrier represented by the difference of the values of the hamiltonian in those two basic states .
[ angularmomentumcurve ] shows this can be achieved by increasing the value of @xmath81 in the diagonal state to @xmath109 , which is very close to the value found numerically ( see right panel in fig . [ squarerelations ] ) .
similar phenomena have been identified , for instance , in discrete nonlinear systems where the discreteness introduces the so called peierls - nabarro potential which has to be overcome by a soliton to become mobile @xcite .
( red curve ) as a function of @xmath81 and hamiltonian of @xmath110 ( blue line ) ( for the same power @xmath107 ) .
the intersection of the two curves defines @xmath92 ( depicted by the black square ) .
right panel : the angular momenta of the rotating superposed dipoles ( corresponding to fig .
[ square_r_evolution ] , blue curve ) , and the twisting ones ( corresponding to fig . [ square_t_evolution ] , red curve ) .
the other curves depict the cases of @xmath81 slightly above ( black ) and bellow ( green ) @xmath92 .
all curves are obtained for @xmath78 .
, title="fig:",scaledwidth=45.0% ] ( red curve ) as a function of @xmath81 and hamiltonian of @xmath110 ( blue line ) ( for the same power @xmath107 ) .
the intersection of the two curves defines @xmath92 ( depicted by the black square ) .
right panel : the angular momenta of the rotating superposed dipoles ( corresponding to fig .
[ square_r_evolution ] , blue curve ) , and the twisting ones ( corresponding to fig . [ square_t_evolution ] , red curve ) .
the other curves depict the cases of @xmath81 slightly above ( black ) and bellow ( green ) @xmath92 .
all curves are obtained for @xmath78 .
, title="fig:",scaledwidth=45.0% ] it should be emphasized that the crucial difference between azimuthon dynamics in circular and square waveguide originates from the fact that angular momentum of the beam is conserved in the circular waveguide , whereas it changes considerably in the square waveguide . in the right panel of fig .
[ angularmomentumcurve ] we display the angular momentum of the beam in the square waveguide for four different values of modulation parameter @xmath81 .
if @xmath81 is above the critical value @xmath92 , the angular momentum does not change its sign as the blue curve shows .
if @xmath96 , the angular momentum changes periodically its sign in propagation ( red curve ) , which indicates the swinging motion of the soliton . for values @xmath81 close to the critical value the variation of the angular momentum are largest as shown by the green and black curves
at the same time , the appearance of broad plateaus indicates slow propagation dynamics .
it is worth mentioning that we also investigated other waveguide symmetries , such as hexagonal waveguides , and found the behavior of the dipole azimuthons to be qualitatively similar to that of the square waveguide , i.e. , they either rotate or wobble , depending on modulation depth and initial orientation . however , the difference of the values of the hamiltonian for the two extremal dipoles is now smaller , resulting in a lower threshold value @xmath92 .
we believe that this threshold behavior in the propagation dynamics of azimuthons is generic for any non - circular - symmetric waveguide structure which supports degenerated dipole modes .
in analogy to dipole - azimuthons discussed so far , it is possible to consider dynamics of higher order azimuthons constructed by using pairs of degenerated higher order waveguide modes . in particular , in a circular waveguide degenerate higher order modes have the form of quadrupoles , hexapoles , octapoles , decapoles , etc . and other modes with even number of lobes ( optical necklaces ) . in a square waveguide , one can identify pairs of degenerate modes in a form of hexapoles , octapoles , decapoles , dodecapoles , etc .
( optical matrices ) .
interestingly , the quadrupoles found in square waveguides are not degenerate and hence the propagation dynamics of their corresponding superposed states in weakly nonlinear regime is dominated by linear mode beating . in the circular waveguide .
left panel shows the original hexapole azimuthon and the corresponding phase ; right panel shows the iso - surface plot of the propagation.,title="fig:",scaledwidth=45.0% ] in the circular waveguide .
left panel shows the original hexapole azimuthon and the corresponding phase ; right panel shows the iso - surface plot of the propagation.,title="fig:",scaledwidth=50.0% ] fig .
[ cquadr_evolution ] displays an example of hexapole azimuthon in a circular waveguide .
the left panel represents the initial intensity and phase distribution of the azimuthon . the 3d surface plot in the right panel depicts stable and smooth rotation of the hexapole azimuthon . as far as the higher order azimuthons of the square waveguide are concerned
, they share the dynamical properties of dipole azimuthons discussed before . as an example , fig .
[ hexa_evolution ] shows the rotating ( top row ) and swinging ( middle row ) hexapoles . due to the beam deformation
, the hexapole with @xmath79 transforms into a ( @xmath111 ) matrix while rotating .
the left two panels in the third row of fig .
[ hexa_evolution ] display the rotating dodecapole azimuthon and its deformation into a ( @xmath112 ) solitonic matrix ; the right panels show rotating icosapole azimuthon and its deformation into a ( @xmath113 ) solitonic matrix .
( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ] ( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ] + ( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ] ( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ] + ( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ] ( media 1 ) , and the second row shows a twisting azimuthon with @xmath114 ( media 2 ) .
the superposed dodecapoles ( left two panels ) , and superposed icosapoles ( right two panels ) as well as their deformations are shown in the third row ; in both cases we choose @xmath78 and @xmath79.,title="fig:",scaledwidth=35.0% ]
in conclusion , we demonstrated numerically stable propagation of azimuthons , i.e. localized rotating nonlinear waves in weakly nonlinear waveguides .
depending on the waveguide profile , different nonlinearity induced propagation dynamics can be observed .
we showed that azimuthons in circular waveguides rotate continuously .
the analytically predicted dependence of rotation frequency @xmath38 as a function of soliton parameters was found to be in excellent agreement with numerical simulations .
further on , we discussed propagation of azimuthon - like structures in square waveguides .
we showed that their dynamics critically depends on the initial conditions .
in particular , we found a threshold - like behavior in the propagation dynamics , separating rotating and wobbling solutions .
we showed that this effect is associated with different values of the hamiltonian for different azimuthon orientations .
as our analysis relies on physical parameters of actual multi - mode waveguides , our findings may open a relatively easy route to experimental observations of stable azimuthons .
numerical simulations were partly performed on the sgi xe cluster and the sun constellation vayu cluster of the australian partnership for advanced computing ( apac ) .
this research was supported by the australian research council . | we show that weakly guiding nonlinear waveguides support stable propagation of rotating spatial solitons ( azimuthons ) .
we investigate the role of waveguide symmetry on the soliton rotation .
we find that azimuthons in circular waveguides always rotate rigidly during propagation and the analytically predicted rotation frequency is in excellent agreement with numerical simulations . on the other hand , azimuthons in square waveguides may experience spatial deformation during propagation .
moreover , we show that there is a critical value for the modulation depth of azimuthons above which solitons just wobble back and forth , and below which they rotate continuously .
we explain these dynamics using the concept of energy difference between different orientations of the azimuthon . 10 g. i. stegeman and m. segev , `` optical spatial solitons and their interactions : universality and diversity , '' science * 286 * , 15181523 ( 1999 ) .
a. s. desyatnikov , a. a. sukhorukov , and y. s. kivshar , `` azimuthons : spatially modulated vortex solitons , '' phys .
rev .
lett . * 95 * , 203904 ( 2005 ) .
p. coullet , l. gil , and f. rocca , `` optical vortices , '' opt . commun . * 73 * , 403408 ( 1989 ) . v. m. lashkin , `` two - dimensional multisolitons and azimuthons in bose - einstein condensates , '' phys .
rev .
a * 77 * , 025602 ( 2008 ) .
d. buccoliero , a. s. desyatnikov , w. krlikowski , and y. s. kivshar , `` spiraling multivortex solitons in nonlocal nonlinear media , '' opt . lett . * 33 * , 198200 ( 2008 ) .
s. lopez - aguayo , a. s. desyatnikov , and y. s. kivshar , `` azimuthons in nonlocal nonlinear media , '' opt .
express * 14 * , 79037908 ( 2006 ) .
d. buccoliero , a. s. desyatnikov , w. krlikowski , and y. s. kivshar , `` laguerre and hermite soliton clusters in nonlocal nonlinear media , '' phys .
rev .
lett . *
98 * , 053901 ( 2007 ) .
a. minovich , d. n. neshev , a. s. desyatnikov , w. krlikowski , and y. s. kivshar , `` observation of optical azimuthons , '' opt .
express * 17 * , 2361023616 ( 2009 ) .
v. i. kruglov , y. a. logvin , and v. m. volkov , `` the theory of spitral laser beams in nonlinear media , '' j. mod . opt . * 39 * , 22772291 ( 1992 ) .
d. v. skryabin and w. j. firth , `` dynamics of self - trapped beams with phase dislocation in saturable kerr and quadratic nonlinear media , '' phys .
rev .
e * 58 * , 39163930 ( 1998 ) .
d. suter and t. blasberg , `` stabilization of transverse solitary waves by a nonlocal response of the nonlinear medium , '' phys .
rev .
a * 48 * , 45834587 ( 1993 )
. o. bang , w. krlikowski , j. wyller , and j. j. rasmussen , `` collapse arrest and soliton stabilization in nonlocal nonlinear media , '' phys .
rev .
e * 66 * , 046619 ( 2002 ) .
w. krlikowski , o. bang , n. i. nikolov , d. neshev , j. wyller , j. j. rasmussen , and d. edmundson , `` modulational instability , solitons and beam propagation in spatially nonlocal nonlinear media , '' j. opt .
b : quantum semiclass . opt .
* 6 * , s288s294 ( 2004 ) .
d. briedis , d. petersen , d. edmundson , w. krlikowski , and o. bang , `` ring vortex solitons in nonlocal nonlinear media , '' opt .
express * 13 * , 435443 ( 2005 )
. s. lopez - aguayo , a. s. desyatnikov , y. s. kivshar , s. skupin , w. krlikowski , and o. bang , `` stable rotating dipole solitons in nonlocal optical media , '' opt .
lett . * 31 * , 11001102 ( 2006 ) .
s. skupin , m. grech , and w. krlikowski , `` rotating soliton solutions in nonlocal nonlinear media , '' opt .
express * 16 * , 91189131 ( 2008 ) .
f. maucher , d. buccoliero , s. skupin , m. grech , a. s. desyatnikov , and w. krolikowski , `` tracking azimuthons in nonlocal nonlinear media , '' opt .
quant .
electron . * 41 * , 337348 ( 2009 ) .
m. peccianti , k. a. brzdkiewicz , and g. assanto , `` nonlocal spatial soliton interactions in nematic liquid crystals , '' opt
. lett .
* 27 * , 14601462 ( 2002 ) . c. conti , m. peccianti , and g. assanto , `` route to nonlocality and observation of accessible solitons , '' phys .
rev .
lett . * 91 * , 073901 ( 2003 ) . c. conti , m. peccianti , and g. assanto , `` observation of optical spatial solitons in a highly nonlocal medium , '' phys .
rev .
lett . * 92 * , 113902 ( 2004 ) .
r. nath , p. pedri , and l. santos , `` soliton - soliton scattering in dipolar bose - einstein condensates , '' phys .
rev .
a * 76 * , 013606 ( 2007 ) .
v. m. lashkin , e. a. ostrovskaya , a. s. desyatnikov , and y. s. kivshar , `` vector azimuthons in two - component bose - einstein condensates , '' phys .
rev .
a * 80 * , 013615 ( 2009 ) .
a. g. litvak , v. mironov , g. fraiman , and a. yunakovskii , `` thermal self - effect of wave beams in plasma with a nonlocal nonlinearity , '' sov .
j. plasma phys .
* 1 * , 3137 ( 1975 ) . c. rotschild , o. cohen , o. manela , m. segev , and t. carmon , `` solitons in nonlinear media with an infinite range of nonlocality : first observation of coherent elliptic solitons and of vortex - ring solitons , '' phys
. rev .
lett . * 95 * , 213904 ( 2005 ) .
s. skupin , u. peschel , l. berg , and f. lederer , `` stability of weakly nonlinear localized states in attractive potentials , '' phys .
rev .
e * 70 * , 016614 ( 2004 ) .
r. w. boyd , _ nonlinear optics _ ( academic press , amsterdam , 2008 ) , 3rd ed .
n. kuzuu , k. yoshida , h. yoshida , t. kamimura , and n. kamisugi , `` laser - induced bulk damage in various types of vitreous silica at 1064 , 532 , 355 , and 266 nm : evidence of different damage mechanisms between 266-nm and longer wavelengths , '' appl . opt . * 38 * , 25102515 ( 1999 ) .
g. p. agrawal , _ nonlinear fiber optics _ ( academic press , singapore , 2009 ) , 4th ed . d. v. skryabin , j. m. mcsloy , and w. j. firth , `` stability of spiralling solitary waves in hamiltonian systems , '' phys .
rev .
e * 66 * , 055602 ( 2002 ) .
f. maucher , s. skupin , m. shen , and w. krlikowski , `` rotating three - dimensional solitons in bose - einstein condensates with gravitylike attractive nonlocal interaction , '' phys .
rev .
a * 81 * , 063617 ( 2010 ) .
y. s. kivshar and d. k. campbell , `` peierls - nabarro potential barrier for highly localized nonlinear modes , '' phys .
rev .
e * 48 * , 30773081 ( 1993 ) . |
the formation and evolution of galaxies remains one of the big questions in astronomy . in the currently favoured @xmath2cdm model ,
galaxies are built up over time via the accretion of smaller systems ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the picture is not wholly satisfactory and some parts of the milky way may have formed in an initial collapse of baryonic material , somewhat akin to the model of galaxy formation proposed by @xcite .
one firm prediction of the @xmath2cdm model is that this accretion of smaller systems should still be ongoing and that the milky way halo should contain a large number of satellite systems .
it has been suggested that , given the model , there are too few satellites actually within the milky way halo @xcite , although the extent of this discrepancy has been a matter of some debate .
many attempts have been made to resolve this issue ( e.g. * ? ? ?
* ) by correcting the measured velocity dispersion of the local satellites to a velocity dispersion of the dark matter components , while more recently @xcite used numerical methods to highlight a scenario whereby these low mass systems are evaporated at the epoch of re - ionization . in the meantime ,
more and more surveys are probing the halo of the milky way revealing structures which may point to the formation history of our galaxy . the tidal dismemberment of a dwarf galaxy as it falls through the milky way halo is a slow process , with extensive streams of tidal debris existing for long periods of time @xcite .
while ancient remnants have been identified , via phase - space analysis , in our own galactic neighbourhood @xcite , more extensive surveys of the galactic halo , such as the spaghetti survey @xcite and utilizing 2mass @xcite have concluded that there is only a single , major on - going accretion event , that of the sagittarius dwarf galaxy @xcite . while this accretion event is adding mass to the galactic halo and provides an important probe of the shape of the dark matter potential ( e.g. * ? ? ?
* ; * ? ? ?
* ) , the lack of other other major accretion events is somewhat disconcerting given the predictions from @xmath2cdm . as will be discussed in detail in @xmath3[ring ]
, the recently discovered monoceros ring ( mri ) can be interpreted as an additional on - going accretion event within the milky way . investigating the
density and extent of this structure is important when trying to fully understand the impact this type of event is having on the evolution of our galaxy both in the past and into the future .
if the mri is instead the outermost edge of the milky way , mapping the outer reaches of the disk will provide insight into the milky way s past . to this end , we have used the isaac newton telescope wide field camera to continue a campaign to detect this stellar population around the galactic plane mapping out the extent of the mri .
this paper presents the results of a wide - field camera survey of the extensive stellar population thought to represent a continuation of the previous work in this field .
the layout of the paper is as follows ; @xmath3[ring ] summarises the extant knowledge of the mri and the associated population of stars while @xmath3[obs ] describes the observational procedure and data reduction .
@xmath3[analysis ] describes the analysis procedure , and the conclusions of this study are presented in @xmath3[conclusions ] .
the first sign of a new structure in the galaxy , the monoceros ring , came from a study of colour selected f - stars drawn from commissioning data from the sloan digital sky survey [ sdss , @xcite ] .
obtained in a narrow strip around the celestial equator these revealed two major stellar overdensities in the galactic halo , consistent with an intersection of the streams of tidal debris torn from the sagittarius dwarf galaxy @xcite . accompanying these , however , was an additional significant overdensity in the direction of the galactic anti - centre , interpreted as being another tidal stream about the galaxy , just past the edge of the stellar disk , at a galactocentric distance of @xmath4kpc and with a thickness of @xmath5kpc , covering over @xmath6 of sky within @xmath7 of the galactic equator
. given the results of the study of the sdss , @xcite searched for the signature of this stellar population in fields obtained for the isaac newton telescope wide field camera ( int / wfc ) survey ( see * ? ? ?
identifying the stream as a distinct population in colour - magnitude diagrams ( cmds ) , this study found the stars to be extensively distributed ; over 100@xmath0 of the sky within @xmath8 of the galactic equator , suggesting that the stream completely rings the galaxy .
main sequence fitting reveals that the galactocentric distance to the stream varies between @xmath9kpc and @xmath10kpc , with an apparent scale - height of @xmath11kpc .
@xcite suggested that this stream represents debris of an accreting dwarf galaxy , although pointed out that the extant data did not rule out alternatives such as an outer spiral arm or unknown warp / flare of the galactic disk . in fact , they favoured the interpretation of it being a perturbation of the disc , possibly the result of ancient warps .
simultaneously , @xcite presented an analysis of a larger sdss catalog of halo stars obtaining a number of radial velocities in several fields over the stream , finding a velocity dispersion of @xmath12km / s , inconsistent with any known galactic component .
furthermore , these kinematics indicate that the stream possesses a prograde orbit about the milky way with a circular velocity of @xmath13km / s [ note : this revised value was presented in @xcite ] .
metallicity estimates from these spectra indicate the stars are relatively metal poor @xmath14\sim-1.6)$ ] .
@xcite also concluded the stream represents a cannibalized dwarf galaxy undulating about the edge of the galactic disk .
several other studies have focused upon this stream population ; using the 2-micron all sky survey ( 2mass ) @xcite identified m - giant stars beyond the disk of the milky way , consistent with the population detected in the optical . again , at a galactocentric distance ( r@xmath15 ) of @xmath16kpc , this arc of stars covers @xmath17 , with a higher metallicity than previously determined @xmath14=-0.4\pm0.3)$ ] .
@xcite extended this study , obtaining stellar velocities of m - giants selected from 2mass . confirming a velocity dispersion of @xmath10km / s , this study concluded the stream orbits the milky way in a prograde fashion on an orbit with very little eccentricity .
while this may seem problematic for accretion models preferring elliptical orbits , numerical simulations by @xcite suggest that tidal streams in the plane of the milky way can possess quite circular orbits ; these numerical studies , however , also suggest that the accretion event must be young , less than @xmath18gyr since its first perigalactic passage , or any coherent structure would have dissolved .
finally , @xcite noted five globular clusters which are apparently associated with the stellar stream , as well as @xmath9 outer , old stellar clusters , bolstering the argument that it represents an accreting dwarf galaxy .
@xcite also employed the 2mass catalogue to identify m - giants beyond the disk of the milky way . by considering the projected density of these stars
, this study uncovered north - south anisotropies in the density of m - stars , with arcs above and below the plane of the galaxy .
significantly , @xcite identified a strong overdensity of these stars , @xmath19 of them in roughly 20@xmath0 of the sky at @xmath20 , the constellation of canis major ( cma ) .
this is a similar number of m - stars to that seen in the sagittarius dwarf galaxy and @xcite similarly interpret this population of stars as a dwarf galaxy .
it is probably the progenitor , with an original mass of @xmath21 , of the extensive stream of stars around the edge of the disk of the milky way .
given its mass , if cma does represent an equatorial accretion event , it will , when fully dissolved , increase the mass of the thick disk by @xmath22 .
additional evidence for this accretion interpretation of the cma dwarf comes from @xcite who found the signature of the main body of the dwarf in the background to several galactic open clusters , with the analysis of this population suggesting that it is somewhat metal rich with an age of @xmath23 gyrs , although a blue plume indicates a younger population .
this study also identified several globular and open clusters associated with the dwarf .
recently , @xcite have shown that the globular clusters associated with cma possess their own age - metallicity relationship which is distinct from that of the galactic population .
furthermore , the clusters are smaller than expected , if drawn from the galactic population , strengthening the interpretation that they are of non - galactic origin .
, by re - analysing the 2mass data claim the overabundance of m - giants in cma is simply a signature of the warp in the milky way . in response ,
@xcite used 2df data of the canis major region to highlight the velocity disparities between the milky way disk stars and the m - giant overdensity stars .
this has been repeated more recently by @xcite suggesting the dwarf galaxy cma can now be tentatively accepted as a real entity despite the current debate . connecting cma to the mri
is more problematic due to gaps in the detection of the ring and poor kinematic knowledge .
a recently completed 2df kinematic survey @xcite of the canis major region may yield results in this area although currently any general conclusions linking it with the mri are speculative .
more information about halo substructure has been recently discovered as @xcite have identified another structure in triangulum - andromedae ( triand ) which extends much further out than the mri .
currently it is not known whether the two are related , although interestingly the detection of the mri by @xcite resides at the edge of this new structure in triand suggesting a connection . the latest evidence supporting the interpretation of the canis major dwarf galaxy has been presented by @xcite , showing the colour - magnitude diagram of a region ( 0.5@xmath0 x 0.5@xmath0 ) centred on ( _ l_,_b _ ) = ( 240,-8)@xmath0 , the overdensity proposed by @xcite . an upper limit to the heliocentric distance of the galaxy
is found to be [email protected] to [email protected] kpc ; this is comparable with previous estimates . by measuring the surface brightness and total luminosity of the dwarf galaxy an estimate of the mass range
is found to be 1.0@xmath25 m@xmath26 @xmath27 5.5@xmath2810@xmath29 .
the tightness of the main sequence they found contradicts the claims that the canis major overdensity is the signature of the galactic warp .
@xcite have completed extensive modelling of the mri and while not conclusively establishing a connection between canis major and the mri , their results are highly suggestive of such a link .
@xcite using all of the current information known about the mri , undertook thousands of simulations , prograde and retrograde , in an attempt to find a model which best fitted the data . while some retrograde models were marginally acceptable , their preferred model was a prograde orbit which includes multiple wraps of the milky way .
this allows for the scenario that the tidal stream has both near and far components .
@xmath3[123m19 ] discusses a field in which this phenomenon appears .
if the dwarf galaxy has completed more than one orbit then the stream must be a much older structure than previously assumed [ cf .
@xcite ] , raising the possibility that the newly discovered triand structure @xcite could be the distant arm of a multiply wrapped tidal stream .
the model of @xcite and the continuing work of those studying this new galactic feature are slowly piecing together this structure , although currently there is no real coverage around the entire galactic plane , these observations hopefully extend the knowledge of the mri and it s potential progenitor the canis major dwarf .
the data was obtained on the isaac newton telescope wide field camera ( int / wfc ) at roque de los muchachos in la palma , canary islands . mounted at the telescope prime focus , this covers 0.29 square degrees field per pointing , imaging onto four 4k@xmath282k ccds .
these possess 13.5@xmath30 m pixels , corresponding to a pixel scale of 0.33 arcsec per pixel .
nine survey regions were chosen , roughly equally spaced between _
_ = @xmath31 . to aid in determining the location of the fields , the model of @xcite ( see figure [ figaitoff ] )
was used to predict where we might expect the spread of debris from an equatorial accretion . for each galactic
longitude , two regions symmetrically placed above and below the galactic plane , were imaged .
each target region is a composite of overlapping fields , the number of which depended on time available to observe in each region ( details of the observations are summarised in table [ obstable ] and presented graphically in figure [ figaitoff ] ) , but typically with a total area of @xmath5 square degrees . with the representative integration times , the limiting magnitudes are on average for these observations @xmath32 23.3 , @xmath33 22.2 , @xmath34 23.8 and @xmath35 22.8 .
archival data of ( 123,-19)@xmath0 , taken from the m31 survey @xcite , is used as the opposing field to ( 118,+16)@xmath0 .
de - biassing and trimming , vignetting correction , astrometry and photometry were all undertaken with the casu data reduction pipeline @xcite .
the flat fielding employed a master twilight flat generated over the entire observing run .
each star was individually extinction corrected using the dust_getval.c program supplied by schlegelschlegel / dust / data / data.html ] which interpolates the extinction using the maps presented by @xcite .
observing several standard fields per night allows the calibration of the photometry to be determined , as described by @xcite , deriving the ccd zero - points .
these are consistent to within a few percent on photometric nights .
the data reduction pipeline produces a catalogue of all images in each colour - band . rejecting non - stellar images
, the catalogues can be cross - correlated and the colour for individual stars can be determined .
near the limiting magnitude , however , galaxies can appear stellar and so galactic contamination is expected for the faintest sources .
[ cols=">,^,^,^,^,^,^,^ , < " , ] [ disttable ]
bcc would like to thank his wife , lll , for kindly supplementing his scholarship income , the university of sydney for the university postgraduate award and the cambridge astronomical survey unit at cambridge university and mike irwin for their hospitality during my week there .
bcc would also like to thank jorge pearrubia , for access to his monoceros ring model and the anonymous referee for their many helpful suggestions .
gfl acknowledges the support of the discovery project grant dp0343508 .
the research of amnf has been supported by a marie curie fellowship of the european community under contract number hpmf - ct-2002 - 01758 .
gfl would also like to thank triple j for their chillout session on sunday mornings which drowns out his fighting children and also their three hours of power in case they choose to fight between the hours of 11 and 1 at night . | we present the results of a wide - field camera survey of the stars in the monoceros ring , thought to be an additional structure in the milky way of unknown origin . lying roughly in the plane of the milky way , this may represent a unique equatorial accretion event which is contributing to the thick disk of the galaxy . alternatively , the monoceros ring may be a natural part of the disk formation process . with ten pointings in symmetric pairs above and below the plane of the galaxy , this survey spans 90 degrees about the milky way s equator .
signatures of the stream of stars were detected in three fields , ( _ l_,_b _ ) = ( 118,+16)@xmath0 and ( 150,+15)@xmath0 plus a more tentative detection at ( 150,-15)@xmath0 .
galactocentric distance estimates to these structures gave @xmath117 , @xmath117 , and @xmath113 kpc respectively .
the monoceros ring seems to be present on both sides of the galactic plane , in a form different to that of the galactic warp , suggestive of a tidal origin with streams multiply wrapping the galaxy .
a new model of the stream has shown a strong coincidence with our results and has also provided the opportunity to make several more detections in fields in which the stream is less significant .
the confirmed detection at ( _ l_,_b _ ) = ( 123,-19)@xmath0 at @xmath114 , kpc from the galactic centre allows a re - examination revealing a tentative new detection with a galactocentric distance of @xmath121 kpc .
these detections also lie very close to the newly discovered structure in triangulum - andromedae hinting of a link between the two .
the remaining six fields are apparently non - detections although in light of these new models , closer inspection reveals tentative structure . with the overdensity of m - giant stars in canis major
being claimed both as a progenitor to the monoceros ring and alternatively a manifestation of the milky way warp , much is still unknown about this structure and its connection to the monoceros ring .
further constraints are needed for the numerical simulations to adequately resolve the increasingly complex view of this structure .
latexl-.36em.3ex-.15em t-.1667em.7ex-.125emx [ section ] [ firstpage ] galaxy : formation galaxy : structure galaxies : interactions |
probably the most attractive feature of string theory is that it incorporates both quantum gravity and gauge interactions in a consistent framework .
a consistent string model should therefore have the appropriate particle content and interactions such that all apparent field - theoretic gauge and/or gravitational anomalies are absent . in four dimensional string models
, there can be @xmath5 gauge symmetries which are pseudo - anomalous .
however , the green - schwarz mechanism @xcite ensures that counterterms are generated from the exchange of pseudoscalar fields and any field - theoretic @xmath5 triangle anomaly is canceled by the corresponding non - trivial @xmath5 transformation of the pseudoscalars .
the analysis of anomalous @xmath6 have mainly be focused on perturbative heterotic string theory @xcite where many semi - realistic string models can be easily constructed . in recent years , however , tremendous progress has been made in constructing four - dimensional @xmath7 type i string models @xcite .
moreover , it was recently suggested that if the standard model particles are localized on a set of @xmath8-branes where @xmath9 whereas gravity lives in higher dimensional spacetime , then the string scale can be as low as a few tev @xcite .
type i string theory provides a natural framework to realize this `` brane world '' scenario @xcite . in view of these developments , it is therefore important to carry out similar analysis for type i string vacua .
anomaly cancelations in some four - dimensional @xmath7 type i string models were studied in some recent papers @xcite .
anomaly cancelations in @xmath10 type iib @xmath11 orientifolds without the b - flux background were studied in @xcite .
the analysis involves techniques in computing couplings of the twisted moduli with the gauge fields which were also recently studied in @xcite .
these simple examples illustrate some new features in type i string vacua that are distinct from that of the perturbative heterotic string .
first , the r - r scalars which participate in the green - schwarz mechanism come from the twisted sectors which can couple differently to different gauge factors .
therefore , in contrast to perturbative heterotic string , there can be more than one anomalous @xmath6 in these type i vacua
. moreover , the fayet - iliopoulos term already appears at tree - level and is dependent on the blowing - up modes of the orbifolds . in this paper
, we consider orientifold models with a non - vanishing ns - ns sector b - field background @xcite . although the fluctuations of the ns - ns b - field are projected out of the orientifold spectrum ( since it is odd under worldsheet parity reversal ) , a quantized vacuum expectation value of this b - field is allowed .
furthermore , the presence of this b - field gives rise to some novel features .
in particular , the rank of the gauge group is reduced by @xmath4 @xcite where @xmath1 is the rank of the background b - field . moreover , in models where there are both @xmath12 and @xmath13-branes , the @xmath14 sector states come with a multiplicity of @xmath4 @xcite . these reduced rank
orientifold models are dual to the chl strings @xcite in heterotic string theory . for models with both @xmath12 and @xmath13-branes ,
the corresponding heterotic duals contain ns fivebranes .
therefore , the analysis of these orientifolds may shed some lights on the non - perturbative properties of the chl strings . following the work of sen @xcite
, we also expect a close relation between orientifolds and f theory vacua . in ref @xcite , some eight - dimensional f theory vacua with non - zero background b - flux
were studied .
naively , the presence of this ns - ns sector b - field background is incompatible with the @xmath15 symmetry ( the s - duality ) of type iib theory .
however , the presence of this b - flux freezes 8 of the moduli of the elliptic @xmath16 , leaving only a 10-dimensional subspace .
the monodromy group on this subspace is reduced from @xmath15 to the congruence group @xmath17 which is the largest subgroup of @xmath18 that keeps the b - flux invariant . in this paper , we consider models with both @xmath12 and @xmath13-branes . they can be viewed as generalizations of f theory compactifications of this kind to lower dimensions when there are more than one type of 7-branes . on the other hand , this class of models is closely related to the setup considered in @xcite . according to @xcite ,
the positions of the branes cease to commute in the presence of a b - field background .
it is possible that our results here can be understood from the point of view of noncommutative geometry .
type iib orientifolds with quantized background b - field are also interesting from the phenomenological point of view . in the absence of b - field ,
the residual gauge symmetries are typically too large for the models to be phenomenologically interesting . since the rank of the gauge group is reduced by @xmath4 in the presence of b - field , it is possible to construct string models containing the standard model with fewer additional gauge symmetries .
in fact , the three - family pati - salam like model in ref .
@xcite was constructed by turning on a non - zero background b - field with @xmath19 in a @xmath20 orientifold .
there are both @xmath12 and @xmath13-branes in this model . under t - duality
, they become 2 sets of @xmath13-branes : @xmath21 and @xmath22 whose intersection is our four - dimensional non - compact spacetime .
the standard model @xmath23 lives on the @xmath21-branes and the three chiral families of fermions in the standard model come from open strings with at least one end attached to the @xmath21-branes .
one of the families comes from @xmath24 open strings whereas the other two families are @xmath25 open string states .
the fact that there are two families in the @xmath25 sector depends crucially on the multiplicity of @xmath26 in the @xmath25 sector .
since the main new features in models with non - zero b - field that we discuss here already appear in six dimensions , we will focus our attention to six - dimensional models in this paper .
moreover , as we will see , anomaly cancelations in six dimensions provide rather stringent constraints on the consistencies of these models .
in particular , we will show that the multiplicity of @xmath4 in the @xmath14 sector is crucial for the cancelation of the leading anomalies ( _ i.e. _ , the @xmath27 and @xmath28 terms ) .
moreover , the remaining anomalies are properly factorized which can then be canceled by the green - schwarz mechanism .
the analysis of four - dimensional @xmath7 orientifold models with background b - field will appear in a separate publication @xcite .
this paper is organized as follows . in section [ 6d ]
, we describe in detail the construction of some six - dimensional orientifold models with non - zero background b - field considered in ref.@xcite . in section [ anomaly ] , we discuss anomaly cancelations in these models .
we end with some discussion in section [ discussion ] .
some of the technical details are relegated to the appendices .
in this section , we discuss in detail the construction of six - dimensional type iib orientifolds with background b - flux @xcite . in particular ,
we consider the orbifold limits of k3 : @xmath0 where @xmath29 .
since we are interested in models with both @xmath12 and @xmath13-branes , we only consider the cases @xmath30 .
however , turning on the ns - ns antisymmetric two - form background seems to render the @xmath31 type iib perturbative orientifold inconsistent type iib orientifold with rank two b - flux does not contain states @xmath32 and @xmath33 given in table ii of @xcite . as a result , the @xmath28 and @xmath27 anomalies do not cancel .
however , it is consistent @xcite to turn on b - field in the @xmath34 orientifold in ref.@xcite since the corresponding @xmath35 twist is different . ] , so we will concentrate on the @xmath36 and @xmath37 models .
type i vacua on smooth k3 with non - zero @xmath38-field have been studied in @xcite .
toroidal compactification of type iib string theory on a four dimensional torus @xmath2 gives rise to a six dimensional model with @xmath39 supersymmetry . by gauging the world - sheet parity @xmath40 @xmath41 of type iib strings reduces by half the number of supersymmetries .
one can further reduce the number of supersymmetries to @xmath42 by orbifolding .
specifically , the @xmath11 orbifold action is realized by powers of the twist generator @xmath43 ( @xmath44 ) which can be written in the form @xmath45 where @xmath46 are @xmath47 cartan generators and @xmath48 represents the twist . in terms of the complex bosonic coordinates @xmath49 , @xmath50 that parametrize the torus , @xmath51 acts diagonally as @xmath52 similarly , we define complex fermionic fields @xmath53 as @xmath54 and @xmath55 . to derive the massless spectra of the orientifolds
, we will work in light - cone gauge .
for example , in the closed untwisted sector the ns massless states are @xmath56 which are invariant under @xmath51 , and @xmath57 which transforms as ^k _ -^i = e^2ikv_i _ -^i .
[ thpsi ] complex conjugates @xmath58 transform with a phase @xmath59 .
the untwisted massless ramond states are of the form @xmath60 with @xmath61 . to implement the gso projection we retain only the states with @xmath62 .
these states transform as ^k |s_0 s_1 s_2 s_3 = e^2ikvs |s_0 s_1 s_2 s_3 .
[ twistr ] the close string spectrum is obtained by retaining only those states in the untwisted sector which are invariant under the orientifold group action and by including twisted sector states .
this will be discussed in more detail in sec .
[ closedsector ] .
although type iib theory is a theory of closed strings , the orientifold projection requires both closed and open string sectors for consistency .
the klein bottle amplitude ( which is present due to the orientifold projection ) in the closed string sector generically gives rise to tadpole divergences .
these divergences are canceled by the new contributions from the open string sector @xcite . alternatively , orientifold fixed planes are sources for the ramond - ramond ( r - r ) @xmath63-forms .
charge cancelation can be generically achieved by including the right number of d@xmath8-branes , carrying opposite charge with respect to these forms @xcite .
the endpoints of an open string are labeled by @xmath64 and @xmath1 which lie on a @xmath65 and @xmath66-brane respectively ( the corresponding excitations are @xmath67 sector states ) .
the models we discuss here have both d9 and d5-branes .
the open string sector of the models will be discussed in sec .
[ opensector ] .
the open string spectrum of type iib orientifolds on @xmath0 are determined by the type and the number of d - branes necessary to cancel the tadpole divergences in the closed string klein bottle amplitude .
the worldsheet parity element @xmath68 of the orientifold group produces the tadpole which must be canceled by the 9-branes . for even @xmath3 , the closed string sector tadpole from @xmath69 orientifold element
is canceled by introducing @xmath70-branes , whose worldvolume spans the uncompactified @xmath10 space - time .
let @xmath71 be a world - sheet excitation and @xmath72 represent chan - paton indices associated with the string endpoints on d@xmath8-brane and d@xmath73-brane ( @xmath74 ) .
these chan - paton indices must be contracted with a hermitian matrix @xmath75 .
the action of the group elements on the open string state @xmath76 is given by @xcite ^k : & & | , ab ( _ k , p)_aa |^k.,ab ( _ k , q)_bb^-1 , ^k : & & | , ab ( _ _ k , p)_aa
|^k.,ba ( _ _ k , q)_bb^-1 , [ groupaction ] where @xmath77 and @xmath78 are unitary matrices associated with @xmath79 . in order to be consistent with group multiplication ( [ groupaction ] ) , we can choose _ _ k , p = _ k , p_,p , [ gamom ] and _ k , p=_1,p^k .
[ gamk ] the absence of pure gauge twists requires @xmath80 @xcite . because @xmath81 , _ 1,p^n=1 .
[ gton ] otherwise @xmath82 would be a pure gauge twist .
similarly , from @xmath83 , _ _ k , p=_2k , p ^t__k , p .
[ orcon1 ] finally , using eqs .
( [ gamom ] ) , ( [ gamk ] ) , ( [ orcon1 ] ) and the unitarity of the @xmath84 matrices we obtain _ k , p^ * = _ , p^ * _ k , p _ , p .
[ afam ] since different types of branes are present it is also necessary to consider the action of @xmath85 on @xmath67 states . in ref.@xcite , it was argued that @xmath86 on @xmath87 states .
this implies that in eq .
( [ orcon1 ] ) there are opposite signs for 9 and 5-branes . since a simultaneous @xmath88-duality on @xmath89 interchanges 9 and 5- branes along with @xmath90 and @xmath91 , _ _ 0,p=^t__0,p _ _ n/2,p=^t__n/2,p , [ tduality ] where the various signs are now correlated . from eqs .
( [ orcon1])-([tduality ] ) it further follows that _ 1,p^n=-1 .
[ gton1 ] for both 9 and 5- branes .
consider first the 99 open string states .
the massless ns states include gauge bosons @xmath92 and matter scalars @xmath93 .
the chan - paton matrices must be such that the full states are invariant under the action of the orientifold group .
hence , ^(0 ) & = & _ 1,9^(0)^-1_1,9 ; ^(0 ) = -_,9 ^(0)^t ^-1_,9 , + ^(i ) & = & e^2i v_i_1,9^(i)^-1_1,9 ; ^(i ) = -_,9^(i)^t^-1_,9 .
[ cons99 ] the massless ns states of the 55 open strings also include gauge bosons @xmath92 and scalars @xmath93 .
we consider models where all @xmath70-branes sit at the origin of the orientifold .
the chan - paton matrices then satisfy ^(0 ) & = & _ 1,5^(0)^-1_1,5 ; ^(0 ) = -_,5 ^(0)^t ^-1_,5 , + ^(i ) & = & e^2i v_i_1,5 ^(i ) ^-1_1,5 ; ^(i ) = _ , 5 ^(i)^t ^-1_,5 .
[ cons55 ] the sign change in the @xmath40 projection is due to the dd boundary conditions on the directions transverse to the @xmath70-branes . finally , consider @xmath14 open string states .
for @xmath70-branes , the compact coordinates obey mixed dn boundary conditions and have expansions with half - integer modded creation operators . by world - sheet supersymmetry
their fermionic partners in the ns sector are integer modded .
their zero modes form a representation of a clifford algebra and can be labeled as @xmath94 , @xmath95 , with @xmath96 .
the gso projection further imposes @xmath97 .
since @xmath94 are invariant under @xmath98 , the orientifold projection on these states implies @xmath99 notice that here the index @xmath64 ( @xmath1 ) lies on a 5-brane ( 9-brane ) .
@xmath100 relates @xmath14 with @xmath87 sectors and does not impose extra constraints on @xmath101 .
the perturbative spectrum of type iib theory contains two massless antisymmetric tensor fields : one coming from the ns - ns sector , and another coming from the r - r sector . under the world - sheet parity reversal ,
the ns - ns two - form @xmath102 is projected out , while the r - r two - form @xmath103 is kept .
although the fluctuations of @xmath104 ( the components of @xmath102 in the compactified dimensions ) are projected out of the perturbative unoriented closed string spectrum , a quantized vacuum expectation value of @xmath104 is allowed . to see this ,
consider the left- and right - moving momenta in the @xmath105 dimensions compactified on a torus @xmath106 : @xmath107 where @xmath108 and @xmath109 are integers , @xmath110 are constant vielbeins such that @xmath111 is the constant background metric on @xmath106 , @xmath112 are compactification radii of the @xmath2 , and @xmath113 . in ( [ bmomentum ] ) @xmath114 .
note that the components of @xmath115 are defined up to a shift @xmath116 ( which can be absorbed by redefining @xmath117 ) . with this normalization ,
only the values @xmath118 and @xmath119 are invariant under @xmath68 , and hence the vacuum expectation values of @xmath115 are quantized .
let @xmath120 ( @xmath121 ) be the rank of ns - ns antisymmetric tensor , it is clear that @xmath122 .
the unitary matrices @xmath77 and @xmath123 must be chosen so that ns - ns tadpoles from the mbius strip ( ms ) and the cylinder ( c ) of the open string sector cancel the ns - ns tadpole of the closed string sector coming from the klein bottle ( kb ) .
the specific form of @xmath77 and @xmath123 , and so the massless open string sector of the orientifold , is sensitive to the ns - ns antisymmetric tensor background .
consider first the case of zero b - flux , @xmath124 .
the 1-loop vacuum amplitudes of the models are given in appendix [ 1loopb0 ] .
the @xmath125 divergences of the kb , c and ms ( c.f .
, eqs.([kamp]),([camp ] ) and ( [ mamp ] ) ) produce tadpoles for the untwisted sector r - r 6- and 10-form as well as the twisted sector r - r 6-form .
relating the tree channel modulus @xmath126 to the kb , ms and c loop moduli through t_=14 , t_=18 , t_=12 , [ tvsl ] all the tadpoles vanish provided t^4/_2:&&_0,9=_0,5=32 , & & _ _ 0,9=^t__0,9 , _ _
0,5=-^t__0,5 , & & _ 1,9=_1,5=0 , t^4/_4:&&_0,9=_0,5=32 , & & _ _ 0,9=^t__0,9 , _ _
0,5=-^t__0,5 , & & _ 1,9=_2,9=_1,5=_2,5=0 .
[ bzero ] the solutions are given by t^4/_2:&&_1,9=_1,5=(i i_16,-i i_16 ) , t^4/_4:&&_1,9=_1,5=(i_8,^3 i_8 , i_8,^3 i_8),=e^i/4 .
[ bzerosolv ] where @xmath127 denotes an identity matrix of rank @xmath128 .
furthermore , we can choose @xmath129 to be purely real and @xmath130 purely imaginary .
( [ bzero ] ) then imply _ _ 0,9=,__0,5=. [ omegasol ] the tadpole cancelation ( [ bzero ] ) requires the introduction of 32 9-branes and 32 @xmath70-branes . in principle
, one can directly solve the constraints ( [ cons99])-([cons59 ] ) on the @xmath131 hermitian chan paton matrices to determine the open string spectrum of the model .
alternatively , when the @xmath101 matrices are written in a cartan - weyl basis , the constraints on the chan paton matrices become restrictions on the weight vectors . in this formalism @xcite
, the 99-sector gauge bosons correspond to both the cartan generators which trivially satisfy the @xmath132 constraint , together with charged generators belonging to a subset of @xmath133 root vectors selected by @xmath134 where @xmath135 is the `` shift '' vector of the gauge twist @xmath136 _ 1,9= e^-2iv_(99 ) h , [ vdef ] and @xmath137 are @xmath133 roots .
( here , underlining indicates that all permutations must be considered ) . in eq .
( [ vdef ] ) @xmath138 is a vector of @xmath133 cartan generators @xmath139 and @xmath135 is determined from ( [ bzerosolv ] ) to be t^4/_2:&&v_(99)=14(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1 ) , t^4/_4:&&v_(99)=18(1,1,1,1,1,1,1,1,3,3,3,3,3,3,3,3 ) .
[ 99shift ] similarly , from the equation for @xmath140 in ( [ cons99 ] ) it follows that matter states correspond to charged generators with @xmath141 since @xmath142 is antisymmetric , @xmath68 projection in the 55-sector constraints gauge bosons to both the cartan generators plus charged generators of @xmath143 satisfying ^a v_(55)= 0 mod , [ v5p ] where @xmath144 include all the @xmath133 roots plus long roots @xmath145 and @xmath146 .
matter fields in the 55-sector correspond to charged generators with ^a v_(55)&= & v_i mod .
[ m5p ] note that only a subset of short roots is used in determining the matter content of the 55-sector .
this is due to the extra minus sign in ( [ cons55 ] ) compare to the corresponding projection in ( [ cons99 ] ) .
finally , the 95 sector is handled using an auxiliary @xmath147 algebra . since we have generators acting simultaneously on both 9-branes and 5-branes , only roots of the form w _
( 95)= w_(9)w_(5)= (; ) [ w95def ] must be considered . here
the first ( second ) @xmath148 components transform under @xmath149 ( @xmath150 ) . the shift in this sector
is defined to be @xmath151 . from ( [ cons59 ] )
we learn that massless states correspond to @xmath152 roots satisfying @xmath153 the open string spectrum of the @xmath0 orientifolds for @xmath124 is included in table [ 6dtable1 ] .
our convention for the @xmath6 charges , as explained in appendix [ lambdatraces ] , differs from that in @xcite .
consider now the tadpoles of @xmath0 orientifolds with non - zero b - flux , @xmath154 .
tadpoles with a volume factor @xmath155 , correspond to the untwisted r - r 10-form exchange in the closed string tree channel .
all three surfaces kb , ms and c contribute to the tadpole .
the kb contribution comes from the @xmath156 untwisted closed string sector .
since the parity projection @xmath68 in this sector requires all the windings to be zero , from ( [ bmomentum ] ) we conclude that the compact momentum sum does not change , and the overall tadpole is the same as in the absence of the @xmath104 : ^10-form_u=(1 - 1)i_0^ dtt^2 ( 32n v_6 v_t^4 ^ 5 ) .
[ bklien ] the c contribution comes from the @xmath156 99-string sector .
the partition function in this sector involves the summation over the 99-string momentum modes . upon poisson resummation , which is necessary in order to extract the tadpole , the momentum sum in the open string loop channel
is reinterpreted as a sum over the closed string winding modes in the tree channel . unlike the momentum states which remain unaffected by the @xmath104 background , only winding states with even winding numbers are allowed along the compact directions with @xmath157 .
indeed , closed string winding states satisfy p_l =- p_r , [ winding ] which with the quantization ( [ bmomentum ] ) and for a generic choice of metric on @xmath2 implies n_i - b_ijm^j=0 .
[ windingquant ] eq .
( [ windingquant ] ) requires @xmath158 for compact directions @xmath159 such that @xmath160 . even winding numbers in the closed tree channel further
imply an effective reduction by half the basis momentum lattice vectors along these directions compare to the @xmath124 case @xcite . in the nontrivial @xmath104 background ,
the c tadpole is thus increased by @xmath161 ^10-form_u = i(1 - 1)_0^ dtt^2 ( 2^b16n v_6 v_t^4 ^ 5 ( _ 0,9)^2 ) .
[ bc ] alternatively , the @xmath161 factor of the c tadpole can be interpreted as a sum over @xmath161 different sub - lattices within the @xmath124 case unit volume , shifted with respect to each other by half the generating vectors . a priori
, the ms tadpole can have a different @xmath68 projection @xmath162 , depending on a specific sub - lattice .
in fact , the tadpole factorization requires that the ms contribution increases by @xmath4 , which is achieved by choosing the same @xmath68 projection for @xmath163 sub - lattices and the opposite projection for the other @xmath164 sub - lattices ^10-form_u = i(1 - 1)_0^ dtt^2 ( ( _ 0,9 ) ) , [ bm ] where a choice of the sign correlates with the choice of sign in @xmath40 projection for the larger set of momentum sub - lattices . using the loop / tree channel moduli translation ( [ tvsl ] ) ,
the full tadpole of the untwisted 10-form becomes [ + + ] ^10-form_u = i(1 - 1)_0^ d .
[ full10 ] tadpoles with a volume factor @xmath165 represent the untwisted r - r 6-form exchange in the closed string tree channel .
the kb contribution comes from the @xmath166 untwisted closed string sector . in this sector
the parity projection requires all the momenta to be zero and the ns - ns background allows only even windings along the compact directions with @xmath160 .
the net effect of the nontrivial background is thus a reduction of the corresponding @xmath124 tadpole by @xmath161 ^6-form_u=(1 - 1)i_0^ dtt^2 ( 32n2^b v_64 ^ 2 v_t^4 ) .
[ bklien6 ] since @xmath167 tadpole measures the square of the total r - r charge of the sixteen orientifold 5-planes , its reduction compare to the zero ns - ns background tadpole implies that @xmath168 @xmath169-planes have the same charge as in the @xmath124 case while the other @xmath170 @xmath169-planes turn into @xmath171-planes which carry r - r charge opposite to that of the @xmath169-plane @xcite .
the c contribution comes from the @xmath156 55-string sector . here , the partition function sums over the 55-string winding modes .
the 55-string winding modes in the open string loop channel are reinterpreted upon the poisson resummation as momentum modes in the closed string tree channel . since the latter are insensitive to the ns - ns tensor background , the c tadpole is unchanged ^6-form_u = i(1 - 1)_0^ dtt^2 ( 116n v_64 ^ 2 v_t^4 ( _ 0,5)^2 ) .
[ bc6 ] the ms amplitude relevant to the 6-form exchange in the tree channel is interpreted as a closed string exchange between 5-branes and sixteen orientifold 5-planes , and so is proportional to the total r - r charge of the orientifold 5-planes . it is suggested by ( [ bklien6 ] ) that the absolute value of the r - r charge of the orientifold 5-planes is reduced by @xmath4 , so ^6-form_u = i(1 - 1)_0^ dtt^2 ( v_64 ^ 2 v_t^4 ( _ 0,5 ) ) .
[ bm6 ] where a choice of a sign correlates with the choice of a sign in @xmath40 projection in the 55-open string sector .
the full tadpole of the untwisted 6-form in terms of the tree channel cylinder modulus @xmath126 ( [ tvsl ] ) becomes [ + + ] ^6-form_u = i(1 - 1)_0^ d v_64 ^ 2 v_t^4 .
[ full6 ] the untwisted tadpole cancelation from ( [ full10 ] ) and ( [ full6 ] ) requires _ 0,9&=&_0,5=322^b/2 , [ gammatraceun ] and the same @xmath68 projection for both the larger set of the momentum sub - lattices in the 99-open string sector and the 55-open string sector as in the @xmath124 case : @xmath172 since traces of @xmath173 count the number of d - branes , ( [ gammatraceun ] ) implies the rank reduction of a gauge group in the open string sector when the b - flux is turned on @xcite .
in addition to the untwisted r - r 6- and 10-form tadpoles which are canceled as discussed above , the orientifolds of interest have potential tadpoles corresponding to the twisted sector r - r 6-form exchange in the tree channel .
these tadpoles come from the cylinder vacuum amplitudes @xmath174 , ( [ camp ] ) with @xmath175 and so do not get contributions from momentum and winding modes on @xmath2 .
since the nonzero ns - ns tensor background effects only the quantization of the momentum and winding modes on @xmath2 , one would expect the twisted r - r 6-form tadpole to be the same as in the @xmath124 case . as we will discuss in detail in sec .
[ anomaly ] , anomaly cancelations in six dimensions requires the @xmath4 multiplicity of massless states in the 59 open string sector . in type
iib orientifolds , gauge degrees of freedom are carried by the chan - paton indices , so the multiplicity of massless states is extended to the multiplicity of chan - paton indices in the 59-sector and thus the multiplicity of massive states in this sector as well .
the latter implies that the cylinder vacuum amplitudes @xmath176 , @xmath177 are unaffected by the b - flux while @xmath178 and @xmath179 get an extra factor of @xmath4 .
this raises a question : once the relative factor of @xmath176 and @xmath178 vacuum amplitudes is changed in the nonzero ns - ns background , how one maintains the factorization of the twisted r - r 6-form tadpoles ? recall that going to the tree channel interpretation in the kb vacuum amplitude we related ( for @xmath124 ) the tree channel modulus @xmath126 and the loop channel modulus @xmath180 through @xmath181 .
this kb tree channel `` cylinder '' has length @xmath126 and two crosscaps .
the c and ms vacuum amplitudes of the open string sector in the tree channel picture were represented by an identical length cylinder and a cylinder with a crosscap .
the fact that we used the same tree channel modulus for c and ms implicitly assumed the ability to relate kb , c and ms amplitudes to each other by adding / removing crosscaps .
this is possible only for @xmath176 and @xmath177 since only they have the same type boundaries , which allows putting the crosscaps .
the above arguments do not fix the relation between @xmath126 and @xmath182 .
we claim that this relation must be fixed by demanding anomaly cancelation along with factorization of twisted r - r 6-form tadpoles .
thus we require = 2^b-1t__59 .
[ tc59 ] the twisted r - r 6-form tadpoles vanish provided t^4/_2 : & & _ 1,9=_1,5=0 , t^4/_4:&&_1,9=_2,9=_1,5=_2,5=0 .
[ bnonzero ] eqs .
( [ gammatraceun ] ) and ( [ bnonzero ] ) are solved by t^4/_2:&&_1,9=_1,5=(i i_2
^ 4-b ,- i i_2 ^ 4-b ) , t^4/_4:&&_1,9=_1,5=(i_2 ^ 3-b,^3 i_2 ^ 3-b , i_2 ^ 3-b,^3 i_2 ^ 3-b ) , = e^i/4 . [ bnonzerosolv ]
the `` shift '' vectors corresponding to the above gauge twists are t^4/_2:&&v_(99)=14 ( ^2 ^ 4-b times ) , t^4/_4:&&v_(99)=18 ( _ 2 ^ 3-b times , _ 2 ^ 3-b times ) .
[ shiftz4 ] with @xmath146 and @xmath183 . as for the @xmath124 case , we can choose @xmath129 purely real and @xmath130 to be purely imaginary .
( [ ompro ] ) then imply _ _
0,9= , _ _ 0,5=. [ omegasolun ] given ( [ bnonzerosolv ] ) and ( [ omegasolun ] ) , the massless open string states can be determined as in the @xmath124 case .
the open string spectra are given in table [ 6dtable1 ] .
in this section we discuss the closed string sector of type iib orientifolds on @xmath0 , ( @xmath184 ) , in the presence of b - flux .
first consider the @xmath36 orientifold models .
the untwisted sector of the models is constructed by keeping states of type iib orbifold on @xmath185 invariant under the world sheet parity @xmath40 .
this gives rise to a @xmath42 supergravity multiplet in six dimensions , accompanied by one tensor multiplet and four hypermultiplets .
the twisted sectors will produce additional multiplets .
the bosonic content of a hypermultiplet is four scalars transforming as @xmath186 under the six dimensional little group @xmath187 , while that of a tensor multiplet as @xmath188 .
there are overall 16 fixed points in the @xmath36 orbifold with an orientifold 5-plane sitting at each fixed point . as we explained in sec .
[ opensector ] , out of 16 orientifold 5-planes , each one of @xmath168 @xmath169-planes carries @xmath189 units of r - r charge while the other @xmath170 @xmath171-planes have opposite r - r charge .
this implies that precisely @xmath168 @xmath36 fixed points are even under the @xmath40 projection while the other @xmath170 @xmath36 fixed points pick up a minus sign under the @xmath40 projection . at each fixed point of the twisted sector ,
the ns and r sector massless states transform as @xmath190 the twisted sector spectrum is obtained by taking products of states from the left- and right - moving sectors .
the orientifold projection @xmath68 keeps symmetric ( antisymmetric ) combinations in the ns - ns sector and antisymmetric ( symmetric ) combinations in the r - r sector for each @xmath40-even ( @xmath40-odd ) @xmath36 fixed point .
this gives the bosonic content of a hypermultiplet for each @xmath40-even @xmath36 fixed point and the bosonic content of a tensor multiplet for each @xmath40-odd @xmath36 fixed point .
overall , the twisted sectors of @xmath36 orientifold contribute @xmath168 hypermultiplets and @xmath170 extra tensor multiplet ( see table [ 6dtable1 ] ) .
the untwisted sector of @xmath191 orientifold contains @xmath42 supergravity multiplet in six dimensions , accompanied by one tensor multiplet and two hypermultiplets .
the four @xmath37 invariant fixed points of @xmath191 orientifold give four hypermultiplets and four tensor multiplet .
they are also @xmath36 fixed points and so supply an additional 4 hypermultiplets .
the other 12 @xmath36 fixed points form 6 @xmath37 invariant pairs . out of 12 @xmath36 fixed points which are not @xmath37 invariant , @xmath170 fixed points are odd under the @xmath40 projection and the other @xmath192 are even .
thus they contribute @xmath193 extra tensor multiplets and @xmath194 hypermultiplets .
overall , twisted sectors of @xmath37 orientifold contribute @xmath195 extra tensor multiplets and @xmath196 hypermultiplets ( see table [ 6dtable1 ] ) .
an important check on the consistency of the @xmath42 six dimensional vacua constructed in the previous sections is provided by the cancelation of all gauge , gravitational , and mixed anomalies .
cancelation of tadpoles in type iib orientifolds on @xmath0 ( @xmath184 ) discussed in sec .
[ opensector ] guarantees 1-loop consistency of the equations of motion for the untwisted r - r 6- and 10-form as well as the twisted r - r 6-form .
recall that the tadpoles were extracted from the kb , ms and c vacuum amplitudes .
each of these amplitudes involves a summation over four spin structures ( _ i.e. _ , summation over @xmath197 in appendix [ 1loopb0 ] ) .
actually , the r - r spin structure , @xmath198 in appendix [ 1loopb0 ] , does not contribute to the 1-loop vacuum amplitudes because of the fermionic zero modes .
on the other hand , it will contribute to the 2-loop vacuum amplitudes , as well as to various scattering processes .
in fact , being odd under the spacetime parity , it is the only spin structure ( at 1-loop level ) that would generate potential chiral anomalies in the low - energy effective field theory . in this sense ,
the cancelation of six dimensional chiral anomalies follows from the absence of tadpoles coming from the r - r spin structure . in this section
we show that @xmath0 ( @xmath184 ) type iib orientifolds with b - flux are free from chiral anomalies .
first , we explain anomaly cancelation in the framework of the six dimensional effective field theory .
then we discuss how the green - schwarz terms can be computed from the corresponding non - planar diagrams in string theory . in the latter approach
we confirm the necessity of @xmath4 multiplicity of 59-cylinder diagrams .
some useful facts on six dimensional chiral anomalies are given in appendix [ anomalyreview ] .
we start with the @xmath185 orientifold models . from the spectrum of the models
given in table [ 6dtable1 ] it is straightforward to check using the results of appendix [ anomalyreview ] that the leading anomalies , _ i.e. _ , the @xmath28 and @xmath27 terms vanish . to see this explicitly ,
notice that the leading term in the pure gravitational anomalies is proportional to @xmath199 where @xmath200 is the number of hypermultiplets from both open and closed string sectors . from section [ closedsector ] , @xmath201 and
the number of closed string hypermultiplets is @xmath202 .
the gauge group is @xmath203 where @xmath204 , and hence @xmath205 .
the number of open string hypermultiplets is @xmath206 .
it is easy to see that the above term ( [ r4 ] ) vanishes .
notice that it is crucial for the cancelation of the @xmath28 gravitational anomaly that the multiplicity of 59-open string matter states is @xmath4 .
in addition , the hypermultiplet representations plus the @xmath4 multiplicity in the 59-sector ensure that the @xmath27 term is absent .
this can be seen from ( [ tracesrelations ] ) that the leading pure gauge non - abelian anomalies for each @xmath207 gauge group is proportional to @xmath208 which vanishes since the multiplicity of @xmath14 states is @xmath4 which is equal to @xmath209 .
the remaining gravitational @xmath210 gauge anomaly may be written in the form i(2)^3
i=-1162^b/2 x_4^(9)x_4^(5 ) + x_2x_6^(9)-_2x_6^(5 ) , [ z2anomaly ] where x_4^(a)&=&r^2 - 22^b/2 ( f_a^2+f_a^2 ) , x_6^(a)&=&16 , [ 46z2 ] and x_2&=&42^b/2f_9-f_5 , _ 2&=&42^b/2f_5-f_9 .
[ 2z2 ] the subscript @xmath211 refers to the gauge group factors in the 55- and 99-open string sectors models were discussed previously in @xcite .
the difference between the corresponding anomaly polynomials of ( [ 2z2 ] ) and the one given in @xcite is due to a different convention of the @xmath6 charges .
see table [ 6dtable1 ] for our convention . ] .
the anomaly polynomial ( [ z2anomaly ] ) contains terms that involve only nonabelian field strength as well as terms that also involve abelian field strength .
the former can be canceled by exchange of a two - form , which involves green - schwarz interactions of the form _
c.t.=i16 2^b/2 ( 2)^3b_2x_4^(5 ) .
[ counter1 ] the @xmath212 field is the 2-form with gauge invariant field strength @xmath213 where @xmath214 .
the counterterm ( [ counter1 ] ) arises from the tree - level diagram which involves coupling of the @xmath212 field and a pair of gauge bosons .
such couplings can be computed from the disk diagram in string theory . when @xmath154 , a disk diagram involving two gauge bosons and a two - form potential of the twisted sector tensor multiplet is proportional to @xmath215 , where @xmath216 refers to the 55- or 99-open string sector and @xmath217 denotes the chan - paton generator of the corresponding gauge boson .
using the results in appendix [ lambdatraces ] , this trace is zero .
thus we identify the 2-form @xmath212 as a combination of the two - form potential of the gravity multiplet with a self - dual field strength and the two - form potential with an anti - self - dual field strength of the untwisted sector tensor multiplet .
the remaining terms of the anomaly polynomial ( [ z2anomaly ] ) can be canceled by @xmath218 if we assign the anomalous transformation laws : @xmath219 under the @xmath220 gauge transformations @xmath221 . as in ( [ counter1 ] ) , the counterterm ( [ counterbl ] ) arises from a tree - level diagram .
this diagram involves the coupling of the @xmath222 scalars with a single @xmath220 gauge boson . from the disk diagram ,
the coupling of the untwisted sector r - r scalars to @xmath220 gauge boson is proportional to @xmath223 which vanishes , see appendix [ lambdatraces ] . on the other hand , the corresponding coupling of the twisted sector r - r scalars is proportional to @xmath224 .
consider first the @xmath124 case .
the number of @xmath36 twisted scalars @xmath225 is @xmath148 .
all of them couple with the same relative strength to the 99-sector @xmath6 , because @xmath226 ( which measures the coupling strength ) is the same for all fixed points .
since all d5-branes sit at the origin , only the twisted scalar at the origin ( which we denote by @xmath227 ) will couple .
this can be seen explicitly from the tadpole solution : the coupling of the scalar at a fixed point @xmath228 to a 55-sector gauge boson is proportional to @xmath229 , which is zero unless @xmath230 is the origin , where all the d5-branes are located .
therefore , b_0^(9)&=&_i=1
^ 16_i , b_0^(5)&= & _ 1 , [ b0z2 ] for some coefficients @xmath231 and @xmath232 to be determined . since @xmath227 couples both to 55-sector and 99-sector gauge fields , it transforms under @xmath6 gauge transformations as _ 1_1+a_1_9+a_2_5 , [ f1z2 ] for some coefficients @xmath233 and @xmath234 to be determined .
the other scalars @xmath235 , couple only to the 99-sector , but all in a same way , so _
[ fiz2 ] requiring that ( [ b0z2])-([fiz2 ] ) generate ( [ anomshift ] ) uniquely fixes all the coefficients : @xmath236 up to field normalization .
we have b_0^(9)&=&_i=1 ^ 16_i , b_0^(5)&=&-4 _ 1 , where _ 1 & & _ 1+(_9 - 4_5)/(4 ) , _ i & & _ i+_9/(4),i=2,,16 .
[ finalz2b0 ] since the @xmath225 s are in linear multiplets , the choice of normalization @xmath231 is irrelevant .
the choice @xmath237 reproduces the result of @xcite .
one can always attain the canonical normalization of @xmath238 and @xmath239 with an appropriate change of the coupling constants in the counterterm ( [ counterbl ] ) .
this is also generic to other orientifold models : a linear combination of twisted scalars that couple to the abelian gauge fields is unique , up to field normalization . in the @xmath19 case ,
the number of @xmath36 twisted scalars is @xmath240 . as in the previous case ,
one scalar is singled out as living at the fixed point where we put all d5-branes .
using the same logic as before , we find a unique solution b_0^(9)&= & _
i=1 ^ 12_i , b_0^(5)&=&-2 _ 1 , where _ 1 & & _ 1+(_9 - 2_5)/(2 ) , _ i & & _ i+3 _ 9/(22 ) , i=2,,12 .
[ finalz2b2 ] in the @xmath241 case , the number of @xmath36 twisted scalars is @xmath242 .
therefore , we find b_0^(9)&= & _ i=1 ^ 10_i , b_0^(5)&=&- _ 1 , where _ 1 & & _ 1+(_9-_5)/ , _ i & & _ i , i=2,,10 .
[ finalz2b4 ] we now consider @xmath191 orientifold models . from the open string spectrum of the models given in table [ 6dtable1 ] and the closed string spectrum in section [ closedsector ] , one can easily see that the @xmath28 and all @xmath27 terms vanish in the 8-form anomaly polynomial .
the remaining gravitational @xmath210 gauge anomaly may be written in the form i(2)^3
i=&&-1322^b/2 x_4^(9)x_4^(5 ) -2^b/28_4^(9)_4^(5)+ 18_4^(5)_4^(5)+ 18_4^(9)_4^(9 ) & & + x_2^(1,9)x_6^(1,9)+x_2^(3,9)x_6^(3,9 ) + x_2^(1,5)x_6^(1,5)+x_2^(3,5)x_6^(3,5 ) , [ anomz4 ] where x_4^(a)&=&r^2 - 22^b/2 ( f_1,a^2+f_3,a^2+f_1,a^2+f_3,a^2 ) , _ 4^(a)&=&f_1,a^2+f_1,a^2-f_3,a^2-f_3,a^2 , x_6^(,a)&=&1122^b/2r^2 f_,a-16f_,a^3 -12f_,a^2 f_,a-2^(2-b)/43 f_,a^3 , [ 46z4 ] and x_2^(1,9)&=&3 f_1,9-f_3,9 - 2^b/2 f_1,5 , x_2^(3,9)&=&3 f_3,9-f_1,9 - 2^b/2 f_3,5 , x_2^(1,5)&=&3 f_1,5-f_3,5 - 2^b/2 f_1,9 , x_2^(3,5)&=&3 f_3,5-f_1,5 - 2^b/2 f_3,9 .
[ 2z4 ] in the @xmath243 orientifolds there are two nonabelian gauge factors in the 99-sector and 55-sector .
one comes from the shift - vector @xmath244 components and the other come from the shift - vector @xmath245 components , ( c.f .
( [ shiftz4 ] ) ) . in ( [ anomz4 ] ) , we use this to label different gauge factors .
for example , the 99-sector nonabelian field strengths will be denoted by @xmath246 and @xmath247 .
the nonabelian factors in the 55-sector are denoted by @xmath248 and @xmath249 .
we use similar notation @xmath250 for the abelian factors .
the terms of the anomaly polynomial ( [ anomz4 ] ) that involve only nonabelian field strengths can be canceled by exchange of three two - forms with green - schwarz interactions _
c.t.=i32 2^b/2 ( 2)^3 b_2^(1)x_4^(5)+i 2^b/28 ( 2)^3 b_2^(2)_4^(5)-i 8 ( 2)^3 b_2^(3)_4^(9 ) , [ z42form ] where @xmath251 s have gauge invariant field strengths h^(1)&=&d b_2^(1 ) - x_3^(9),x_4^(9)=d x_3^(9 ) , h^(2)&=&d b_2^(2 ) - x_3^(9,5 ) , _
4^(9)-12^b/2_4^(5)=d x_3^(9,5 ) , h^(3)&=&d b_2^(3 ) - _ 3^(9 ) , _ 4^(9)=d _
[ 2fromtransform ] the counterterm ( [ z42form ] ) arises from a tree - level diagram which involves coupling of @xmath251 fields and a pair of gauge bosons .
recall that these models have a single tensor multiplet from the untwisted sector , @xmath105 tensor multiplets from the @xmath37 twisted sector and @xmath252 extra tensor multiplets coming from the @xmath36 twisted sector . by the same arguments as for the @xmath185 orientifolds ,
the @xmath36 twisted sector tensor multiplets do not couple to a pair of gauge bosons . on the contrary , the coupling of the @xmath37 twisted sector tensor multiplets from the disk diagram is nonzero .
with all the d5-branes at the origin , only the @xmath37 twisted tensor multiplet at the origin will couple to the 55-sector gauge bosons .
thus we expect that @xmath253 and @xmath254 are combinations of the untwisted sector two - form potentials of the gravity and the tensor multiplets plus two - form potentials of all four @xmath37 twisted sector tensor multiplets .
@xmath255 must contain the untwisted sector two - form potentials and a two - form potential from the @xmath37 twisted sector tensor multiplet at the origin .
the remaining abelian terms of the anomaly polynomial ( [ anomz4 ] ) may be canceled by additional counterterms @xmath256 if we assign the anomalous transformation laws : @xmath257 under the @xmath6 gauge transformations @xmath258 .
notice that in the presence of d - branes , the total twist of a non - vanishing @xmath3-point function does not have to be zero .
therefore , the anomalous @xmath6 gauge bosons can couple to closed string states from different twisted sectors ( in this case @xmath259 and @xmath260 twisted sectors ) . as a result , there are mixings in the kinetic terms of the @xmath259 and @xmath260 twisted sector states even when @xmath261 .
we now proceed to identify the scalars @xmath262 in terms of the twisted r - r states .
we start with the @xmath124 case .
there are in total 14 twisted scalars : 4 scalars ( we denote them by @xmath263 ) coming from the @xmath37 twisted sector , the other 10 scalars ( we denote them @xmath264 ) coming from the @xmath36 twisted sector . out of 10 @xmath36 twisted fields ,
4 come from the @xmath36 fixed points which are also @xmath37 fixed points , and the other 6 come from 6 pairs of the @xmath36 fixed points which are @xmath37 invariant .
let us analyze the coupling of these r - r twisted fields to the various @xmath6 gauge bosons .
all twisted fields couple to the 99-sector vector bosons .
these couplings can be computed from the disk diagram : for the @xmath37 twisted fields the coupling is proportional to @xmath265 while for the @xmath36 twisted fields the coupling is proportional to @xmath266 .
since & = & 1,&=&-1 , [ t1 ] the coupling of @xmath267 to the @xmath268 abelian gauge boson is identical to that of the @xmath269 , while the coupling of @xmath270 to @xmath268 is opposite to its coupling to @xmath269 . as in the case of @xmath36 orientifold , only twisted fields that live at the fixed point where the d5-branes are located will couple to the 55-sector gauge bosons .
we call these fields @xmath271 and @xmath272 . their coupling to different @xmath6
can be deduced as discussed above .
collecting the information about the coupling , we can easily generalize ( [ b0z2 ] ) to : b_0^(1,9)&=&_i=1 ^ 4 _ i^(4)+ _
j=1 ^ 10_j^(2 ) , b_0^(3,9)&=&_i=1 ^ 4 _ i^(4)- _
j=1 ^ 10_j^(2 ) , b_0^(1,5)&= &
_ 1^(4)+ _
1^(2 ) , b_0^(3,5)&= & _ 1^(4)- _
1^(2 ) , [ b0z4 ] where @xmath273 are coefficients to be determined . using the same logic as for the @xmath36 orientifolds , we assign the following gauge transformations to the twisted fields & & _
1^(4)_1^(4)+a_1_1,9+a_2_3,9+a_3_1,5 + a_4_3,5 , & & _ i^(4)_i^(4)+b_1_1,9+b_2_3,9 , i=2,4 , & & _ 1^(2)_1^(2)+c_1_1,9+c_2_3,9+c_3_1,5 + c_4_3,5 , & & _ j^(2)_j^(2)+d_1_1,9+d_2_3,9 , j=2,10 .
[ trans ] as we already mentioned , requiring that ( [ b0z4])-([trans ] ) satisfy ( [ anomshiftz4 ] ) , fixes uniquely all coefficients up to the field normalization .
( the system is overconstraint and so generically would not have solution at all . )
we end up with b_0^(1,9)&=&_i=1 ^ 4 _ i^(4)+ _
j=1 ^ 10_j^(2 ) , b_0^(3,9)&=&_i=1 ^ 4 _ i^(4)- _
j=1 ^ 10_j^(2 ) , b_0^(1,5)&=&-2
_ 1^(4)-4 _
1^(2 ) , b_0^(3,5)&=&-2
_ 1^(4)+4 _
1^(2 ) , where _ 1^(4 ) & & _
1^(4)+(_1,9+_3,9)/(4 ) -(_1,5 + _
3,5)/(2 ) , _ i^(4 ) & & _
i^(4)+(_1,9+_3,9)/(4 ) , i=2,4 , _
1^(2 ) & & _
1^(2)+(_1,9-_3,9)/(8 ) -(_1,5- _ 3,5)/(2 ) , _
j^(2 ) & & _
j^(2)+5(_1,9-_3,9)/(24 ) , j=2,10 .
[ finalz4b0 ] the appearance of two normalization constants in the final expression comes from an independent choice of normalizations for the @xmath37 and @xmath36 twisted fields . turning on rank two ns - ns two form field @xmath104 converts two of the @xmath36 twisted hypermultiplets into tensor multiples ,
so the total number of @xmath36 twisted scalars becomes @xmath274 .
repeating identical analysis as in the @xmath124 case , we find a unique solution ( up to field normalization ) b_0^(1,9)&=&_i=1 ^ 4 _ i^(4)+ _
j=1 ^ 8_j^(2 ) , b_0^(3,9)&=&_i=1 ^ 4 _ i^(4)- _
j=1 ^ 8_j^(2 ) , b_0^(1,5)&=&-
_ 1^(4)-2 _
1^(2 ) , b_0^(3,5)&=&-
_ 1^(4)+2 _
1^(2 ) , where _ 1^(4 ) & & _ 1^(4)+(_1,9+_3,9)/-(_1,5 + _ 3,5)/ , _
i^(4 ) & & _
i^(4 ) , i=2,4 , _
1^(2 ) & & _
1^(2)+(_1,9-_3,9)/(2 ) -(_1,5- _ 3,5)/ , _
j^(2 ) & & _
j^(2)+3(_1,9-_3,9)/(14 ) , j=2,8 .
[ finalz4b2 ] finally , the rank four b - flux converts three of the @xmath36 twisted hypermultiplets into tensor multiples , so the total number of @xmath36 twisted scalars becomes @xmath275 . in this case
we find b_0^(1,9)&=&_i=1 ^ 4 _ i^(4)+ _
j=1 ^ 7_j^(2 ) , b_0^(3,9)&=&_i=1 ^ 4 _ i^(4)- _
j=1 ^ 7_j^(2 ) , b_0^(1,5)&=&-(/2 ) _ 1^(4)- _
1^(2 ) , b_0^(3,5)&=&-(/2 ) _ 1^(4)+ _
1^(2 ) , where _ 1^(4 ) & & _ 1^(4)+4(_1,9+_3,9)/-2(_1,5 + _ 3,5)/ , _
i^(4 ) & & _ i^(4)-(_1,9+_3,9)/ , i=2,4 , _
1^(2 ) & & _ 1^(2)+2(_1,9-_3,9)/-2(_1,5- _ 3,5)/ , _
j^(2 ) & & _
j^(2 ) , j=2,7 .
[ finalz4b4 ] in the field theory framework the cancelation procedure of the six dimensional chiral anomalies involves one - loop and tree - level diagrams .
the former generates the anomaly polynomial , while the latter cancels it through the generalized green - schwarz mechanism .
this cancelation has a simple interpretation from the string theory point of view .
consider 1-loop open string diagrams whose low - energy limit reproduces the 1-loop field theory anomalies .
the factorized form of the anomaly polynomial comes from the appropriate non - planar diagrams in string theory .
let @xmath276 denotes the standard 1-loop cylinder modulus , the field theory result comes from the @xmath277 boundary of the moduli space .
since the non - planar diagrams are finite @xcite , they can not generate chiral anomalies : the anomalous contribution of the 1-loop amplitude from the large @xmath276 region of the moduli space is canceled exactly by the contribution from the small @xmath276 region . through the conformal transformation @xmath278 , the @xmath125 region of the 1-loop open string amplitude is interpreted as @xmath279 limit of a tree - level closed string exchange between two boundary states .
this is precisely the string theory picture of the green - schwarz mechanism . in the rest of this section
, we identify the string diagrams whose @xmath280 limit produces purely gauge anomalies of the @xmath185 orientifold models . for illustrative purposes , we consider terms of the anomaly polynomial ( [ z2anomaly ] ) in the @xmath259 models which are sensitive to the @xmath4 multiplicity of states in the 59-open string sector .
the analysis for the other terms in ( [ z2anomaly ] ) as well as that for the @xmath191 orientifold models is similar and so will not be repeated here . from the anomaly polynomial ( [ z2anomaly ] ) ,
the ratio of the coefficients of @xmath281 and @xmath282 terms is @xmath283 .
this is the result we want to confirm by a string computation .
we start with the @xmath281 term .
the 1-loop open string diagram that would generate this term is shown on fig .
[ 6dfigure1 ] .
this amplitude is given by a_(_1;_1_2_3)^(99)&= & & & _ _
1,9(x_1 ) _ _ 1,9(y_1)__2,9(y_2 ) _ _ 3,9(y_3)_\{,}=\{,}^_99 , [ ab9ab9 ] where the 4 99-sector abelian gauge bosons are split into groups of three and one between the two cylinder boundaries .
the subscript of the correlator indicates the relevant spin structure ) are chosen to be in the @xmath284-picture .
this can be achieved by inserting an appropriate number of worldsheet supercurrents . ] .
the correlation function in ( [ ab9ab9 ] ) decomposes into the non - compact part which depends on the vertex insertion coordinates @xmath285 but is independent of the twists @xmath79 , and the compact part .
since the vertex insertions involve only the non - compact excitations , this compact part is identical to the compact part of the @xmath176 tadpole with @xmath286 spin structure ( [ z991 ] ) .
altogether , we can rewrite ( [ ab9ab9 ] ) as a_(_1;_1_2_3)^(99 ) & = & 13 c_(_1;_1_2_3)(t ) & & \ { _ k=0 ^ 1 _ i=1 ^ 2 ( -2k v_i ) ( _ k,9_9)(_k,9 ^ -1 _ 9 ^ 3 ) s__99(k ) } & = & 13 c_(_1;_1_2_3)(t ) \ { -4 ( _ 1,9_9)(_1,9 ^ -1 _ 9 ^ 3 ) } , [ ab9ab9f ] where @xmath287 is the generator of the @xmath6 in the 99-sector . a symmetry factor @xmath288 in ( [ ab9ab9f ] ) accounts for the fact that all three gauge bosons on one of the boundaries of the cylinder come from the same @xmath6 .
the 1-loop open string diagram that generates @xmath282 term in the anomaly polynomial ( [ z2anomaly ] ) is shown on fig .
[ 6dfigure2 ] .
this amplitude is given by a_(_1;_1_2_3)^(59)&= & 2^b/2 & & _ _
1,5(x_1 ) _ _ 1,9(y_1)__2,9(y_2 ) _ _ 3,9(y_3)_\{,}=\{,}^_59 , [ ab5ab9 ] where @xmath4 accounts for the multiplicity of 59-cylinders . as before
, the correlation function in ( [ ab5ab9 ] ) decomposes into the non - compact part which depends on the vertex insertion coordinates @xmath285 but is independent of the twists @xmath79 , and the compact part .
the non - compact part of this correlation function is identical to the non - compact part of the correlation function in ( [ ab9ab9 ] ) since 99- and 59-cylinders have the same boundary conditions for the non - compact excitations .
the compact part coincides with the compact part of the @xmath179 tadpole with @xmath286 spin structure ( [ z591 ] ) .
altogether , the can rewrite ( [ ab5ab9 ] ) as a_(_1;_1_2_3)^(59 ) & = & 2^b/23 c_(_1;_1_2_3)(t ) & & \ { _ k=0 ^ 1 _ i=1 ^ 2 ( _ k,5_5)(_k,9 ^ -1 _ 9 ^ 3 ) } & = & 2^b/23 c_(_1;_1_2_3)(t ) \ { ( _ 1,5_5)(_1,9 ^ -1 _ 9 ^ 3 ) } , [ ab5ab9f ] where @xmath289 is the generator of the @xmath6 in the 55-sector . from ( [ ab9ab9f ] ) and ( [ ab5ab9f ] ) = -42^b/2 , [ 99ratio59 ] which is the desired result .
in this paper we have considered type iib orientifolds on @xmath0 ( @xmath184 ) with discrete b - flux , previously constructed in @xcite .
we have analyzed the gravitational , gauge and mixed anomalies of the resulting six dimensional vacua and showed that they all cancel .
the cancelation required the @xmath4 multiplicity of states in the 59-open string sector . in a field theory framework , we have identified the twisted sector r - r scalars and tensor multiplets involved in the green - schwarz mechanism responsible for this cancelation .
we presented details of the construction of these models and argued that consistency with @xmath4 multiplicity of 59-sector states requires a modification of the relation between the open string 1-loop channel modulus and the closed string tree channel modulus for the 59- ( 95- ) cylinder amplitudes .
the latter should not be surprising , since only for cylinders with the same boundary conditions ( 99- or 55-cylinders ) can one relate their loop moduli to the loop moduli of the klein bottle and the mbius strip by adding / removing crosscaps .
in fact , we argued that it is precisely the anomaly cancelation condition that should be used to determine the relation between the 59-cylinder 1-loop and tree channel modulus .
the reason is that chiral anomalies are sensitive to the r - r spin structure of 1-loop amplitudes , however this spin structure does not contribute to the 1-loop tadpoles . on the other hand , the r - r spin structure will contribute to 2-loop tadpoles , and hence the cancelation of anomalies probes the 2-loop tadpole consistency of the models .
we showed that the anomaly polynomial computed in field theory can be extracted from an appropriate limit of non - planar cylinder amplitudes .
we identified certain 1-loop non - planar diagrams which are sensitive to the @xmath4 multiplicity of 59-sector states and showed that they correctly reproduce the field theory result , when the multiplicity of 59-cylinders is taken to be @xmath4 as required from the cancelation of the tadpoles .
our results here can be extended to four dimensional orientifolds @xcite . in @xcite a @xmath34 orientifold with @xmath19 b - flux was proposed as an explicit string realization of the `` brane world '' scenario .
the model constructed there contains three chiral families and a pati - salam @xmath290 gauge group .
the fact that there are three chiral families depends crucially on the @xmath4 multiplicity of states in the 59-open string sector . in four dimensions ,
anomaly constraints are much weaker and so are not stringent enough to determine the doubling of the 59-sector states @xcite . however , by compactifying the @xmath19 @xmath185 orientifold considered in this paper on an extra @xmath291 and further orbifolding it by @xmath35 , we obtain the @xmath34 orientifold . before the @xmath35 orbifold projection ,
there is doubling of states in the 59 sector of the orientifold since it is connected to the six - dimensional @xmath36 orientifold discussed in this paper by decompactifying the @xmath291 . , it is possible that some momentum / winding states become massless . however , these states do not come from the @xmath14 sector . ]
this doublet of states can only transform as a singlet representation under @xmath35 .
thus either all the 59-states of @xmath34 orientifolds are projected out , or all of them are kept . since tadpole cancelation requires the presence of 59 open string sector states @xcite , this implies that the 59 sector states come with a multiplicity .
in addition to the two anomalous @xmath6 gauge fields which contribute to the mixed @xmath6-gravitational anomalies @xcite , two more anomalous @xmath6 s arise when other mixed @xmath6 gauge anomalies are taken into account , giving rise to a total of four anomalous @xmath6 gauge fields@xcite .
the presence of the ns - ns sector b - field background is rather generic .
for example , when one considers orientifolding an asymmetric orbifold @xcite , the background b - field is generically non - zero .
the fact that that @xmath115 defined in eq.([momenta ] ) is invariant under @xmath292 and that it takes only quantized values of @xmath293 or @xmath119 suggests that it should be possible to associate with it a @xmath36 action .
indeed , one also finds a @xmath4 multiplicity of states in the 59-sector in this picture @xcite .
we thank philip argyres , mirjam cvetic , luis ibanez , clifford johnson , zurab kakushadze , robert shrock , angel uranga and piljin yi for discussions .
the research of g.s .
is partially supported by the nsf grant phy-97 - 22101 .
the research of a.b . and
s .- h.h.t .
is partially supported by the national science foundation .
s .- h.h.t would also like to thank the aspen center for physics for hospitality while part of this work was completed .
[ appendix6d ]
in this section we review the computation of 1-loop vacuum amplitudes of the @xmath0 orientifolds @xcite .
the 1-loop vacuum amplitudes come from the klein bottle @xmath294 , mbius strip @xmath295 , and cylinder @xmath296 .
the klein bottle amplitude is given by = _
n , k=0^n-1 _ 0^ t ( 4 ^ 2 ^t)^-3 _ ( ^n,^k ) , [ kamp ] where _ ( ^n,^k ) = \{(1+(-1)^f ) ^k e^-2t[l_0(^n ) + _ 0(^n ) ] } .
[ zkdef ] the contribution of the uncompactified momenta is already extracted in ( [ kamp ] ) . here
, @xmath297 denotes the regularized space - time volume .
since @xmath40 exchanges @xmath298 with @xmath299 , only @xmath300 and @xmath301 terms survive the trace .
the trace in @xmath302 can be evaluated in a standard way using @xmath303 functions to write the contributions of complex bosons and fermions . also , the gso projection is implemented by summing over spin structures .
then , taking into account the insertion of @xmath40 we find _ ( 1,^k ) = _ , = 0 , _ , _
i=1 ^ 2 ( -22k v_i ) s_(k ) , [ k1 ] where @xmath304 .
the @xmath303 and @xmath305 functions are defined in appendix [ thetaproperties ] .
the tilde indicates that the argument is @xmath306 .
note that for @xmath307 , ( [ k1 ] ) has a well - defined limit .
the factors @xmath308 account for the sums over quantized momenta ( @xmath309integer ) or winding ( @xmath309half - integer ) in the @xmath310 direction : s_(k)&=&1 , k\{0,n/2 } , s_(k)&=&_k,0_p^2_4q^ p^2/2 + _ k , n/2_^2_4q^^2/(2 ) , k=\{0,n/2}. [ skb ] in ( [ skb ] ) , @xmath311 and @xmath312 denote the winding and momentum lattice of @xmath2 correspondingly .
the @xmath125 divergence of ( [ k1 ] ) represents the exchange of the massless ns - ns ( terms with @xmath313 and @xmath314 in ( [ k1 ] ) ) and r - r forms in the closed string tree channel .
the @xmath315 twisted sector yields _ ( ^n/2,^k ) = ( ^n/2 , ^k ) _ , = 0 , _ , _
[ kr ] in eq .
( [ kr ] ) , @xmath316 is a factor that takes into account the fixed point degeneracy @xcite , _
i.e. _ , it counts the number of fixed points of @xmath317 on @xmath2 invariant under @xmath79 : ( ^n/2 , ^k)&=&16,k=\{0,n/2 } , ( ^2 , ^k)&=&4,k=\{1,3 } , and n=4 .
[ z2fixed ] as ( [ k1 ] ) , ( [ kr ] ) vanishes by virtue of supersymmetry .
the ns - ns exchange in the closed string tree channel is represented by @xmath313 and @xmath314 terms in ( [ kr ] ) .
the cylinder amplitudes are given by _
pq = _ k=0^n-1 _ 0^ t ( 8 ^ 2 ^t)^-3 _
pq(^k ) , [ camp ] where _
pq(^k ) = _ pq \{(1+(-1)^f ) ^k e^-2 t l_0 } .
[ zcdef ] the trace is over open string states with boundary conditions according to the d@xmath8 and d@xmath73-branes at the endpoints . in @xmath318 ,
boundary conditions are nn in all directions .
99(^k ) = _ , = 0 , _ , _
i=1 ^ 2 ( -2k v_i ) ( _ k,9)^2 s__99(k ) .
[ z991 ] the factors @xmath319 account for the sums over quantized momenta in @xmath310 when @xmath309integer : s__99(k)&=&1 , k0 , s__99(0)&=&_p^2_4q^ p^2 .
[ scyl ] eq .
( [ z991 ] ) vanishes by supersymmetry .
the terms @xmath313 and @xmath314 represent ns - ns exchange in the closed string tree channel . in the 55-sector
there are dd boundary conditions in directions @xmath320 transverse to the 5-branes .
the oscillator expansions with dd boundary conditions have integer modes but include windings instead of momenta .
then , @xmath321 has a form similar to ( [ z991 ] ) _
55(^k ) = _ , = 0 , _
i=1 ^ 2 ( -2k v_i ) _ i=1^n(n , k ) ( _ k,5,i)^2 s__55(k ) , [ z55 ] where @xmath228 refers to the fixed points of @xmath79 .
thus , & & n(2,1)=n(4,2)=16 , & & n(4,1)=n(4,3)=4 .
[ 55fixed ] @xmath322 account for the sums over winding in @xmath310 when @xmath309integer : s__55(k)&=&1 , k0 , s__55(0)&=&_^2_4q^^2/ .
[ scyl55 ] in ( [ z55 ] ) terms @xmath313 and @xmath314 represent ns - ns exchange in the closed string tree channel .
in the 59-sector there are dn boundary conditions in coordinates @xmath320 .
hence , their oscillator expansions include half - integer modes . for fermions
, world - sheet supersymmetry requires that in neveu - schwarz ( ramond ) , the moddings are opposite ( same ) to that of the corresponding bosons .
59(^k ) = _ , = 0 , _ , _
i=1 ^ 2 _
i^n(n , k ) _ k,5,i .
[ z591 ] the @xmath313 and @xmath323 contributions of ( [ z591 ] ) represent the ns - ns exchange in the closed string tree channel .
the mbius strip amplitudes are given by _
p = _ k=0^n-1 _ 0^ t ( 8 ^ 2 ^t)^-3 _
p(^k ) , [ mamp ] where _
p(^k ) = _ p \{(1+(-1)^f ) ^k e^-2 t l_0 } .
[ zmdef ] here the trace is over open string states with boundary conditions according to the d@xmath8-branes at both endpoints .
the main difference between @xmath324 and @xmath325 is the insertion of @xmath68 that acts on the various bosonic and fermionic oscillators thereby introducing extra phases in the expansions in @xmath73 .
more precisely , @xmath68 acts on oscillators as @xcite _ r ^ir ; _ r ^ir .
[ apsi ] the upper ( lower ) sign is for nn ( dd ) boundary conditions . furthermore ,
@xmath68 acts as @xmath326 on the ns vacuum .
this ensures that @xmath327 as needed for the orientifold projection on gauge vectors . to derive the mbius trace
we can use ( [ apsi ] ) and the results for 99-cylinders . after using @xmath303 identities
we obtain _
9(^k ) & = & -(1 - 1 ) _ i=1 ^ 2 ( _ _ k,9 ^ -1 _ _
k,9^t ) & & s__9(k ) . [ zm9 ] in ( [ zm9 ] ) we separate the ns - ns and r - r exchange in the closed string tree channel .
@xmath328 account for a sum over quantized momentum in @xmath310 for @xmath309integer : s__9(k)&=&1 , k0 , s__9(0)&=&_p^2_4q^ p^2 .
[ sm9 ] for @xmath70-branes , we find _
5(^k ) & = & ( 1 - 1 ) _
i=1 ^ 2 _ i ( _ _ k,5,i^-1 _ _ k,5,i^t ) & & s__5(k ) .
[ zm5 ] in ( [ zm5 ] ) we separate the ns - ns and r - r exchange in the closed string tree channel . @xmath329 account for a sum over winding in @xmath310 for @xmath309half - integer : s__5(k)&=&1 , kn/2 , s__5(n/2)&=&_^2_4q^^2/ .
the @xmath303 function of rational characteristics @xmath330 and @xmath331 is given by ( t ) = _ n q^(n+)^2 e^2i(n+ ) .
[ vts ] here the variable @xmath73 is @xmath332 .
the @xmath303 function also has the product form = e^2i q^^2 - 124 _
n=1^(1 + q^n+- e^2i ) ( 1 + q^n-- e^-2i ) , [ vtp ] where the dedekind @xmath305 function is = q^124 _ n=1^(1 - q^n ) .
[ deta ] notice that _ 0 = 1 ^ 3 . [ limu ] the @xmath303 and @xmath305 functions have the modular transformation properties ( t ) & = & e^2i t^- ( 1/t ) , + ( t ) & = & t^- ( 1/t ) .
[ modt ] the @xmath303 s satisfy several riemann identities @xcite .
in particular , & & _ , _ , _
i=1 ^ 3 = 0 , + & & _ , _ , _
i=1 ^ 2 = 0 , [ abtru ] provided that @xmath333 .
in this section we collect some facts about anomalies in six dimensional @xmath42 supergravity theories .
details can be found in ref .
@xcite .
@xmath42 supersymmetry in six dimensions involves four types of massless multiplets . in terms of representations of the little group @xmath334
( we label the @xmath335 representations by their multiplicities @xmath336 ) , the massless multiplets are \(i ) gravity multiplet @xmath337 : @xmath338 , \(ii ) tensor multiplet @xmath88 : @xmath339 , \(iii ) vector multiplet @xmath340 : @xmath341 , \(iv ) hypermultiplet @xmath138 : @xmath342 . in six dimensions
chiral anomalies are characterized by an 8-form polynomial constructed from the gauge and gravitational field strengths .
the basic diagram to be examined is the box diagram with an even number of external gravitons and gauge fields .
the resulting anomalous diagrams can be classified as purely gravitational , purely gauge or mixed gauge and gravitational .
all four @xmath42 supergravity multiplets contribute to purely gravitational anomaly .
their separate contributions to the anomaly polynomial are given by i(2)^3 i_g^grav.&=&15760 ( 273 r^4 - 54 51 ( r^2)^2 ) , i(2)^3 i_t^grav.&=&15760 ( -29 r^4 + 54 7 ( r^2)^2 ) , i(2)^3 i_v^grav.&=&15760 ( r^4 + 54 ( r^2)^2 ) , i(2)^3 i_h^grav.&=&15760 ( -r^4 - 54 ( r^2)^2 ) .
[ grav ] where the trace over curvature matrices in @xmath343 is in the vector representation of @xmath344 .
the cancelation of the leading term @xmath28 in the pure gravitational anomalies implies @xmath345 mixed gauge and gravitational anomalies get contributions from gauginos and matter fermions : i(2)^3 i^mixed=-196 r^2 ( & & _ af^2_a-_i , as_a^i(_r^if^2_a ) & & -_i , as_a^i q_i , a^2 ( f^2_a ) -2_ij , a <
bs_ab^ij q_ij , a q_ij , b ( f_a ) ( f_b ) ) , [ mixed ] where @xmath346 represents the nonabelian field strength of a gauge group factor @xmath347 and @xmath348 represents the abelian field strength . @xmath349 denotes the number of hypermultiplets transforming in representation @xmath350 of the nonabelian factor @xmath347 , while @xmath351 counts the hypermultiplets with @xmath352 charge under @xmath220 .
@xmath353 is a multiplicity of states with charges @xmath354 under @xmath355 .
@xmath356 refers to the trace in the respective adjoint representation .
finally , purely gauge anomaly is given by i(2)^3 i^gauge&=&124 ( _ af_a^4-_i , as_a^i ( _ r^i f_a^4)-6_ij , a < b s_ab^ij ( _ r^if_a^2 ) ( _ r^jf_b^2 ) & & -4_ij , aa s_aa^ij q_ij , a ( _ r^if_a^3 ) ( f_a ) & & -6_ij , aa s_aa^ij q_ij , a^2 ( _ r^if_a^2 ) ( f_a)^2 & & -12_iij , a , a < b s_aab^iij q_iij , a q_iij , b ( _ r^if_a^2 ) ( f_a ) ( f_b ) & & -_i , as_a^i q_i , a^4 ( f_a)^4 - 4_ij , abs^ij_ab q_ij , a^3 q_ij , b ( f_a)^3 ( f_b ) & & -6_ij , a < b s^ij_ab q_ij , a^2 q_ij , b^2 ( f_a)^2 ( f_b)^2 ) , [ gauge ] where @xmath357 is the multiplicity of states in the @xmath358 representation of @xmath359 , @xmath360 is the multiplicity of states in the representation / charge @xmath361 of @xmath362 , and @xmath363 is the multiplicity of states in the representation / charges @xmath364 of @xmath365 . in ( [ gauge ] )
we included only hypermultiplet representations which are relevant to the orientifold models of interest ( see table [ 6dtable1 ] for the hypermultiplet spectrum ) .
it is convenient to express all traces over nonabelian field strengths in the fundamental representation of respective gauge groups .
here we list out the relations between traces in different representations for @xmath366 groups @xcite ( since they are the nonabelian gauge groups arise in the type iib orientifolds of interest in this paper ) . for @xmath367 , f^4&=&2n f^4 + 6 ( f^2)^2 , _ a^ij f^4&=&(n-8 ) f^4 + 3 ( f^2)^2 , _ a^ij f^3&=&(n-4 ) f^3 , f^2&=&2n f^2 , _
a^ij f^2&=&(n-2 ) f^2 , [ tracesrelations ] where @xmath368 denotes a second rank antisymmetric tensor representation .
@xmath369 refers to the trace in the fundamental representation .
furthermore , for @xmath335 : f^4&= & ( f^2)^2 . [ su2traces ]
in sec . [ opensector ] we discuss how the `` shift '' vector formalism of @xcite can be used to represent gauge twists @xmath370 and determine massless states in the open string sector of the type iib orientifold models . here
, we collect the formulas relevant to the computation of traces for the anomaly cancelation . in sec .
[ opensector ] we rewrite the chan - paton matrices @xmath371 and @xmath372 in a cartan - weyl basis .
the subscript @xmath373 refers to the nonabelian gauge group factor @xmath347 in the @xmath216 open string sector , while @xmath64 labels the generator of @xmath220 in open string sector @xmath8 .
the cartan generators are represented by tensor product of @xmath374 @xmath375 submatricies .
we chose the normalization of the @xmath133 generators @xmath101 in such a way that @xmath376 .
in all the models we discuss , there is a @xmath6 factor for each of the @xmath366 s in the model . the generator @xmath372 of a given @xmath220 is given by a linear combination _
a^p = q_a^ph , [ u1gen ] where @xmath138 is a vector of cartan generators and @xmath377 is a @xmath148-dimensional real vector of the form q_a^p=1(0,0 , , _
n_a times , 0,0,,0 ) , [ chargevector ] where the non - zero entries sit at the positions where the corresponding @xmath378 lives . with these convention and normalization & & ( _ k , p _
a^p)=(e^-2ik v_(pp)h q_a^ph ) = ( -i ) 2k v_(pp)^a , & & ( _ k , p _ a^p)=0 , & & ( _ k , p ( _ a^p)^2)=2k v_(pp)^a , & & ( _ k , p ( _ a^p)^2)=2k v_(pp)^a , & & ( _ k , p ( _ a^p)^3)=(-i)1 2k v_(pp)^a , & & ( _ k , p ( _ a^p)^3)=(-i)1 2k v_(pp)^a , & & ( _ k , p _ a^p ( _ a^p)^2)=(-i)1 2k v_(pp)^a , v_(pp)^a = v_(pp)^a .
[ tracescollection ] where @xmath379 is the rank of the @xmath380 group containing @xmath220 and @xmath381 is a component of the @xmath382 shift vector along any of the overlapping entries with that @xmath380 , @xmath383 is a component of @xmath382 along any entry overlapping with the group @xmath347 in the @xmath8-open string sector .
.the massless spectrum of the six dimensional type iib orientifolds on @xmath0 for @xmath184 , and various values of @xmath1 ( the rank of @xmath104 ) .
the semi - colon in the column `` charged hypermultiplets '' separates 99 and 55 representations . [ cols="^,^,^,<,^,^ " , ] for a partial list , see , _ e.g. _ , + m. berkooz and r.g .
leigh , nucl .
* b483 * ( 1997 ) 187 ; + c. angelantonj , m. bianchi , g. pradisi , a. sagnotti and ya.s .
stanev , phys .
* b385 * ( 1996 ) 96 ; + z. kakushadze , nucl . phys .
* b512 * ( 1998 ) 221 ; + z. kakushadze and g. shiu , phys .
* d56 * ( 1997 ) 3686 ; nucl . phys .
* b520 * ( 1998 ) 75 ; + g. zwart , nucl . phys .
* b526 * ( 1998 ) 378 ; + z. kakushadze , g. shiu and s .- h.h .
tye , nucl .
* b533 * ( 1998 ) 25 ; + r. blumenhagen and a. wisskirchen , phys . lett .
* b438 * ( 1998 ) 52 ; + z. kakushadze , phys . lett . *
b434 * ( 1998 ) 269 ; nucl . phys . *
b535 * ( 1998 ) 311 ; phys . rev . * d58 * ( 1998 ) 101901 ; + z. kakushadze and s .- h.h .
tye , phys . rev .
* d58 * ( 1998 ) 126001 .
l. e. ibanez , r. rabadan , a. m. uranga , nucl . phys .
* b542 * ( 1999 ) 112 ; hep - th/9905098 + g. aldazabal , d. badagnani , l.e . ibanez and a.m.
uranga , jhep * 9906 * ( 1999 ) 031 ; + e. poppitz , nucl .
b542 * ( 1999 ) 31 ; + m. cvetic , l. everett , p. langacker and j. wang , hep - th/9903051 ; + z. lalak , s. lavignac and h.p .
nilles , hep - th/9903160 . | we consider anomaly cancelations in type iib orientifolds on @xmath0 with quantized ns - ns sector background b - flux . for a rank @xmath1 b - flux on @xmath2
( @xmath1 is always even ) and when @xmath3 is even , the cancelation requires a @xmath4 multiplicity of states in the 59-open string sector .
we identify the twisted sector r - r scalars and tensor multiplets which are involved in the green - schwarz mechanism .
we give more details of the construction of these models and argue that consistency with the @xmath4 multiplicity of 59-sector states requires a modification of the relation between the open string 1-loop channel modulus and the closed string tree channel modulus in the 59-cylinder amplitudes .
= 10000 epsf 2 |
the dominant , non - instrumental background source for infrared observations in space is the zodiacal light ( zl ) , which comes from both scattered and thermal interplanetary dust ( ipd ) emission .
this dust originates mainly from comets and asteroids , as well as a minimal amount from the interstellar medium @xcite .
it has many components , including a smoothly distributed dust cloud along with various clumps and gaps generated by interactions and resonances with the large bodies of the solar system .
studies of the zl are applicable to at least three different , unrelated fields of astronomical research .
understanding the zl allows more accurate studies of the cosmic infrared background ( cib ) because it is the dominant error source in those measurements , which has implications for pop iii stars .
in addition , any absolute measure of extended surface brightness , for example the intracluster light in galaxy clusters , or the outer expanses of nearby galaxies , needs to take into account the contribution from zodiacal light .
lastly , zodiacal models are important in comparison to extrasolar ipd models , especially in using structure in the ipd of stellar systems to find planets . measurements of the zl and models for its properties are based on iras ( 12 - 100@xmath1 ) and , primarily , cobe dirbe ( 1 - 240@xmath1)@xcite . the model most commonly used to date , based on dirbe data , is @xcite .
the dirbe data have been subsequently analyzed in conjunction with other data sets to calibrate inconsistencies in various zl models .
motivated by a desire to measure the cib , @xcite and @xcite re - derived a zl model , based on dirbe data only , making changes to the scattering function and forcing the darkest regions to zero zodiacal emission .
using infrared telescope in space ( irts ) data , @xcite note the need for a correction of order a few percent at near - infrared wavelengths to the @xcite model which they attribute to calibration differences between cobe and irts , and quoted uncertainties in the model parameters .
used a hybrid approach to fix most zl model parameters to be consistent with the @xcite model while still allowing their akari infrared camera data to constrain certain parameters .
they find an underestimate by @xcite of the earth trailing cloud component and a possible warping of the ipd cloud .
we present here an independent evaluation of the zl model of @xcite using @xmath2 irac data @xcite .
@xmath3 measurements of the zl are unique because 1 ) the spacecraft is not in the same place as the earth , which gives us new positional information on the ipd cloud and 2 ) we have a multi - year baseline with approximately weekly cadence .
@xmath2 is in an earth - trailing orbit , slowly drifting behind the earth at a rate of @xmath40.1 au per year . as of january 2012
, @xmath2 is approximately 1 au from the earth .
we know its position to 500 km , considerably under mission specifications .
the weekly cadence allows us to measure the annual variation in the irac background levels due to the zl signal , which in turn will allow for generation of a more accurate model than what could be generated from the ground or a stationary satellite with short duration .
the measured surface brightness of the zl depends on time of year , direction of observation , and location within the dust cloud .
the irac wavelength range ( 3.6 - 8@xmath1 ) probes regimes of both scattered light and thermal emission ; both contribute equally at 3.6 , but the longer wavelengths are dominated by thermal emission @xcite .
the dirbe data show that the zodiacal light typically accounts for just over 50% of the non - instrumental measured sky levels at @xmath5 , which is about five times the predicted cib levels .
that number jumps to the zodiacal light being 70% of the non - instrumental sky levels at @xmath6 . at longer wavelengths the sky levels are a few hundred times brighter than the cib signal .
only redward of @xmath7 do other foregrounds become significant @xcite .
as part of the calibration program of irac , a field @xmath8 from the nep , with relatively low zl and no bright stars or extended galaxies is observed with a regular cadence for the purpose of having a shutterless measurement of the bias and dark current in the arrays @xcite . when looking towards the nep , @xmath9 observes a near vertical line of sight through the ipd cloud , approximately perpendicular to the dust plane , at a distance of about 1au from the sun . during the cryogenic mission from dec 2003 through may 2009 ,
dark field data were observed twice per campaign , whenever the irac instrument was on , which was roughly every two to three weeks .
@xmath2 cryogen was depleted in may 2009 , leaving only the 3.6 and 4.5@xmath1 channels functional . from may 2009 through jan 2010 , as the instrument was changing temperature and bias levels , the dark fields flux levels were not comparable to those at a steady temperature , and therefore are not used in this analysis . during the warm mission , from january 2010 to the present , dark fields are observed once per week , with relatively few exceptions .
the data used in this analysis are observed as a set of 18 , dithered , 100s fowler - sampled exposures processed with the calibration pipeline version s18.8(cryo ) and s19.0(warm ) .
the longest possible irac exposure times were chosen as the best measure of the zl variations .
for the 8@xmath1 data , 50s observations are the longest possible exposure time .
similar to the basic calibrated data pipeline detailed in the irac handbook , each of the raw frames was bias subtracted , linearized , flat - fielded , and corrected for various instrumental effects .
the bias correction removes the first frame effect which is a dependency of the bias pattern on the time since last readout , or delay
time. for the cryogenic mission , a pre - launch library of frames taken at different exposure delay times was interpolated and used as a first frame correction .
the warm mission has no such ground data , and therefore no first frame correction .
however , we do not expect the lack of a warm first frame correction to effect the zl measure because it should be a constant offset in the level of the frames since all data was taken with the same delay times in both cryogenic and warm data .
after these corrections , a separate program creates a median image of all the frames ( 18 ) within each set of observations using outlier rejection , including a sophisticated spatial filtering stage to reject all stars and galaxies generating a final `` skydark '' image .
the calibration pipeline was designed as the best measure of the dark and bias level in the frames without signal from resolved sources .
diffuse stray light is a contaminant at the 1% level in both the raw and flat - field images used to correct the pixel to pixel gain effects .
we expect this contaminant to be mitigated to less than 0.25% by using 18 dithered frames in the median combine .
all data has been converted into physical units of mjy / sr by applying a calibration based on point sources . because the 5.8 and 8.0@xmath1 detectors suffer from internal scattering caused by photons diffusing around on the chip , the point source calibration is not appropriate for extended sources in these bands .
we therefore apply an extended source aperture correction which is a @xmath10 increase .
this correction , which has associated errors , is not critical to the below analysis because it only effects the overall level of the background , and not the annual variation . to measure the background level in the data , which is composed of zl , cib , and any residual instrumental dark current
, a gaussian distribution is fitted to the histogram of all pixel values .
the mean of that gaussian is considered the true background level for that observation .
the mode , as calculated with the technique described in @xcite , is different from the mean by only 0.35% , which is insignificant to our conclusions .
therefore , we use the mean as an adequate representation of the background . daily observations of calibration stars throughout the mission show that irac photometry in all four channels is stable to 1% , or better , as a function of time .
therefore , it is reasonable to compare photometry over the entire mission without large time - scale systematic drifts .
a completely different calibration product exists to measure the flat fields for irac , where observations are taken of high background ecliptic fields .
those observations are not a clean zl measurement in the same sense as the dark field data because they are taken of a different field every month , and so they do not have the same sampling as the dark field and therefore wo nt have good enough statistics to compare the sinusoidal variation to the dirbe data / models .
we make a direct comparison between irac bands 1 , 2 , 3 , & 4 ( 3.6 , 4.5 , 5.8 , 8.0@xmath1 ) and cobe dirbe bands 3 , 4 , and 5 ( 3.5 , 4.9 , and 12@xmath1 ) .
dirbe operated in cryogenic mode from november 1989 to september 1990 , scanning half the sky every day , building an all - sky map with high coverage and measuring absolute flux using a zero - flux internal calibrator .
the dirbe background has contributions from zl , astrophysical sources , and cib . for this analysis we used the calibrated individual observations ( cio ) , which are the unbinned data for the full cryogenic mission ( 285 days ) , with data taken every @xmath11 second .
dirbe pixels are roughly @xmath12 on a side .
we measured surface brightness in the single pixel that includes the center of the irac dark field . at the location of that pixel ,
over the course of the cryogenic cobe mission , there are roughly 495 individual observations per channel after outlier rejection , where @xmath13 of observations were rejected because they were greater than five sigma from the mean of the distribution .
we binned the data to give roughly 34 observations .
this binning level was chosen to reduce noise while still retaining the annual zl signature .
errors are fixed per channel at 3.3% , 2.7% and 5.1% for the three channels respectively @xcite .
figures 1 - 4 display the irac background levels over the entire 8.5 year lifetime , to date , of @xmath2 .
cryogenic mission data are shown in green , and warm mission data are shown in blue .
the 5.8 and 8@xmath1 channels , shown in figures 3 and 4 , were no longer viable after cryogenic operations ended @xmath45.5 years after launch . between the cryogenic and warm missions ,
the detector temperature increased from 15 to 28.7k , consequently increasing the dark current in the images by about 0.6mjy / sr .
therefore , the 3.6@xmath1 surface brightnesses in the cryo and warm data are shown on different scales .
the 3.6@xmath1 cryogenic data are much noisier than the @xmath14 warm data , and all of the 4.5@xmath1 data .
the cause of this is uncertain , but could be that the instrument was less stable due to constant power cycling and annealing that only occurred during the cryogenic mission .
the dirbe data are shown in figures 1,2 , & 4 as solid black points .
irac 8@xmath1 data are shown with the dirbe 12@xmath1 channel for comparison .
the irac 5.8 micron data do not have a good counterpart in the dirbe filter set . in all cases
the dirbe data is phased in time to fit on top of the irac observations even though they were not taken simultaneously .
the dirbe data are also shifted in surface brightness to match the irac data .
the two data sets do not have the same absolute level because 1 ) filter responses are not the same , 2 ) irac data includes an instrumental dark current because irac does not have an absolute calibration ( see [ irac_observations ] ) , and 3 )
while dirbe data do have absolute calibration , they include a signal from stars and galaxies that are unresolved in its relatively large beam .
@xcite measure the contribution of stars to be roughly 10% of the total flux at @xmath5 .
the @xcite model is based on dirbe spectral , temporal and angular information .
briefly , this is a complex , 3d , physical model with over 90 free parameters .
it includes contributions from a smooth cloud , three asteroidal debris dust bands , and a circumsolar dust ring near 1au .
documented sources of uncertainty include non - uniqueness of the model , use of circular , flat orbits when ellipticity and warping are known to exist , and simplistic assumptions about the dust distributions ; among others .
we emphasize that this is an extremely difficult problem to solve with many components and limited data .
the solid black line in figures 1 - 4 shows the predicted zl level based on @xcite .
the model values have been shifted in surface brightness by @xmath15 , @xmath16 , @xmath17 , and @xmath18 mjy / sr in the four channels respectively , to match the irac mean cryogenic levels .
the bottom plot of each figure shows the residuals after subtracting the zodiacal light model from the data .
the y - axis is shown in units of percent of the surface brightness data .
the sinusoidal variation in our figures is the zl signature .
we see an anual variation in the zl contribution to irac data because the location of the dark field precesses around the real north ecliptic pole over the course of a year , the dust particle orbits near 1au are eccentric , and the zodiacal cloud is tilted relative to the earth s orbital plane @xcite .
spitzer is moving above and below the ecliptic dust plane , so that when the telescope is below the plane , it views a larger column density of material towards the nep , and we see a maximum .
six months later , when the telescope is above the plane , the column towards the nep is much smaller and we see a minimum zl signature .
because these data were taken at the nep , we expect much less zl than we would observe edgewise through the ecliptic plane .
the absolute surface brightness measured in irac data is affected by both astronomical sources ( zodiacal , interstellar medium , and extragalactic ) , and instrumental effects .
while dirbe made an absolute measurement of the instrumental background level , irac can not . without use of a shutter
, irac has no direct way of disentangling the dark current and bias level from astronomical sources .
we therefore ignore the absolute level in the plots and focus only on the sinusoidal shape as the measurement of zl variation as a function of time . the irac data reduction process ( see [ irac_observations ] ) removes all stars and galaxies from the measurement , so the seasonal variation seen is not the result of the resolved star and galaxy content changing .
we assume that the zl is constant over the 25 square arcminute irac field of view , and the larger , 1800 square arcminute , dirbe beam .
figures 1 - 4 show significant residuals between the irac data and the model , implying inaccuracies in the model .
the model curve should fit the irac data to higher precision because it has been tailored to the irac bands , @xmath2 orbital position around the sun , and the pointing and time that the data were taken . at 3.6@xmath1 ,
the cryogenic and dirbe data are too noisy to glean much information .
however the warm mission data show that the model under - predicts the amplitude of the variation by @xmath19 percent .
an underestimate of the amplitude could imply either an underestimate of the amount of dust at 1au , or that the scale height of the dust disk is flatter than modelled leading to spitzer traversing further above and below the concentrated area than predicted . at 4.5@xmath1 ,
all irac data show a deviation from a sinusoid shape in addition to an underestimate of the amplitude of that sinusoid .
the shape change is most clearly evident in the warm mission residuals , which are not sinusoidal in shape .
this is the only channel where this shape deviation is seen .
a possible explanation for the shape change is that the dust disk is warped .
the 4.5@xmath1 residuals are larger than the 3.6@xmath1 residuals , which could be giving us color information about the model implying a better knowledge of the dust grain properties .
however , it could also be connected to the difference in the scattered and thermal components of the zl to the two different wavelength bands , and we have no way of separating those effects .
the 4.5@xmath1 residuals also show the overdensity behind the earth as seen previously at 8@xmath1(see below ) . at 5.8@xmath1 ,
the data are too noisy to be able to glean information from the residuals .
surface brightness values can be negative because a ground - based estimate of the dark current is removed from this data , which is known to differ from the true dark current , of which we have no good measure without use of a shutter
. we can rule out large scale changes from the predicted surface brightnesses . at 8.0@xmath1 ,
deviations from a simple sinusoidal annual variation over the first year and a half of the mission have been associated with the telescope travelling through an overdensity in dust behind the earth @xcite .
this overdensity is seen in figure 4 where the residuals are mainly negative for the first 1.5 years , and then switch to oscillating around zero . beyond that overdensity , the residuals look very smooth and constant , implying no further large over- or under - densities at 1au .
in addition to the overdensity , our data show that the model under - predicts the amplitude of the variation by @xmath20 , similar to that seen at 3.6 and 4.5@xmath1 .
we used irac calibration data taken roughly weekly of the nep to study the zl component at 3.6 , 4.5 , 5.8 , and 8.0@xmath1 over the course of the currently 8.5 year mission of the instrument .
we compare the irac data to both cobe dirbe data and the zl model of @xcite based thereon .
cobe dirbe data are taken from 9.4 months of observations of the same region of the sky as the irac data at 3.5 , 4.9 , and 12@xmath1 .
the @xcite model is a 90 parameter fit to the dirbe all sky data at multiple wavelengths from 1 - 240@xmath1 .
all data are shown in figures 1 - 4 .
the sinusoidal variation in the plots is the zl signature .
the spitzer irac data show a deviation from the @xcite model at most at the few percent level .
we see an under - prediction of the amplitude of the yearly variation by the model , the presence of an overdensity behind the earth , and possible evidence for a warping in the ipd cloud .
these data show both that irac can be used for zl studies and that the zl model would benefit from the additional information gathered here
. a better understanding of the zodiacal light will have broad impacts on studies of the cib , low surface brightness observations , and extrasolar planets , among other things . generating a new zl model is beyond the scope of this work .
it is difficult to know the effect of the few percent discrepancies discussed here on work which uses the @xcite model .
because the contribution of the zl to the background of any given image will change as a function of direction , time of year , and wavelength , there is no easy prescription for the application of these residuals to the conclusions of other papers .
it is worth noting that work on the cosmic infrared background is very sensitive to models of the zl as that type of science is often working at only the few percent level for detections of their signal .
we thank the anonymous referee for useful suggestions on the manuscript .
this research has made use of data from the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation .
this work was based on observations obtained with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .
we acknowledge the use of the legacy archive for microwave background data analysis ( lambda ) .
support for lambda is provided by the nasa office of space science . , n. w. , mather , j. c. , weiss , r. , bennett , c. l. , cheng , e. s. , dwek , e. , gulkis , s. , hauser , m. g. , janssen , m. a. , kelsall , t. , meyer , s. s. , moseley , s. h. , murdock , t. l. , shafer , r. a. , silverberg , r. f. , smoot , g. f. , wilkinson , d. t. , & wright , e. l. 1992 , , 397 , 420 , g. g. , hora , j. l. , allen , l. e. , ashby , m. l. n. , barmby , p. , deutsch , l. k. , huang , j .- s . , kleiner , s. , marengo , m. , megeath , s. t. , melnick , g. j. , pahre , m. a. , patten , b. m. , polizotti , j. , smith , h. a. , taylor , r. s. , wang , z. , willner , s. p. , hoffmann , w. f. , pipher , j. l. , forrest , w. j. , mcmurty , c. w. , mccreight , c. r. , mckelvey , m. e. , mcmurray , r. e. , koch , d. g. , moseley , s. h. , arendt , r. g. , mentzell , j. e. , marx , c. t. , losch , p. , mayman , p. , eichhorn , w. , krebs , d. , jhabvala , m. , gezari , d. y. , fixsen , d. j. , flores , j. , shakoorzadeh , k. , jungo , r. , hakun , c. , workman , l. , karpati , g. , kichak , r. , whitley , r. , mann , s. , tollestrup , e. v. , eisenhardt , p. , stern , d. , gorjian , v. , bhattacharya , b. , carey , s. , nelson , b. o. , glaccum , w. j. , lacy , m. , lowrance , p. j. , laine , s. , reach , w. t. , stauffer , j. a. , surace , j. a. , wilson , g. , wright , e. l. , hoffman , a. , domingo , g. , & cohen , m. 2004 , , 154 , 10 , m. g. , arendt , r. g. , kelsall , t. , dwek , e. , odegard , n. , weiland , j. l. , freudenreich , h. t. , reach , w. t. , silverberg , r. f. , moseley , s. h. , pei , y. c. , lubin , p. , mather , j. c. , shafer , r. a. , smoot , g. f. , weiss , r. , wilkinson , d. t. , & wright , e. l. 1998 , , 508 , 25 , r. f. , hauser , m. g. , boggess , n. w. , kelsall , t. j. , moseley , s. h. , & murdock , t. l. 1993 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
2019 , society of photo - optical instrumentation engineers ( spie ) conference series , ed .
m. s. scholl , 180189 , m. w. , roellig , t. l. , low , f. j. , rieke , g. h. , rieke , m. , hoffmann , w. f. , young , e. , houck , j. r. , brandl , b. , fazio , g. g. , hora , j. l. , gehrz , r. d. , helou , g. , soifer , b. t. , stauffer , j. , keene , j. , eisenhardt , p. , gallagher , d. , gautier , t. n. , irace , w. , lawrence , c. r. , simmons , l. , van cleve , j. e. , jura , m. , wright , e. l. , & cruikshank , d. p. 2004
, , 154 , 1 | the dominant non - instrumental background source for space based infrared observatories is the zodiacal light .
we present @xmath0 infrared array camera ( irac ) measurements of the zodiacal light at 3.6 , 4.5 , 5.8 , and 8.0@xmath1 , taken as part of the instrument calibrations .
we measure the changing surface brightness levels in approximately weekly irac observations near the north ecliptic pole ( nep ) over the period of roughly 8.5 years .
this long time baseline is crucial for measuring the annual sinusoidal variation in the signal levels due to the tilt of the dust disk with respect to the ecliptic , which is the true signal of the zodiacal light .
this is compared to both cosmic background explorer diffuse infrared background experiment ( cobe dirbe ) data and a zodiacal light model based thereon .
our data show a few percent discrepancy from the @xcite model including a potential warping of the interplanetary dust disk and a previously detected overdensity in the dust cloud directly behind the earth in its orbit .
accurate knowledge of the zodiacal light is important for both extragalactic and galactic astronomy including measurements of the cosmic infrared background , absolute measures of extended sources , and comparison to extrasolar interplanetary dust models .
irac data can be used to further inform and test future zodiacal light models . |
anderson localization @xcite was discovered 50 years ago in disordered crystals as an accumulation of single particle electronic wavefunctions and can be interpreted as an interference effect between multiple scatterings of the electron by random defects of the potential . as a consequence eigenstates are no longer spatially extended , but exponentially localized .
anderson localization is a universal phenomenon of wave physics , unrestricted to quantum mechanics .
experimental observations were made in noninteracting bose - einstein condensates ( bec ) expanding in random optical potentials @xcite , light propagation in spatially random nonlinear optical media @xcite , and in microwave cavities filled with randomly distributed scatterers @xcite .
anderson localization is a linear wave effect , i.e. it is well - established for wave equations which are linear in the wave amplitude .
however , in many cases one is confronted with a nonlinear response of the wave - carrying medium ; for instance , large light intensities induce a nonlinear response of the optical medium .
electron - electron and electron - phonon interactions also result in substantial deviations from anderson localization in solids . in experiments of bose - einstein condensates
the interatomic interactions are always present , although they can be diminished by either decreasing atomic densities or by exploiting magnetically tunable feshbach resonances
. from a mathematical perspective , a linear wave equation is integrable , with each normal mode evolving independently in time .
a localized wave packet in the presence of anderson localization will therefore stay localized as time evolves .
nonlinearity will usually destroy the integrability of a system and induce mode - mode interactions .
it was observed numerically that wave packets in such nonlinear disordered wave equations delocalize in time without respecting anderson localization limits @xcite .
thus , there are several intriguing questions which have attracted much attention during the last few years : ( i ) will anderson localization be destroyed by arbitrary small strength of nonlinearity or is there a threshold below which the localization is restored ?
( ii ) will wave packet spreading , if observed , last forever or will it stop at certain ( though probably very large ) time ? ( iii ) is the shape of the initial wave packet crucial for the details of spreading
? we will mainly address question ( ii ) here .
johansson et al .
@xcite conjectured that spreading must eventually stop and dynamics will become close to regular , assuming that in these limits the kolmogorov - arnold - moser ( kam ) theorem is applicable , i.e. that for small wave density regular nonergodic phase space structures predominate and the dynamics develops along kam tori .
other attempts consist in a numerical scaling analysis , in order to predict and extend results beyond computational ability @xcite .
analytical studies perform perturbation theory to higher order by treating the strength of nonlinearity as a small parameter @xcite , conflicting with the explosive growth of secular terms in higher orders of perturbation theory .
this theory states that for the disordered discrete nonlinear schrdinger model with nonlinearity strength exceeding a finite threshold , any initial localized wave packet can not fully spread to zero amplitudes at infinite time . in this case , a part of the excitation is selftrapped as a result of nonlinearity induced frequency shifts , which tune a localized excitation out of resonance with its surrounding non - excited linear modes .
however , even in the case of strong nonlinearity , subdiffusion of the non - selftrapped part is observed @xcite .
when strong nonlinearities are avoided numerical studies showed a rather universal asymptotic subdiffusive spreading of initial single site excitations @xcite , which is characterized by a growth of the second moment of the wave packet as @xmath0 , with @xmath1 @xcite .
the selftrapping theorem @xcite holds irrespective of the strength of disorder , therefore it is reflecting the properties of a strongly nonlinear lattice wave equation , rather than peculiarities of waves propagating in disordered media .
also the selftrapping theorem is crucially depending on the presence of at least two integrals of motion , and fails for most nonlinear wave equations with only one integral of motion . in @xcite the observed wave packet spreading
was assumed to be due to an incoherent excitation of the wave packet exterior , induced by the chaotic dynamics of the wave packet interior .
the number of resonant modes in the packet was estimated by considering from quadruplet and triplet mode - mode interactions @xcite . a generalization to higher dimensions @xmath2 and different nonlinearity powers were performed .
this led to a quantitative prediction for the subdiffusive wave packet spreading characteristics @xmath3 @xcite .
its validity was confirmed numerically in @xcite .
recently , it has been predicted theoretically @xcite and verified numerically @xcite that a potentially long - lasting strong chaos regime induces faster ( though still subdiffusive ) spreading , which is followed by the asymptotic and slower weak chaos subdiffusive spreading .
notably , published numerical data did not report on a further slowing down of spreading , when starting from the weak chaos regime . in this paper we present results of extensive numerical studies of
wave packet spreading in various models of disordered nonlinear one - dimensional lattices .
in particular , we consider different initial excitations and scan the parameter space of disorder strength and nonlinearity over a wide region .
the main aim is to test the applicability of previously derived spreading laws , and to search for indications of a continuation of the weak chaos spreading , or for indications of a slowing down , as conjectured by others .
in our study we consider various one - dimensional lattice models .
the first one is the disordered discrete nonlinear schrdinger equation ( dnls ) described by the hamiltonian function @xmath4 in which @xmath5 are complex variables , @xmath6 are the lattice site indices and @xmath7 is the nonlinearity strength .
the random on - site energies @xmath8 are chosen uniformly from the interval @xmath9 $ ] , with @xmath10 denoting the disorder strength .
the equations of motion are generated by @xmath11 : @xmath12 this above set of equations conserves both the energy of eq .
( [ rdnls ] ) , and the norm @xmath13 . the second model we consider is the quartic klein - gordon ( kg ) lattice , given as @xmath14 where @xmath15 and @xmath16 respectively are the generalized coordinates and momenta on site @xmath6 , and @xmath17 are chosen uniformly from the interval @xmath18 $ ] .
the equations of motion are @xmath19 and yield @xmath20 this set of equations only conserves the energy of eq .
( [ rqkg ] ) .
the scalar measure of energy resulting from eq .
( [ rqkg ] ) we shall henceforth label as @xmath21 .
this scalar value @xmath22 serves as a control parameter of nonlinearity , similar to @xmath23 for the dnls case . for @xmath24 and @xmath25 ,
( [ rdnls - eom ] ) reduces to the linear eigenvalue problem @xmath26 the normalized eigenvectors @xmath27 ( @xmath28 are the corresponding normal modes ( nms ) , and the eigenvalues @xmath29 are the frequencies of these nms .
the width of the eigenfrequency spectrum @xmath29 in eq .
( [ evequation ] ) is @xmath30 with @xmath31 $ ] .
the coefficient @xmath32 in eq .
( [ rqkg ] ) was chosen so that the linear parts of the hamiltonians , eqs .
( [ rdnls],[rqkg ] ) would correspond to the same eigenvalue problem . in the limit @xmath33 ( in practice by neglecting the nonlinear term @xmath34 ) the kg model of eq .
( [ rqkg ] ) - with @xmath35 - is reduced to the same linear eigenvalue problem of eq .
( [ evequation ] ) , under the substitutions @xmath36 and @xmath37 .
the width of the squared frequency @xmath38 spectrum is @xmath39 with @xmath40 $ ] .
note that @xmath41 .
as in the case of dnls , @xmath10 determines the disorder strength .
the asymptotic spatial decay of an eigenvector is given by @xmath42 where @xmath43 is the localization length .
in the case of weak disorder , @xmath44 , the localization length is approximated @xcite as @xmath45 . the nm participation number @xmath46 characterizes the spatial extend of the nm .
an average measure of this extent is the localization volume @xmath47 , which is of the order of @xmath48 for weak disorder and unity in the limit of strong disorder , @xmath49 @xcite .
the average spacing of eigenvalues of nms within the range of a localization volume is then @xmath50 , with @xmath51 being the spectrum width .
the two frequency scales @xmath52 determine the packet evolution details in the presence of nonlinearity . in order to write the equations of motion of hamiltonian ( [ rdnls ] ) in the normal mode space of the system we insert @xmath53 in ( [ rdnls - eom ] ) , with @xmath54 denoting the time - dependent amplitude of the @xmath55th nm .
then , using eq .
( [ evequation ] ) and the orthogonality of nms the equations of motion ( [ rdnls - eom ] ) read @xmath56 with the overlap integral @xmath57 the frequency shift of a single site oscillator induced by the nonlinearity is @xmath58 for the dnls model .
the squared frequency shift of a single site oscillator induced by the nonlinearity for the kg system is @xmath59 , with @xmath60 being the energy of the oscillator .
since all nms are exponentially localized in space , each effectively couples to a finite number of neighbor modes .
the nonlinear interactions are thus of finite range ; however , the strength of this coupling is proportional to the norm ( energy ) density for the dnls ( kg ) model .
if the packet spreads far enough , we can generally define two norm ( energy ) densities : one in real space , @xmath61 ( @xmath60 ) and the other in nm space , @xmath62 ( @xmath63 ) .
averaging over realizations , no strong difference is seen between the two , and therefore , we treat them generally as some characteristic norm ( @xmath64 ) or energy ( @xmath65 ) density . the frequency shift due to nonlinearity is then @xmath66 for the dnls model , while the squared frequency shift is @xmath67 for the kg lattice .
the basic characteristics of both models are summarized in table [ tab : char ] .
we order the nms in space by increasing value of the center - of - norm coordinate @xmath68 @xcite . for dnls
we follow normalized norm density distributions @xmath69 with @xmath70 , where @xmath71 is the amplitude of the @xmath55th nm and @xmath72 its squared frequency .
we measure the second moment @xmath73 ( with @xmath74 ) , which quantifies the wave packet s spreading width ; the participation number @xmath75 , i.e. the number of the strongest excited modes in @xmath76 ; and the compactness index @xmath77 , which quantifies the inhomogeneity of a wave packet .
thermalized distributions have @xmath78 , while @xmath79 indicates very inhomogeneous packets , e. g. sparse ( with many holes ) or partially selftrapped ones ( see @xcite for more details ) .
in addition , following anderson s definition of localization @xcite , we measure the fraction @xmath80 ( @xmath81 ) of the wave packet norm ( energy ) in a localization volume @xmath47 around the initially excited state in real space . for a localized state
this fraction asymptotically tends to a constant nonzero value , while it goes to zero in the case of delocalization . in the limit of strong disorder ( @xmath82 )
the dnls and kg models suffer from increasing computational times needed to observe any nontrivial dynamics .
this is because the eigenvectors tend to single site profiles , i.e. the overlap integrals become very small .
frhlich , spencer and wayne ( fsw ) suggested considering a modified hamiltonian , which operates directly in normal mode space for the strong disorder limit , but considers artificial rescaled anharmonic interactions between neighboring nms in order to rescale time @xcite : @xmath83 where the nms are equivalent to the single site oscillators .
the nm eigenvalues @xmath84 are considered to be uncorrelated , also for nearest neighbors .
this is different from the dnls and kg models .
also the fsw chain has only pair interactions between nms ( sites ) .
note also that the nonlinear part of the fsw hamiltonian is invariant under any shift @xmath85 , as opposed to the kg model .
we also consider two variants of dnls and kg models with spatially inhomogeneous nonlinearity .
the first type of lattices is composed of linear coupled oscillators except for a central region of length @xmath86 where nonlinearities are present .
we refer to this type as the lnl ( linear - nonlinear - linear ) model .
the second type is called nln ( nonlinear - linear - nonlinear ) and is the exact counterpart of the previous one , since the linear part of the lattice is located at the central @xmath86 sites .
the evolution of wave packets in nonlinear disordered chains can be expected to be selftrapped for strong nonlinearities , or show no selftrapping for weaker nonlinearities .
the existence of the selftrapping regime was theoretically predicted for the dnls model in @xcite ( see also @xcite for more details ) . according to the theorem stated in @xcite , for large enough nonlinearities ( @xmath87 ) single site excitations
can not uniformly spread over the entire lattice .
consequently , a part of the wave packet will remain localized , although the theorem does not prove that the location of this inhomogeneity is constant in time .
if the nonlinear shift @xmath88 moves the frequencies of some of the initially excited oscillators out of the linear spectrum , it tunes them out of resonance and part of the wave packet will be selftrapped . in our study
we consider initial block
wave packets , where @xmath86 central oscillators of the lattice are excited having the same norm ( energy ) .
since we consider many random disorder realizations ( of the order of few hundreds ) we expect that , on average , the linear frequencies of the initially excited lattice sites @xmath89 ( @xmath17 ) cover the whole range of permitted values @xmath9 $ ] ( @xmath90 $ ] ) . thus , some of these frequencies are tuned out of resonance if @xmath91 ( @xmath92 ) .
these conditions for the possible appearance of selftrapping are less strict than the theoretically defined ones @xcite and are , in general , in good agreement with numerical simulations . in particular , the selftrapping regime was numerically observed for single - site excitations @xcite and for extended excitations @xcite , both for the dnls and the kg models , despite the fact that the kg system conserves only the total energy @xmath65 , and the selftrapping theorem can not be applied there .
when selftrapping is avoided for @xmath93 ( @xmath94 ) , two different spreading regimes were predicted , having different dynamical characteristics : an asymptotic weak chaos regime , and a potential intermediate strong chaos one @xcite .
numerical verifications of the existence of these two regimes were presented in @xcite . in the weak chaos regime , for @xmath95 and @xmath96
most of the nms are weakly interacting with each other .
then the subdiffusive spreading of the wave packet is characterized by @xmath97 .
if the nonlinearity is weak enough to avoid selftrapping , yet strong enough to ensure @xmath98 , the strong chaos regime is realized .
wave packets in this regime initially spread faster than in the case of weak chaos , with @xmath99 .
since the norm density drops with further spreading , @xmath88 is dropping in time as well , and eventually the wave packet enters the weak chaos regime , where its evolution is characterized by slower spreading with @xmath97 . the wave packet evolution in both the weak and the strong chaos spreading regimes is also expected to be characterized by an increase of the participation number as @xmath100 when @xmath101 .
let as now discuss the spreading of wave packets when @xmath102 .
the packet will initially spread over the localization volume @xmath47 during a time interval @xmath103 , even in the absence of nonlinearities @xcite .
the initial average norm ( energy ) density @xmath104 ( @xmath105 ) of the wave packet is then lowered to @xmath106 ( @xmath107 ) .
the further spreading of the wave packet in the presence of nonlinearities is then determined by these reduced densities .
note that for single - site excitations ( @xmath108 ) , the strong chaos regime therefore completely disappears and the wave packet evolves either in the weak chaos or in the selftrapping regimes @xcite .
there are no existing theoretical predictions for wave packet dynamics in the fsw chain .
the fsw case can be considered as a strong disorder limit of the kg model , emulating the dynamics of the latter in nm space .
however , in the kg and dnls case , triplet interactions between nms are present and necessary in order to allow for finite ( though small ) resonance probabilities for small ( but finite ) energy and norm densities @xcite .
pair interactions will cease to produce nm resonances for sufficiently small densities due to level repulsion within one localization volume @xcite .
the fsw chain keeps only pair interactions . at the same time ,
level repulsion between neighboring ( interacting ) nms are absent in the fsw case .
therefore , a theoretical analysis similar to the dnls and kg case @xcite appears to be possible .
its details will be considered in a future publication .
the width of the linear spectrum @xmath109 .
the average spacing of nearest neighbor eigenvalues @xmath110 .
therefore , we expect that only the asymptotic regime of weak chaos , and the regime of selftrapping can be expected .
the lnl chain can be expected to start in a chaotic wave packet spreading regime as long as the wave packet is confined mainly to the finite size n ( nonlinear ) part .
however , the more the wave packet spreads , the more it extends into the infinitely extended l ( linear ) parts .
resonances and chaos are therefore confined to the finite n part . since distant nms in the l part are exponentially weakly interacting with the chaotic nms in the n part , their excitation - if at all - will take times which increase exponentially with growing distance .
therefore , the wave packet will spread ( if at all ) slower than any power law .
thus , the lnl model is the only model we consider here , where almost trivial slowing down of spreading is expected .
recently , the dynamics of a similar system was theoretically investigated in @xcite , where a disordered subsystem of coupled anharmonic oscillators , linearly coupled to an infinite harmonic system , was considered . in this work
the conditions which permit the persistence of the discrete breather of the isolated anharmonic system for small but nonvanishing couplings to the harmonic lattice were derived , and cases characterized by energy transfer to the harmonic system were also discussed . the nln chain is expected to behave differently .
as long as the wave packet is confined mainly to the l region , the dynamics is regular , and no spreading should occur .
for large enough time , some part of the packet will leak out into the n regions .
therefore , finally spreading of the wave packet should occur .
we consider compact dnls wave packets at @xmath111 spanning a width @xmath86 centered in the lattice , such that within @xmath86 there is a constant initial norm of @xmath112 and a random phase at each site , while outside the width @xmath86 the norm density is zero . in the kg case
, we excite each site in the width @xmath86 with the same energy , @xmath113 , i.e. initial momenta of @xmath114 with randomly assigned signs .
we use symplectic integration schemes of the saba family of integrators @xcite for the integration of equations ( [ rdnls - eom ] ) and ( [ kg - eom ] ) .
the particular symplectic scheme used for the dnls model is described in the appendix .
the number of lattice sites @xmath115 and the integration time step @xmath116 varied between @xmath117 to @xmath118 and @xmath119 to @xmath120 , in order to exclude finite size effects in the wave packet evolution , and in order to reach long integration times up to @xmath121 time units with feasible cpu times . in all our simulations , the relative energy and norm errors
are kept smaller than @xmath122 . for each parameter
set we averaged our data over @xmath123 different disorder realizations , unless otherwise stated , and denote this by @xmath124 .
in particular , we compute @xmath125 and @xmath126 , and we smooth @xmath127 and @xmath128 with a locally weighted regression algorithm @xcite , and then apply a central finite difference to calculate the local derivatives @xmath129 in this subsection we considerably extend the reports on the observation of weak chaos , strong chaos , the crossover between both , and the selftrapping regime in ref . @xcite . in fig .
[ fig_dnls_4 ] we show results for the dnls model with @xmath130 and @xmath131 , for six different values of the nonlinearity strength @xmath23 .
the time evolution of @xmath132 and @xmath133 is plotted in figs .
[ fig_dnls_4](a ) and [ fig_dnls_4](b ) , respectively .
the evolution of the compactness index @xmath134 is shown in fig .
[ fig_dnls_4](c ) . in figs .
[ fig_dnls_4](d ) and [ fig_dnls_4](e ) we plot the time dependence of the numerically computed derivatives @xmath135 and @xmath136 ( [ eq : deriv ] ) obtained from the smoothed curves of figs .
[ fig_dnls_4](a ) and [ fig_dnls_4](b ) respectively . finally , in fig .
[ fig_dnls_4](f ) the values of @xmath137 are plotted .
the weak chaos dynamics is observed in fig .
[ fig_dnls_4 ] for @xmath138 ( black curves ) and @xmath139 ( magenta curves ) .
initially , the wave packets remain localized and all quantities of fig .
[ fig_dnls_4 ] are constant with @xmath140 , @xmath141 being practically zero .
after some detrapping time @xmath142 the wave packets start to subdiffusively spread with @xmath97 and @xmath143 ( figs .
[ fig_dnls_4](a ) , ( b ) , ( d ) and ( e ) ) .
in addition , the compactness index @xmath144 ( fig . [ fig_dnls_4](c ) ) , indicating that wave packets are well thermalized inside . the tendency towards complete delocalization of wave packets
is clearly depicted in the evolution of the averaged norm fraction @xmath145 which remains at the @xmath146 initially excited sites ( fig .
[ fig_dnls_4](f ) ) .
after the detrapping time @xmath142 @xmath145 decreases continuously up to the final integration time @xmath147 . increasing the value to @xmath148 , the initial spreading dynamics enters the crossover between weak and strong chaos , and a faster spreading is observed .
spreading sets in earlier , the compactness index again indicates thermalized wave packets , and the local derivatives @xmath140 and @xmath141 increase up to 0.4 and 0.2 respectively , with a possibly very slow decreasing at even larger times .
@xmath145 again continuously decreases to zero indicating complete delocalization . for @xmath149
, we fully enter the strong chaos regime .
most importantly we observe a saturation of the local exponent @xmath140 around the theoretical value 1/2 , with a subsequent decay , again as predicted by theory @xcite , and first observed in @xcite . finally , for @xmath150 and @xmath151 ( green and brown curves respectively in fig . [ fig_dnls_4 ] )
the dynamics enters the selftrapping regime .
we observe that a part of a wave packet remains localized , while the remainder spreads .
the spreading portion results in a continuous increase of @xmath125 ( fig .
[ fig_dnls_4](a ) ) which initially is characterized by large values of @xmath152 ( fig .
[ fig_dnls_4](d ) ) . for larger time @xmath140 decreases below 1/2 .
the evolution appears to be rather complex , and is not captured by the theoretical considerations in @xcite .
the large values of @xmath140 may be due to temporal trapping and detrapping processes in this strongly nonequilibrium dynamics of the wave packet , with a final trapped packet fraction remaining .
therefore , we observe that @xmath153 starts to level off ( fig .
[ fig_dnls_4](b ) ) , @xmath141 tends to very small values ( fig .
[ fig_dnls_4](e ) ) and @xmath154 tends to zero ( fig .
[ fig_dnls_4](c ) ) . in fig .
[ fig_dnls_4](f ) , we observe that the values of @xmath155 saturate to higher values for @xmath151 . for these typical cases of weak chaos ( @xmath139 ) , strong chaos ( @xmath149 ) and selftrapping ( @xmath150 ) we present in figs .
[ fig_dnls_kg_w4_profiles](a)-(c ) the time evolution of the averaged norm density distributions @xmath156 in real space .
all simulations presented in fig . [ fig_dnls_kg_w4_profiles ] started from the same initial profile with size @xmath157 , therefore the width of the localization volume set by the linear case is the width of the distributions at the shortest times in the plots . in the weak chaos regime ( fig .
[ fig_dnls_kg_w4_profiles](a ) ) , the wave packets remain close to their initial configuration for some times , as the high @xmath158 values in the region of the initial excitation indicate , followed by delocalization at larger times .
thus , at @xmath147 the averaged wave packet has spread to about @xmath159 sites with @xmath160 .
therefore , the wave packet spreads continuously over distances which are an order of magnitude larger than the limits set by the linear theory and destroying anderson localization . in the case of strong chaos ( fig .
[ fig_dnls_kg_w4_profiles](b ) ) , spreading is even faster , leading to more extended profiles at @xmath147 : about @xmath161 sites with @xmath162 . in the selftrapping regime ( fig .
[ fig_dnls_kg_w4_profiles](c ) ) , the spreading part of the wave packet covers @xmath163 sites , with another clearly visible part staying selftrapped at the initial excitation region .
the curved fronts in the density plots in fig .
[ fig_dnls_kg_w4_profiles ] follow from the theoretical prediction @xmath101 with which leads to a packet width @xmath164 and @xmath165 . for the kg model ( [ rqkg ] ) with @xmath130 we present in fig .
[ fig_kg_4 ] similar results to the ones for the dnls model . for small values of the initial energy density @xmath166 ,
the characteristics of the weak chaos regime are observed : @xmath167 ( black curve in fig .
[ fig_kg_4](a ) ) after a detrapping time @xmath168 , wave packets remain compact as they spread since @xmath144 ( fig . [ fig_kg_4](b ) ) , and the fraction @xmath169 of the energy of the initially @xmath146 excited sites decreases ( fig .
[ fig_kg_4](d ) ) . for @xmath170 we enter the crossover region between the weak and the strong chaos regimes , with characteristics similar to the dnls case . for @xmath171 ( red curves in fig .
[ fig_kg_4 ] ) we observe the typical behavior of the strong chaos scenario : spreading is characterized by a saturated @xmath172 ( fig .
[ fig_kg_4](c ) ) for about two decades ( @xmath173 ) , followed by a crossover to the weak chaos dynamics with @xmath140 decreasing .
getting closer to the selftrapping regime for @xmath174 ( blue curves in fig .
[ fig_kg_4 ] ) , or being deep inside it for @xmath175 ( green curves in fig .
[ fig_kg_4 ] ) the characteristics of the selftrapping behavior appear , since @xmath176 decreases ( fig . [ fig_kg_4](b ) ) , and @xmath177 tends to stabilize to non - zero small values ( fig .
[ fig_kg_4](d ) ) .
similar to the @xmath150 and @xmath151 cases of the dnls model , the evolution of @xmath125 ( fig . [ fig_kg_4](a ) ) shows an initial phase of fast growth with @xmath152 ( fig .
[ fig_kg_4](c ) ) followed by a lowering in the values of @xmath140 . : evolution of ( a ) @xmath178 , ( b ) @xmath134 , ( c ) @xmath135 , and ( d ) @xmath179 versus @xmath180 for the spreading of initially compact wave packets of width @xmath146 with @xmath181 [ ( bl ) black ; ( m ) magenta ; ( r ) red ; ( b ) blue ; ( g ) green ] . in panels ( a ) , and ( c ) straight lines correspond to the theoretically predicted power laws @xmath101 with @xmath182 ( dashed lines ) and @xmath183 ( dotted lines ) .
error bars in panel ( a ) denote representative standard deviation errors . ]
our numerical results are in accord with the predictions of weak and strong chaos regimes , as well as of the crossover from strong to weak chaos .
selftrapping is observed as well , with less understood strongly nonequilibrium dynamics of the trapped and spreading packet parts .
we vehemently stress that in all our simulations we never observed any evidence of a wave packet transition from the weak chaos regime , characterized by @xmath184 , to a subsequent slowing down of spreading , which would lead to @xmath185 .
to search for potential deviations from the predicted spreading laws , we turn to large values of @xmath10 for the dnls system . in all our simulations we had @xmath186 . in particular
, we set @xmath187 and considered the cases with @xmath188 and @xmath189 . for large values of @xmath10
we expect only the weak chaos and the selftrapping dynamical regimes to be observed @xcite . for each value of @xmath10 three different values of @xmath23
were considered , one being in the weak chaos regime and the other two in the selftrapping regime . in particular
, we considered @xmath190 , @xmath191 and @xmath192 for @xmath188 and @xmath193 , @xmath194 and @xmath195 for @xmath189 .
the obtained results are shown in fig .
[ fig_dnls_w_15 ] for @xmath188 and in fig .
[ fig_dnls_w_40 ] for @xmath189 . : evolution of ( a ) @xmath178 , ( b ) @xmath134 , ( c ) @xmath135 , and ( d ) @xmath196 versus @xmath180 for the spreading of initially compact wave packets of width @xmath187 with @xmath197 [ ( m ) magenta ; ( g ) green ; ( br ) brown ] .
mean values are averaged quantities over 100 disorder realizations . in panels ( a ) , and
( c ) straight lines correspond to theoretically predicted weak chaos behavior @xmath97 . ] : evolution of ( a ) @xmath178 , ( b ) @xmath134 , ( c ) @xmath135 , and ( d ) @xmath196 versus @xmath180 for the spreading of initially compact wave packets of width @xmath187 with @xmath198 [ ( m ) magenta ; ( g ) green ; ( br ) brown ] .
mean values are averaged quantities over 100 disorder realizations . in panels ( a ) , and ( c ) straight lines correspond to theoretically predicted weak chaos behavior @xmath97 . ] for large @xmath10 , the localization volume @xmath47 decreases drastically , so that eventually the overlap integrals become small as well .
therefore , the values of @xmath23 at which spreading becomes visible will increase . in the weak chaos spreading regime
the detrapping time @xmath142 becomes large and we start observing spreading after long time intervals . for @xmath188 and @xmath190
we find @xmath199 ( fig . [ fig_dnls_w_15](a ) ) .
the local derivative @xmath140 increases from zero showing a tendency to approach the theoretically predicted value @xmath182 ( fig . [ fig_dnls_w_15](c ) ) , and both @xmath154 ( fig .
[ fig_dnls_w_15](b ) ) and @xmath200 ( fig . [ fig_dnls_w_15](d ) ) start to decrease .
we note that since @xmath201 for large @xmath10 , we measure the time evolution of the fraction @xmath202 of the norm density of the @xmath187 initially excited sites .
the detrapping time @xmath142 increases as @xmath10 increases .
this is seen from the results for @xmath189 , @xmath193 ( magenta curves in fig .
[ fig_dnls_w_40 ] ) . in this case
, we have to wait at least up to @xmath203 in order to get some evidence that spreading starts , since after that time @xmath204 starts to slightly grow ( fig . [ fig_dnls_w_40](a ) ) , while @xmath205 ( fig . [ fig_dnls_w_40](b ) ) and @xmath206 ( fig .
[ fig_dnls_w_40](d ) ) start to decrease .
this increase of @xmath142 happens despite the fact that the nonlinearity strength @xmath23 also increased by a factor of two as compared to the @xmath188 case . nevertheless even in this case of large @xmath189
we are able to numerically observe the onset of spreading in the weak chaos regime .
increasing @xmath10 to even higher values pushes @xmath142 to values larger than the final integration time @xmath147 used in our simulations .
with increasing @xmath23 we observe selftrapping , but a part of the wave packet spreads and @xmath125 increases ( green and brown curves in figs . [ fig_dnls_w_15](a ) and [ fig_dnls_w_40](a ) ) .
the average compactness index @xmath154 decreases , which is a clear indication that a part of the wave packet remains localized , and reaches smaller final values for larger @xmath23 ( figs .
[ fig_dnls_w_15](b ) and [ fig_dnls_w_40](b ) ) .
the selftrapping of the wave packets is also clearly seen from the evolution of @xmath206 . in fig .
[ fig_dnls_w_15](d ) we see that for @xmath188 , @xmath191 ( green curve ) and @xmath192 ( brown curve ) @xmath206 decreases due to the spreading of a part of wave packets , while , finally it shows a tendency to level off to a positive value , indicating that part of the wave packets remains localized .
similar behaviors of @xmath206 are observed in fig .
[ fig_dnls_w_40](d ) for the @xmath189 case with @xmath194 ( green curve ) and @xmath195 ( brown curve ) , although the plateauing of @xmath206 is not as clear as in fig . [ fig_dnls_w_15](d ) .
the numerically computed exponents @xmath140 exhibit the typical behavior of selftrapping seen in fig .
[ fig_dnls_4](d ) .
for @xmath188 they increase reaching values larger than @xmath207 and afterwards decrease towards @xmath184 ( fig .
[ fig_dnls_w_15](c ) ) . for @xmath189
a similar behavior is observed for @xmath195 , while for @xmath194 @xmath140 seems to approach the theoretically predicted value @xmath207 from below .
therefore , even for strong disorder , the dynamics of wave packets evolves according to the theoretical predictions .
most importantly , we do not observe a slowing down of the wave packet below the limits set by the weak chaos regime . to further probe a possible slowing down of wave packet spreading beyond the limits of weak chaos , we turn to the fsw model ( [ fsw ] ) . in fig . [
fig : fswmodel ] we present results with initial compact wave packets of width @xmath146 , with energy density @xmath208 , similarly to the kg model ( [ rqkg ] ) .
we observe that also for this model subdiffusive spreading occurs , because the second moment and the participation number ( red and blue curves respectively in fig .
[ fig : fswmodel](a ) ) increase continuously , and the corresponding exponents @xmath140 and @xmath141 ( fig .
[ fig : fswmodel](c ) ) tend to eventual constant non - zero values .
the fraction @xmath209 of energy remaining in the @xmath146 initially excited sites ( fig .
[ fig : fswmodel](d ) ) decreases as time increases indicating the delocalization of the wave packets . the compactness index ( fig .
[ fig : fswmodel](b ) ) has a different behavior with respect to what we have seen in the rest of our simulations , as it decreases slowly up to @xmath210 , with a subsequent increase .
therefore , the wave packet is highly inhomogeneous for almost all the integration time , violating the assumptions which are used in the theoretical considerations of weak chaos @xcite .
the observed subdiffusive spreading may still not be in its final asymptotic range .
still , we again do not see any signature of a slowing down of this subdiffusive process , as was reported in @xcite where a similar model was considered .
clearly the fsw model calls for a thorough and independent study . [ ( r ) red curve ] and @xmath211 [ ( b ) blue curve ] , ( b ) @xmath134 , ( c ) @xmath135 [ ( r ) red curve ] and @xmath136 [ ( b ) blue curve ] , and ( f ) @xmath212 versus @xmath180 for initially compact wave packets of width @xmath146 with @xmath208 . ] in fig .
[ fig_kg_hybrid ] we present results for the kg , lnl and nln models for @xmath130 , @xmath146 and @xmath213 . for comparison
we also include the results for the kg model ( [ rqkg ] ) ( magenta curves ) with @xmath213 , for which subdiffusive spreading in the weak chaos regime is observed . the kg - nln model ( green curves in fig
. [ fig_kg_hybrid ] ) exhibits a similar behavior , since both the second moment ( fig .
[ fig_kg_hybrid](a ) ) and the participation number ( fig .
[ fig_kg_hybrid](b ) ) start to grow after some detrapping time @xmath168 .
this time is larger than the detrapping time of the kg model ( @xmath214 ) , because a wave packet in the kg - nln model initially evolves in an almost linear system and only after some large time , when it has spread significantly to the nonlinear part of the lattice , spreading takes on characteristics of the purely nonlinear model . on the other hand the evolution of all quantities of fig .
[ fig_kg_hybrid ] for the kg - lnl system ( red curves ) follows the kg model until @xmath215 , because initially the wave packets evolve in the same nonlinear system .
later on the wave packet enters the l ( linear ) parts of the system .
thus , spreading starts to retard , and both @xmath132 ( fig .
[ fig_kg_hybrid](a ) ) and @xmath211 ( fig . [ fig_kg_hybrid](b ) ) show a characteristic slowing down in the exponents @xmath140 ( fig .
[ fig_kg_hybrid](d ) ) and @xmath141 ( fig . [ fig_kg_hybrid](e ) ) .
in addition , @xmath145 ( fig . [ fig_kg_hybrid](f ) ) saturates at finite non - zero values , indicating that wave packets tend to localize again . for all three kg models ,
the values of @xmath154 ( fig .
[ fig_kg_hybrid](c ) ) show that wave packets do not become sparse and inhomogeneous in the course of time .
we obtained similar results for the lnl and nln dnls models for @xmath130 , @xmath146 and @xmath139 .
thus , also for the lnl and nln models spreading is observed , and only in the case of the lnl models we have observed a slowing down of the spreading , as expected .
we considered several models of disordered nonlinear one - dimensional lattices and performed extensive numerical simulations of norm ( energy ) propagations .
since we focused on the dynamical spreading of fronts , we prepared initial block wave packet profiles , having widths equal or larger than the average localization volume defined by the linear problem . while not performed , we expect similar behaviors for initial gaussian profiles , again where the width ( for gaussians , say the standard deviation ) is on par with the average localization volume .
we carefully studied statistical properties of the dynamics , by varying the values of disorder and nonlinearity strengths over a wide interval , and by averaging results over many disorder realizations .
our results agree quite well with our theoretical expectations for the existence of the weak and strong chaos regimes .
the main outcome of our study is that in the presence of nonlinearities we always observe subdiffusive spreading , so that the second moment grows initially as @xmath216 with @xmath217 , showing signs of a crossover to the asymptotic @xmath97 law at larger times .
remarkably , subdiffusive spreading is also observed for large disorder strengths , when the localization volume ( which defines the number of interacting partner modes ) tends to one .
frhlich - spencer - wayne models which take the disorder strength to its infinite limits , are also showing subdiffusive growth .
most remarkably , in none of our studies ( except the artificial lnl case ) did we encounter a slowing down of spreading beyond the limits set by the weak chaos predictions .
therefore , our numerical data support the conjecture , that the wave packets , once they spread , will do so up to infinite times in a subdiffusive way , bypassing anderson localization of the linear wave equations .
the only cases where spreading shows a tendency to stop are the lnl models , for which nonlinearities are absent everywhere except inside a finite - size central region , where the initial wave packet is launched . in these models ,
when wave packets have spread substantially , their chaotic component in the central region of the lattice becomes weak , and distant normal modes in the linear parts of the system are exponentially weakly coupled to the central nonlinear region .
when the nonlinearity strength tends to smaller values , waiting ( detrapping ) times for wave packet spreading of compact initial excitations increase beyond the detection capabilities of our computational tools . the corresponding question of whether a kam regime can be entered at finite nonlinearity strength was addressed in @xcite and is analyzed in detail in a forthcoming work @xcite .
we thank s. aubry , m. mulansky , a. pikovsky , r. schilling and d. shepelyansky for useful discussions .
s. was partly supported by the european research project `` complex matter '' , funded by the gsrt of the ministry education of greece under the era - network complexity program .
we first discuss a novel pq method which we designed to integrate the dnls equations locally .
previously used methods employ a transformation of the wave function from real into fourier space and back , at each integration step .
these transformations induce small but observable corrections in the tails of the wave packet , which slowly but steadily grow in time . in such a case
we will have to stop the integration once this noisy background reaches a substantial level .
the pq method avoids the generation of this background by simply not performing the fourier transformation .
instead the pq method integrates the dnls equations in real space . the canonical transformation @xmath218 of the complex variable @xmath219 in eq . ( [ rdnls ] )
transforms ( [ rdnls ] ) into @xmath220 where @xmath221 and @xmath16 are generalized coordinates and momenta , respectively .
if a hamiltonian function can be split into two integrable parts , then a symplectic integration scheme can be used for the integration of its equations of motion .
one possible splitting of the dnls hamiltonian ( [ rdnls_pq ] ) into two separate hamiltonian functions @xmath222 and @xmath223 is @xmath224 hamiltonian @xmath222 is integrable and the operator @xmath225 which propagates the set of initial conditions @xmath226 at time @xmath227 , to their final values @xmath228 at time @xmath229 is @xmath230 with @xmath231 .
hamiltonian @xmath223 of eq .
( [ eq : ham_equiv_dnls ] ) is not integrable , thus the operator @xmath232 can not be written explicitly . if we consider @xmath223 as a separate hamiltonian function and again split it as @xmath233 , the component parts @xmath234 are integrable , under the corresponding operators @xmath235 and @xmath236 this technique of splitting the hamiltonian into multiple parts has been used in different applications of symplectic integrators ( see for example @xcite ) . in our simulations
we successively apply the sbab@xmath237 symplectic integrator @xcite twice : first for the split ( @xmath238 ) of the dnls hamiltonian , eq .
( [ rdnls_pq ] ) , and second for the split @xmath233 in eq .
( [ eq : ham_equiv_dnls ] ) .
the solution for the equations of motion from the hamiltonian eq .
( [ rdnls_pq ] ) is thus approximated by the application of @xmath239 simple operators on an initial condition @xmath226 , since r. dalichaouch , j.p .
armstrong , s. schultz , p.m. platzman , s.l .
mccall , _ nature _ * 354 * , 53 ( 1991 ) ; c. dembowski , h .-
grf , r. hofferbert , h. rehfeld , a. richter , t. weiland , _ phys .
e _ * 60 * , 3942 ( 1999 ) ; j.d .
bodyfelt , m. c. zheng , t. kottos , u. kuhl , h .- j .
stckmann , _ phys .
lett . _ * 102 * , 253901 ( 2009 ) . | we probe the limits of nonlinear wave spreading in disordered chains which are known to localize linear waves .
we particularly extend recent studies on the regimes of strong and weak chaos during subdiffusive spreading of wave packets [ epl * 91 * , 30001 ( 2010 ) ] and consider strong disorder , which favors anderson localization .
we probe the limit of infinite disorder strength and study frhlich - spencer - wayne models .
we find that the assumption of chaotic wave packet dynamics and its impact on spreading is in accord with all studied cases .
spreading appears to be asymptotic , without any observable slowing down .
we also consider chains with spatially inhomogeneous nonlinearity which give further support to our findings and conclusions . |
the authors are grateful to l. i. gurvits for sending his compilation of the data as well as for helpful discussions .
we would like to thank gang chen for useful discussions .
this work was partially suported by the conselho nacional de desenvolvimento cientfico e tecnolgico - cnpq , pronex / finep ( no . 41.96.0908.00 ) and fapesp ( 00/06695 - 0 ) .
alcaniz j. s. & lima j. a. s. 1999 , apj , 521 , l87 alcaniz j. s. & lima j. a. s. 2001 , apj , 550 , l133 balbi a. , baccigalupi c. , matarrese s. , perrotta f. , vittorio n. 2001 , apj 547 l89 caldwell r. r. , dave r. & steinhardt p. j. 1998 , phys . rev . lett . 80 , 1582 carretta e. _ et al .
_ 2000 , apj , 533 , 215 carvalho j. c. , lima j. a. s. & waga i. 1992 , phys .
d46 , 2404 chiba t. , sugiyama n. & nakamura t. 1997 , mnras , 289 , l5 corasaniti p. s. & copeland e. j. 2001 , preprint ( astro - ph/0107378 ) de bernardis p. _ et al .
_ 2000 , nature , 404 , 955 dekel a. , burstein d. & white s. , in critical dialogues in cosmology , edited by n. turok world scientific , singapore ( 1997 ) dev a. , jain d. , panchapakesan n. , mahajan m. & bhatia v. 2001 , preprint ( astro - ph/0104076 ) dunlop j. s. _ et al .
_ 1996 , nature , 381 , 581 efstathiou g. 1999 , mnras , 310 , 842 freese k. , adams f. c. , frieman j. a. & mottola e. 1987 , nucl .
b287 , 797 frieman j. a. , hill c. t. , stebbins a. & waga i. 1995 , phys .
75 , 2077 garnavich p. m. _ et al .
_ 1998 , apj , 509 , 74 gurvits l. i. 1994 , ap . j. 425 , 442 gurvits l. i. , kellermann k. i. & frey s. 1999 , a&a 342 , 378 jackson j. c. & dodgson m. 1997 , mnras 278 , 603 jackson j. c. & dodgson m. 1997 , mnras 285 , 806 jain d. , dev a. , panchapakesan n. , mahajan m. & bhatia v. 2001 , preprint ( astro - ph/0105551 ) kellermann k. i. 1993 , nature 361 , 134 krauss l. m. & schramm d. n. 1993 , apj 405 , l43 krauss l. m. 1997 , apj , 480 , 466 krauss l. m. 2000 , phys . rep . 333 , 33 lima j. a. s. & maia j. m. f. 1994 , phys .
rev d49 , 5597 lima j. a. s. & alcaniz j. s. 2000a , mnras , 317 , 893 lima j. a. s. & alcaniz j. s. 2000b , a&a , 357 , 393 lima j. a. s. & alcaniz j. s. 2000c , gen .
32 , 1851 ozer m. & taha o. 1987 , nucl .
phys . , b287 , 776 peebles p. j. e. 1984 , apj 284 , 439 peebles p. j. e. & ratra b. 1988 , apj 325 , l17 podariu s. & ratra b. 2000 , apj 532 , 109 perlmutter s. , _ et al .
_ 1998 , nature , 391 , 51 perlmutter s. , _ et al .
_ 1999 , apj , 517 , 565 perlmutter s. , turner m. s. & white m. 1999 , phys . rev .
lett . , 83 , 670 ratra b. & peebles p. j. e. 1988 , phys .
d37 , 3406 rengel m. , mateu j. & bruzual g. 2001 , iau symp . 207 , _ extragalactic star clusters
e. grebel , d. geisler , d. minniti ( in press ) , preprint ( astro - ph/0106211 ) riess a. _ et al .
_ 1998 , aj , 116 , 1009 saini t. d. , raychaudhury s. , sahni v. & starobinsky a. a. 2000 , phys .
, 1162 turner m. s. & white m. 1997 , phys . rev .
d56 , r4439 sandage a. 1988 , ann .
26 , 561 vishwakarma r. g. 2001 , class .
18 , 1159 waga i. & miceli a. p. m. r. 1999 , phys .
rev d59 , 103507 wang l. , caldwell r. r. , ostriker j. p. , & steinhardt p. j. 2000 , apj 530 , 17 rlll + & & & + + cmb+sne ia .. & turner & white ( 1997 ) & @xmath109 & @xmath110 + & efstathiou ( 1999 ) & @xmath16 & @xmath111 + sne ia ............ & garnavich _ et al . _
( 1998 ) & @xmath16 & @xmath112 + sgl+sne ia .. & waga & miceli ( 1999 ) & @xmath113 & @xmath114 + sne ia+lss ... & perlmutter _ et al . _
( 1999 ) & @xmath16 & @xmath111 + various ............ & wang _ et al . _
( 1999 ) & @xmath115 & @xmath116 + ohrgs .......... & lima & alcaniz ( 2000a ) & @xmath117 & @xmath118 + cmb ............... & balbi _ et al . _
( 2001 ) & @xmath117 & @xmath119 + & corasaniti & copeland ( 2001 ) & @xmath16 & @xmath120 + sgl ................ & jain _ et al . _ ( 2001 ) & @xmath121 & @xmath122 , @xmath123 + | we investigate observational constraints on the cosmic equation of state from measurements of angular size for a large sample of milliarcsecond compact radio - sources .
the results are based on a flat friedmann - robertson - walker ( frw ) type models driven by non - relativistic matter plus a smooth dark energy component parametrized by its equation of state @xmath0 ( @xmath1 ) .
the allowed intervals for @xmath2 and @xmath3 are heavily dependent on the value of the mean projected linear size @xmath4 . for @xmath5
pc , we find @xmath6 , @xmath7 and @xmath8 , @xmath9 ( 68@xmath10 c.l . ) , respectively . as a general result
, this analysis shows that if one minimizes @xmath11 for the parameters @xmath4 , @xmath3 and @xmath2 , the conventional flat @xmath12cdm model ( @xmath13 ) with @xmath14 and @xmath15pc is the best fit for these angular size data .
a large number of recent observational evidences strongly suggest that we live in a flat , accelerating universe composed by @xmath16 1/3 of matter ( barionic + dark ) and @xmath16 2/3 of an exotic component with large negative pressure , usually named dark energy or quintessence " .
the basic set of experiments includes : observations from sne ia ( perlmutter _ et al .
_ 1998 ; 1999 ; riess et al .
1998 ) , cmb anisotropies ( de bernardis _ et al .
_ 2000 ) , large scale structure ( bahcall 2000 ) , age estimates of globular clusters ( carretta _ et al .
_ 2000 ; krauss 2000 ; rengel _ et al .
_ 20001 ) and old high redshift galaxies ( ohrg s ) ( dunlop 1996 ; krauss 1997 ; alcaniz & lima 1999 ; alcaniz & lima 2001 ) .
it is now believed that such results provide the remaining piece of information connecting the inflationary flatness prediction ( @xmath17 ) with astronomical observations , and , perhaps more important from a theoretical viewpoint , they have stimulated the current interest for more general models containing an extra component describing this unknown dark energy , and simultaneously accounting for the present accelerated stage of the universe .
the absence of a convincing evidence concerning the nature of this dark component gave origin to an intense debate and mainly to many theoretical speculations in the last few years .
some possible candidates for quintessence " are : a vacuum decaying energy density , or a time varying @xmath12-term ( ozer & taha 1987 ; freese 1987 ; carvalho _ et al _ 1992 , lima and maia 1994 ) , a relic scalar field ( peebles & ratra 1988 ; frieman _ et al _ 1995 ; caldwell _ et al _ 1998 ; saini _ et al _ 2000 ) or still an extra component , the so - called x - matter " , which is simply characterized by an equation of state @xmath18 , where @xmath19 ( turner & white 1997 ; chiba _ et al _ 1997 ) and includes , as a particular case , models with a cosmological constant ( @xmath12cdm ) ( peebles 1984 ) . for x - matter " models ,
several results suggest @xmath20 and @xmath21 .
for example , studies from gravitational lensing + sne ia provide @xmath22 at 68@xmath10 c.l .
( waga & miceli 1999 ; see also dev _ et al .
_ 2001 ) .
limits from age estimates of old galaxies at high redshifts require @xmath23 for @xmath24 ( lima & alcaniz 2000a ) .
in addition , constraints from large scale structure ( lss ) and cosmic microwave background anisotropies ( cmb ) complemented by the sn ia data , require @xmath25 and @xmath26 ( @xmath27 c.l . ) for a flat universe ( garnavich _ et al _ 1998 ; perlmutter _ et al _ 1999 ; efstathiou 1999 ) , while for universes with arbitrary spatial curvature these data provide @xmath28 ( efstathiou 1999 ) . on the other hand ,
although carefully investigated in many of their theoretical and observational aspects , an overview on the literature shows that a quantitative analysis on the influence of a quintessence " component ( @xmath29 ) in some kinematic tests like angular size - redshift relation still remains to be analysed .
recently , lima & alcaniz ( 2000b ) studied some qualitative aspects of this test in the context of such models , with particular emphasis for the critical redshift @xmath30 at which the angular size takes its minimal value .
as a general conclusion , it was shown that this critical redshift can not discriminate between world models since different scenarios can provide similar values of @xmath30 ( see also krauss & schramm 1993 ) .
this situation is not improved even if evolutionary effects are taken into account .
in particular , for the observationally favoured open universe ( @xmath31 ) we found @xmath32 , a value that can also be obtained for quintessence models having @xmath33 and @xmath34 .
qualitatively , it was also argued that if the predicted @xmath30 is combinated with other tests , some interesting cosmological constraints can be obtained . in this letter , we focus our attention on a more quantitative analysis .
we consider the @xmath35 data of compact radio sources recently updated and extended by gurvits _
et al . _
( 1999 ) to constrain the cosmic equation of state .
we show that a good agreement between theory and observation is possible if @xmath6 , @xmath7 and @xmath36 , @xmath9 ( 68@xmath10 c.l . ) for values of the mean projected linear size between @xmath5 pc , respectively .
in particular we find that a conventional cosmological constant model ( @xmath13 ) with @xmath14 and @xmath37 pc is the best fit model for these data with @xmath38 for 9 degrees of freedom . for spatially flat , homogeneous , and isotropic cosmologies driven by nonrelativistic matter plus an exotic component with equation of state , @xmath39 , the einstein field equations take the following form : @xmath40 , \ ] ] @xmath41 , \ ] ] where an overdot denotes derivative with respect to time , @xmath42 is the present value of the hubble parameter , and @xmath3 and @xmath43 are the present day matter and quintessence density parameters . as one may check from ( 1 ) and ( 2 ) , the case @xmath44 corresponds effectively to a cosmological constant .
in such a background , the angular size - redshift relation for a rod of intrinsic length @xmath4 can be written as ( sandage 1988 ) @xmath45 in the above expression @xmath46 is the angular - size scale expressed in milliarcseconds ( marcs ) @xmath47 where @xmath4 is measured in parsecs ( for compact radio - sources ) , and the dimensionless coordinate @xmath48 is given by ( lima & alcaniz 2000b ) @xmath49^{{1}\over{2}}}.\ ] ] the above equations imply that for given values of @xmath4 , @xmath3 and @xmath2 , the predicted value of @xmath50 is completely determined .
two points , however , should be stressed before discussing the resulting diagrams .
first of all , the determination of @xmath3 and @xmath2 are strongly dependent on the adopted value of @xmath4 . in this case , instead of assuming a especific value for the mean projected linear size , we have worked on the interval @xmath51 pc , i.e. , @xmath52 pc for @xmath53 , or equivalently , @xmath54 marcs .
second , following kellermann ( 1993 ) , we assume that possible evolutionary effects can be removed out from this sample because compact radio jets are ( i ) typically less than a hundred parsecs in extent , and , therefore , their morphology and kinematics do not depend considerably on the intergalactic medium and ( ii ) they have typical ages of some tens of years , i.e. , they are very young compared to the age of the universe .
in our analysis we consider the angular size data for milliarcsecond radio - sources recently compiled by gurvits _
et al . _
( 1999 ) .
this data set , originally composed by 330 sources distributed over a wide range of redshifts ( @xmath55 ) , was reduced to 145 sources with spectral index @xmath56 and total luminosity @xmath57 w / hz in order to minimize any possible dependence of angular size on spectral index and/or linear size on luminosity .
this new sample was distributed into 12 bins with 12 - 13 sources per bin ( see fig .
1 ) . in order to determine the cosmological parameters @xmath3 and @xmath2
, we use a @xmath11 minimization neglecting the unphysical region @xmath58 , @xmath59^{2}}{\sigma_{i}^{2}}},\ ] ] where @xmath60 is given by eqs .
( 3 ) and ( 5 ) and @xmath61 is the observed values of the angular size with errors @xmath62 of the @xmath63th bin in the sample . figure 1 displays the binned data of the median angular size plotted against redshift .
the curves represent flat quintessence models with @xmath64 and some selected values of @xmath2 .
as discussed in lima & alcaniz ( 2000b ) , the standard open model ( thick line ) may be interpreted as an intermediary case between @xmath12cdm ( @xmath13 ) and quintessence models with @xmath65 . in fig .
2 we show contours of constant likelihood ( 95@xmath10 and 68@xmath10 ) in the plane @xmath66 for the interval @xmath51 pc . for @xmath67 pc ( @xmath68 marcs ) , the best fit occurs for @xmath69 and @xmath70 .
as can be seen there , this assumption provides @xmath71 and @xmath72 at 1@xmath73 . in the subsequent panels of the same figure
similar analyses are displayed for @xmath74 pc ( @xmath75 marcs ) , @xmath76 pc ( @xmath77 marcs ) and @xmath78 pc ( @xmath79 marcs ) , respectively . as should be physically expected , the limits are now much more restrictive than in the previous case because for the same values of @xmath61 it is needed larger @xmath80 ( for larger @xmath4 ) and , therefore , smaller values of @xmath2 . for @xmath78 pc , we find @xmath81 and @xmath13 as the best fit model . for intermediate values of @xmath4 , namely , @xmath82 pc ( @xmath83 marcs ) and @xmath84 pc ( @xmath77 marcs ) , we have @xmath85 , @xmath86 and @xmath87 and @xmath13 , respectively . in particular , for smaller values of @xmath4 , e.g. , @xmath88 pc ( @xmath89 marcs ) we find @xmath90 , @xmath91 . as a general result ( independent of the choice of @xmath4 ) ,
if we minimize @xmath11 for @xmath4 , @xmath3 and @xmath2 , we obtain @xmath92 pc ( @xmath93 marcs ) , @xmath14 and @xmath94 with @xmath38 for 9 degrees of freedom ( see table 1 ) .
it is worth notice that our results are rather different from those presented by jackson & dodgson ( 1996 ) based on the original kellermann s data ( kellermann 1993 ) .
they argued that the kellermann s compilation favours open and highly decelerating models with negative cosmological constant .
later on , they considered a bigger sample of 256 sources selected from the compilation of gurvits ( 1994 ) and concluded that the standard flat cdm model is ruled out at @xmath95 confidence level whereas low - density models with a cosmological constant of either sign are favoured ( jackson & dodgson 1997 ) .
more recently , vishwakarma ( 2001 ) used the updated data of gurvits _
et al . _
( 1999 ) to compare varying and constant @xmath12cdm models .
he concluded that flat @xmath12cdm models with @xmath14 are favoured . at this point
it is also interesting to compare our results with some recent determinations of @xmath2 derived from independent methods .
recently , garnavich _ et al . _
( 1998 ) using the sne ia data from the high - z supernova search team found @xmath96 ( @xmath27 c.l . ) for flat models whatever the value of @xmath3 whereas for arbitrary geometry they obtained @xmath26 ( @xmath27 c.l . ) . as commented there , these values are inconsistent with an unknown component like topological defects ( domain walls , string , and textures ) for which @xmath97 , being @xmath98 the dimension of the defect . the results by garnavich
_ et al . _
( 1998 ) agree with the constraints obtained from a wide variety of different phenomena ( wang _ et al .
_ 1999 ) , using the concordance cosmic " method .
their combined maximum likelihood analysis suggests @xmath99 , which is less stringent than the upper limits derived here for values of @xmath100 pc .
more recently , balbi _ et al . _
( 2001 ) investigated cmb anisotropies in quintessence models by using the maxima-1 and boomerang-98 published bandpowers in combination with the cobe / dmr results ( see also corasaniti & copeland 2001 ) .
their analysis sugests @xmath101 and @xmath102 whereas jain _ et al _ ( 2001 ) found , by using image separation distribution function of lensed quasars , @xmath103 , for the observed range of @xmath104 ( dekel _ et al .
_ 1997 ) . these and other recents results are summarized in table 2 .
let us now discuss briefly these angular size constraints whether the adopted x - matter model is replaced by a scalar field motivated cosmology , as for instance , that one proposed by peebles and ratra ( 1988 ) .
these models are defined by power law potentials , @xmath105 , in such a way that the parameter of the effective equation of state ( @xmath106 ) may become constant at late times ( or for a given era ) . in this case , as shown elsewhere ( lima & alcaniz 2000c ) , the dimensionless quantity @xmath48 defining the angular size reads @xmath107^{{1}\over{2}}}.\ ] ] comparing the above expression with ( 5 ) we see that if @xmath108 this class of models may reproduce faithfully the x - matter constraints based on the angular size observations presented here .
however , as happens with the supernovae type ia data ( podariu & ratra 2000 ) , the two sets of confidence contours may differ significantly if one goes beyond the time independent equation of state approximation .
naturally , a similar behavior is expected if generic scalar field potentials are considered .
finally , we stress that measurements of angular size from distant sources provide an important test for world models .
however , in order to improve the results a statistical study describing the intrinsic lenght distribution of the sources seems to be of fundamental importance . on the other hand , in the absence of such analysis but living in the era of _ precision cosmology _
, one may argue that reasonable values for astrophysical quantities ( like the characteristic linear size @xmath4 ) can be infered from the best cosmological model .
as observed by gurvits ( 1994 ) , such an estimative could be useful for any kind of study envolving physical parameters of active galactic nuclei ( agn ) . in principle , by knowing @xmath4 and assuming a physical model for agn , a new method to estimate the hubble parameter could be established . |
let @xmath5 be a hausdorff topological group .
@xmath5 is said to be * amenable * if every jointly continuous action on a compact hausdorff space @xmath6 ( often called a @xmath5-flow ) supports an invariant borel probability measure .
it is normally quite difficult to determine wheter a given topological group @xmath5 is amenable or not .
however , there are some observations we can make to simplify our discussion . if @xmath6 is a @xmath5-flow , a
_ subflow _
@xmath7 is a compact subspace which is invariant under @xmath5-action . using zorn s lemma
, we see that each @xmath5-flow contains a minimal subflow . therefore to show that @xmath5 is amenable
, it is enough to show that every minimal @xmath5-flow supports an invariant borel probability measure .
we can simplify things even further ; if @xmath6 and @xmath8 are @xmath5-flows , we say that @xmath9 is a @xmath5-map if @xmath10 is continuous and @xmath11 for each @xmath12 .
it is a fact that each topological group @xmath5 admits , up to isomorphism , a * universal minimal flow * @xmath13 , a flow which is minimal and such that for any other minimal @xmath5-flow @xmath6 , there is a @xmath5-map @xmath14 ( see [ a ] for further reading ) .
hence by push forward , @xmath5 is amenable if and only if the flow @xmath13 supports an invariant borel probability measure .
kechris , pestov , and todorcevic in [ kpt ] provided a way of explicitly describing @xmath13 in many instances where @xmath5 is the automorphism group of a frass structure .
we will discuss this in much more detail in section 2 .
in particular , having a concrete representation of @xmath13 allows one to address questions of amenability .
kechris and soki in [ ks ] take advantage of this to show that the automorphism groups of the random poset and the random distributive lattice are not amenable . indeed , amenability is not the only property of topological groups @xmath5 for which is it sufficient to verify a property only for the flow @xmath13 .
@xmath5 is said to be * uniquely ergodic * if each minimal @xmath5-flow supports a unique invariant borel probability measure , and @xmath5 is said to have the * generic point property * if for each minimal flow @xmath6 , there is @xmath12 with @xmath15 comeager in @xmath6 .
it is worth noting that any comeager orbit must be unique since the intersection of two comeager sets is comeager , hence nonempty . as promised , @xmath5 is uniquely ergodic iff @xmath13 supports a unique invariant borel probability measure , and @xmath5 has the generic point property iff there is @xmath16 with @xmath15 comeager .
these two definitions provide us with two notions of a canonical choice of `` large '' subset of @xmath13 , namely the support of the unique measure and the comeager orbit . in all examples of groups previously shown to be uniquely ergodic and to have the generic point property , the measure was in fact supported on the generic orbit .
angel , kechris , and lyons in [ akl ] asked whether this phenomenon holds generally .
this paper is divided into two main parts .
the first half generalizes the methods in [ ks ] and finds a necessary and sufficient condition for those groups whose universal minimal flows can be described using the methods of [ kpt ] to be amenable .
armed with this condition , we consider the ultrahomogeneous directed graph @xmath0 and the boron tree structure @xmath1 , which are defined in sections 4 and 5 , respectively .
we show that : the groups @xmath17 and @xmath18 are not amenable . the second half answers in the negative the question of angel , kechris , and lyons by exhibiting the following counterexample ; fix a finite field @xmath4 , and let @xmath3 be the countably infinite vector space over @xmath4 . in section 6 ,
we prove : the unique invariant measure on @xmath19 is not supported on the generic orbit . of course
, after asserting where the unique measure in this example is _ not _ supported , it is natural to ask where it is supported .
this is the focus of sections 7 and 8 .
i would like to thank alekos kechris , miodrag soki , and anush tserunyan for many helpful discussions and comments , as well as ola kwiatkowska and robin tucker - drob for patiently lending an ear .
i would also like to thank an anonymous referee for many helpful comments and suggestions .
this research was supported by the caltech surf office and the logic in southern california nsf grant ems w21-rtg .
this section describes the necessary model theoretic background as well as a brief summary of the material from [ kpt ] . a _ language _
@xmath20 is a set of relation , function , and constant symbols .
each relation and function symbol has an arity @xmath21 for @xmath22 . in this paper , all languages
are assumed to be countable .
an _ @xmath23-structure _ @xmath24 consists of a set @xmath25 called the _ universe _ of @xmath26 , @xmath27 , @xmath28 , and @xmath29 .
an _ embedding / isomorphism _ @xmath30 of @xmath23-structures is an injective / bijective map which preserves structure : @xmath31 , @xmath32 , and @xmath33 .
we write @xmath34 if there is an embedding from @xmath26 into @xmath35 , and we write @xmath36 if there is an isomorphism between the two .
we say @xmath26 is a _ substructure _ of @xmath35 , written @xmath37 , if @xmath38 and @xmath39 is closed under the functions @xmath40 . if @xmath41 and @xmath26 is an @xmath23-structure , we write @xmath42 for the structure obtained by taking @xmath26 and ignoring the interpretations of relation , function , and constant symbols in @xmath43 .
similarly , if @xmath44 is a class of @xmath23-structures , we write @xmath45 for the class @xmath46 .
let @xmath44 be a class of finite @xmath23-structures closed under isomorphism with countably many isomorphism types and such that there are structures in @xmath44 of arbitrarily large finite cardinality .
we call @xmath44 a * frass class * if the following three items hold : * hereditary property ( hp ) : if @xmath47 and @xmath34 , then @xmath48 . *
joint embedding property ( jep ) : if @xmath49 , then there is @xmath50 such that @xmath51 .
* amalgamation property ( ap ) : if @xmath52 and @xmath53 , @xmath54 are embeddings , then there is @xmath55 and embeddings @xmath56 , @xmath57 such that @xmath58 .
a * frass structure * is a countably infinite , locally finite ( finitely generated substructures are finite ) @xmath23-structure @xmath59 which is _ ultrahomogeneous _ , i.e. any isomorphism between finite substructures extends to an automorphism of @xmath59 .
an equivalent definition is that a frass structure satisfies the _ extension property _ :
for any @xmath37 , @xmath60 and any embedding @xmath61 , there is an embedding @xmath62 extending @xmath63 . for any infinite structure @xmath64 ,
let @xmath65 denote the class of finite substructures which embed into @xmath64 .
frass s theorem states that there is a one - to - one correspondence between frass classes and frass structures : the age of each frass structure is a frass class , and each frass class is the age of a frass structure unique up to isomorphism .
for @xmath44 such a class , we write @xmath66 for the associated structure , called the * frass limit * of @xmath44 .
see hodges [ ho ] for a more detailed exposition . for finite structures @xmath34 ,
let @xmath67 denote those substructures of @xmath35 which are isomorphic to @xmath26 .
a class @xmath44 of finite structures satisfies the _ ramsey property _ ( rp ) if for any @xmath68 and any @xmath69 , there is @xmath50 , @xmath70 such that for any coloring @xmath71 , there is @xmath72 for which @xmath73 is constant on @xmath74 . in [ kpt ]
, it is shown that for a frass structure @xmath59 , @xmath75 has the rp and consists of rigid structures ( having no non - trivial automorphisms ) if and only if @xmath76 consists of a single point ; such groups are said to be _
extremely amenable_. it should be noted that we regard @xmath77 as a topological group with the pointwise convergence topology ; a neighborhood basis at the identity is given by subgroups of the form @xmath78 for finite @xmath79 .
let @xmath44 and @xmath80 be frass classes with limits @xmath81 in languages @xmath23 and @xmath82 , with each @xmath83 a relation not in @xmath23 , such that @xmath84 .
we say the pair @xmath85 is _ reasonable _ if for any @xmath86 , @xmath47 , and embedding @xmath87 , there is @xmath88 with @xmath89 and @xmath90 also an embedding . for @xmath85
reasonable , we have @xmath91 . in this case ,
@xmath92 acts on the compact , metrizable space @xmath93 of all relations @xmath94 on @xmath59 with @xmath95 .
the basic neighborhoods of @xmath93 are given by open sets of the form @xmath96 for @xmath97 , @xmath48 , and @xmath98 .
the pair satisfies the _ expansion property _ if for each @xmath48 , there is @xmath47 such that for any @xmath99 with @xmath100 , @xmath89 , we have @xmath101 . if @xmath102 is a reasonable pair , then the expansion property is equivalent to @xmath93 being a minimal flow .
pairs of frass classes which are reasonable , satisfy the expansion property , and with @xmath80 satisfying the rp are called _
excellent_. in this case we call @xmath80 a * companion * of @xmath44 . in [ kpt ] , and later in the generality presented here in work of nguyen van th [ lnvt ] , it is shown that given an excellent pair , @xmath93 is the universal minimal flow of @xmath92 . let us show that for such @xmath59 , @xmath92 has the generic point property as witnessed by any point @xmath103 with @xmath104 .
notice that @xmath105 ; we need to show that this is @xmath106 ( since @xmath93 is minimal , we know that the orbit is dense ) .
so fix @xmath107 , with @xmath108 , @xmath89 and let @xmath61 be an embedding .
let @xmath109 denote those expansions @xmath110 for which @xmath63 embeds @xmath111 into @xmath112 and can be extended to an embedding @xmath113 .
@xmath109 is open and represents a single instance of the extension property .
recall that structures isomorphic to @xmath114 are exactly those which satisfy all instances of extension property ; as there are countably many triples @xmath115 , we see that @xmath116 is @xmath106 .
let @xmath85 be an excellent pair with limits @xmath81 . define @xmath117 to be the set of all finite substructures of @xmath59 .
compare this notion to @xmath118 ; the latter consists of all structures isomorphic to structures in @xmath117 . for each @xmath119 ,
let @xmath120 let @xmath121 , and let @xmath30 be an embedding .
for each @xmath122 , define : @xmath123 let @xmath124 .
though the elements of @xmath125 are structures , for neatness we will often denote elements of @xmath125 with variable names , such as @xmath126 , @xmath127 , etc .
form the vector space @xmath128 , the vector space over @xmath129 with basis @xmath125 .
let @xmath130 be the set of all elements of the form @xmath131 for some @xmath121 , @xmath132 , and @xmath133 an embedding .
let @xmath134 be the subspace generated by @xmath130 .
we are now able to state the first main theorem .
@xmath92 is amenable if and only if for all nonzero @xmath135 , @xmath136 , there are @xmath137 with @xmath138 , @xmath139 . it should be noted that only the forward direction will be needed in sections 4 and 5 .
@xmath140 suppose @xmath141 is an invariant measure on @xmath93 .
recall that the topology on @xmath93 is given by basic neighborhoods of the form @xmath142 for @xmath143 .
define @xmath144 by @xmath145 we see that for @xmath135 , we must have @xmath146 .
as @xmath93 is minimal , @xmath147 for all @xmath148 . if @xmath149 , @xmath150 and without loss of generality @xmath151 for all @xmath148 , then for some @xmath126 we have @xmath138 .
but then @xmath152 , a contradiction .
@xmath153 suppose the conditions of theorem 3.1 hold . as noted by kechris ( private communication ) , it is sufficient to show that for @xmath154 , there exists a consistent probability measure @xmath155 on @xmath156 , i.e. a probability measure such that for @xmath119 , @xmath157 , and any two embeddings @xmath158 , we have @xmath159 . indeed ,
if this is the case , let @xmath160 be finite substructures with @xmath161 , and for each @xmath162 let @xmath163 be a consistent probability measure on @xmath164 .
we will create an invariant measure on @xmath93 as follows : let @xmath119 and @xmath157 ; note that @xmath165 for large enough @xmath162 .
set @xmath166 , where @xmath167 is a non - principal ultrafilter on @xmath168 and @xmath169 is the inclusion embedding .
extend to @xmath93 by additivity . to see that this is invariant , let @xmath170 ; find @xmath154 and @xmath171 an embedding such that @xmath172 .
let @xmath162 be sufficiently large so that @xmath173 .
now we have : @xmath174 let @xmath175 consist of all elements of @xmath176 of the form 2 mm @xmath177 for some @xmath119 , @xmath157 , and embeddings @xmath178 .
consider the following system of inequalities and equalities in real variables @xmath179 , @xmath180 , where for @xmath181 , we let @xmath182 . -1
mm @xmath183 if this system has a solution , the solution is the consistent probability measure we seek , with @xmath184 .
note that the system ( 1 ) , ( 2 ) , ( 3 ) has a solution when the system ( 1 ) , ( 2 ) has a solution .
now we can use the rational form of stiemke s theorem ( [ s ] , see also border [ b ] ) , which states that for a rational - valued matrix @xmath39 , the equation @xmath185 has a rational solution with @xmath186 ( each entry @xmath187 has @xmath188 ) , or the equation @xmath189 has a rational solution ( each entry @xmath190 has @xmath191 , and at least one entry has @xmath192 ) . let @xmath39 be the matrix of coefficients from ( 1 ) . a solution to the system ( 1 ) , ( 2 )
is a vector @xmath193 whose entries are exactly the @xmath179 . hence , if there is no solution to @xmath194 , let @xmath195 be a solution to @xmath196 .
consider the element @xmath197 .
write @xmath198 .
stiemke s theorem tells us that @xmath151 for all @xmath180 and that for some @xmath126 we have @xmath138 .
this contradicts the assumptions of theorem 3.1 .
the directed graph @xmath0 has vertex set @xmath200 ; for vertices @xmath201 , @xmath202 , @xmath203 when there is @xmath204 with @xmath205 , i.e. @xmath206 is less than @xmath207 radians counterclockwise from @xmath208 .
we will write @xmath209 for those points lying in the interval counterclockwise from @xmath208 to @xmath206 .
let @xmath210 .
the appropriate companion class with the expansion and ramsey properties is @xmath211 , where @xmath212 , and the @xmath213 are unary relations with @xmath214 when @xmath208 is between @xmath215 and @xmath216 radians counterclockwise from the top of the circle ( see [ lnvt ] ) . for @xmath217
, we will write @xmath218 when @xmath214 holds . we will use theorem 3.1 to show : @xmath199 is not amenable .
let @xmath119 be a directed graph consisting of two vertices @xmath208 and @xmath206 with no edge between them .
let @xmath154 be a directed graph with vertices @xmath219 , @xmath126 , @xmath127 , and @xmath220 , with edges @xmath221 , @xmath222 , and @xmath223 .
consider @xmath156 ; we list all 12 expansions of @xmath35 below , where @xmath224 stands for @xmath225 with @xmath226 , @xmath227 , @xmath228 , @xmath229 . @xmath230 to arrive at the above , first observe that we can not have @xmath231 nor @xmath232 , as this would require @xmath233 or @xmath234 , respectively . also observe that mod @xmath235 , we can not have any of @xmath236 , @xmath237 , nor @xmath238 ; this is because for any @xmath239 , we have @xmath240 .
lastly , we may not have @xmath241 , as this would imply @xmath242 or @xmath243 . now without loss of generality fix @xmath244 .
there are exactly four possible expansions meeting the necessary conditions , and all four are easily realized .
the other eight expansions are then given by adding @xmath245 or @xmath246 mod 3 to each coordinate let @xmath157 denote the expansion with @xmath247 , @xmath248 .
let @xmath249 denote the embedding @xmath250 , @xmath251 .
let @xmath252 denote the embedding @xmath253 , @xmath254 .
now we have @xmath255 forming the spaces @xmath176 , @xmath134 as above , we have @xmath256 - [ ( 0 , 0 , 1 , 1 ) ] & \in v,\\ \rightarrow ( 0 , 1 , 1 , 2 ) + ( 0 , 0 , 1 , 2 ) & \in v.\end{aligned}\ ] ] hence @xmath199 is not amenable .
a boron tree is a graph - theoretic unrooted tree where each vertex has degree @xmath245 or @xmath235 .
it is possible to interpret boron trees as a model theoretic structure such that the class of finite boron trees forms a frass class ; we will closely follow the exposition of jasiski [ j ] . for a boron tree @xmath257 ,
let @xmath258 be the structure with universe @xmath259 , the leaves of @xmath257 , and @xmath260 a @xmath261-ary relation defined as follows : for leaves @xmath208 , @xmath206 , @xmath73 , @xmath262 , we have @xmath263 if @xmath208 , @xmath206 , @xmath73 , @xmath262 are distinct and there are paths from @xmath208 to @xmath206 and from @xmath73 to @xmath262 which do not intersect .
it is a fact that for finite boron trees @xmath257 and @xmath264 , @xmath265 if and only if @xmath266 .
let @xmath267 be the boron tree structure defined as follows : its universe is @xmath268 .
for any two leaves @xmath208 and @xmath206 , let @xmath269 .
define @xmath270 = \{a|_k:\ ; k\in [ \delta(a , b ) , n]\}\cup \{b|_k:\ ; k\in [ \delta(a , b ) , n]\}$ ] . for distinct @xmath208 ,
@xmath206 , @xmath73 , @xmath262 , set @xmath271 if @xmath272\cap [ c , d ] = \emptyset$ ] . the associated tree has vertices @xmath273 ; vertices @xmath274 and @xmath275 are adjacent if @xmath276 , and additionally the vertices @xmath277 with @xmath278 , @xmath279 are adjacent . on each @xmath267 ,
let @xmath280 be the lexicographic linear ordering of the leaves ; @xmath281 if @xmath282 or @xmath283 and @xmath284 or etc .
also , for any function @xmath285 , @xmath286 , let @xmath287 consist of those @xmath288 with @xmath289 .
let @xmath290 be the class of finite boron tree structures .
it is a fact that the set @xmath291 is cofinal in @xmath290 ; for each @xmath292 , there is an @xmath162 for which there is an embedding @xmath293 .
we will use this fact in defining the companion class @xmath294 such that @xmath295 is an excellent pair .
let @xmath292 , and let @xmath293 be an embedding for some @xmath162 .
define the structure @xmath296 as follows : @xmath130 is a @xmath235-ary relation , where for @xmath297 , we have @xmath298 if @xmath299 and @xmath300 .
we can rephrase this condition in a somewhat easier to grasp way : for @xmath301 , let @xmath302 be as follows : @xmath303 now we have @xmath298 exactly when @xmath304 .
note that @xmath298 holds whenever @xmath305 does .
let @xmath294 be the class of all such @xmath306 .
it is a fact that @xmath307 is an excellent pair ( see [ j ] ) , with limits @xmath308 .
we will show : @xmath309 is not amenable .
we begin with the following proposition .
let @xmath310 , with @xmath293 , and @xmath311 .
then there exists an embedding @xmath312 with @xmath313 .
suppose there exists @xmath314 such that for any @xmath315 , we have @xmath316 , or equivalently that for no @xmath315 do we have @xmath317 .
we will show that there is an embedding @xmath318 with @xmath319 . for @xmath288 , define the map @xmath320 as follows .
@xmath321 define @xmath322 by @xmath323 .
first let us show that @xmath324 is an embedding @xmath325 .
it will be useful to note that @xmath326 let @xmath327 be distinct , and suppose @xmath271 .
as @xmath10 is an embedding , we have @xmath328 \cap [ \pi(c),\pi(d)]= \emptyset$ ] .
suppose , without loss of generality and for sake of contradiction , that @xmath329 \cap [ \psi(c ) , \psi(d)]$ ] .
if @xmath330 , we have that @xmath331 \cap [ \pi(c),\pi(d)]$ ] , a contradiction .
if @xmath332 , then @xmath333 $ ] and @xmath334 $ ] .
therefore we must have @xmath335 .
however , this is only possible if @xmath336 . as @xmath337 ,
this is a contradiction .
therefore we have @xmath338 .
if @xmath339 , we may suppose @xmath340 \cap [ \pi(c),\pi(d)]$ ] . if @xmath330 , then @xmath329 \cap [ \psi(c ) , \psi(d)]$ ] . if @xmath341 , then @xmath342 \cap [ \psi(c ) , \psi(d)]$ ] .
hence we have @xmath343 .
now we will show that @xmath319 .
suppose @xmath344 , or equivalently @xmath345 .
observe that @xmath346 . now
if @xmath347 , we must have @xmath348 ; it follows that @xmath349 . let @xmath350 be least such that @xmath351 .
then we must have @xmath352 .
then @xmath353 is least such that @xmath354 , and @xmath355 .
hence @xmath356 .
now suppose @xmath356 .
let @xmath350 be least such that @xmath357 . then @xmath358 .
now if @xmath359 , then @xmath350 is least such that @xmath351 . if @xmath360 , then @xmath361 is least such that @xmath362 , and @xmath363 .
hence @xmath344 , and @xmath319 .
now suppose no such @xmath364 exists .
we will show that @xmath365 .
more precisely , we will show that for any @xmath350 and embedding @xmath366 , the map @xmath367 with @xmath368 has @xmath369 . for @xmath370
this is clear .
assume the result true for @xmath371 , and suppose @xmath372 .
fix @xmath373 , and let @xmath374 be such that @xmath375 is maximal .
let @xmath376 with @xmath377 and @xmath378 .
then @xmath379 we see that @xmath380 with this result , we can now compute @xmath381 for @xmath382 ; we will exhibit an element of @xmath383 showing that @xmath309 is not amenable .
let @xmath382 be a boron tree with three leaves @xmath384 .
we list the @xmath385 expansions in @xmath381 below ; we will denote an element of @xmath381 by @xmath386 $ ] , where we have @xmath387 for @xmath388 , @xmath389 and @xmath390 for @xmath391 . -1 mm @xmath392,[a < c < b;\ ; \emptyset],[b < a < c;\ ; \emptyset],\\ & [ b < c < a;\ ; \emptyset],[c < a < b;\ ; \emptyset],[c < b
< a;\ ; \emptyset],\\ & [ a < b < c;\ ; ( a , b , c)],[a < c < b;\ ; ( a , c , b)],[b < a < c;\ ; ( b , a , c)],\\ & [ b < c < a;\ ; ( b , c , a)],[c < a < b;\ ; ( c , a , b)],[c < b <
a;\ ; ( c , b , a)]\}.\end{aligned}\ ] ] now consider @xmath393 , where @xmath394 ; this structure has @xmath395 expansions .
these can be split into @xmath396 types , each of which has @xmath397 expansions . 1 .
type @xmath39 : expansions of the form @xmath398 with @xmath399 .
below , @xmath400 stands for the expansion @xmath401 $ ] .
@xmath402 2 .
type @xmath403 : expansions of the form @xmath398 with @xmath404 such that @xmath405 and @xmath406 .
below , @xmath400 stands for the expansion @xmath407 $ ] .
@xmath408 3 .
type @xmath409 : expansions of the form @xmath398 with @xmath404 such that @xmath405 and @xmath410 .
below , @xmath400 stands for the expansion @xmath411 $ ] .
@xmath412 4 .
type @xmath413 : expansions of the form @xmath398 with @xmath404 such that @xmath414 and @xmath415 .
below , @xmath400 stands for the expansion @xmath416 $ ] .
@xmath417 5 .
type @xmath418 : expansions of the form @xmath398 with @xmath404 such that @xmath414 and @xmath419 .
below , @xmath400 stands for the expansion @xmath420 $ ] .
@xmath421 now set @xmath422\in \mathcal{b}^*(\mathbf{a})$ ] .
let @xmath423 with @xmath424 , @xmath425 .
we see that @xmath426 and @xmath427 .
thus in @xmath176 , @xmath428 now let @xmath429 with @xmath430 , @xmath431 .
we have that @xmath432 , and @xmath433 . in @xmath176 , @xmath434 therefore @xmath309 is not amenable .
for the remainder of this paper , we shift our focus to the class @xmath44 , the class of finite dimensional vector spaces over a fixed finite field @xmath4 , which has companion @xmath80 , the class of such vector spaces equipped with a natural linear ordering , an ordering induced antilexicographically by some choice of ordered basis and a fixed ordering of @xmath4 with @xmath435 the least elements .
this is to say that if @xmath436 is the ordered basis we have chosen and @xmath437 , then @xmath438 if @xmath439 , where @xmath440 is the least index where @xmath441 . in particular , notice that there is a 1 - 1 correspondence between ordered bases and natural orderings . in what follows
, all vector spaces are assumed to be over @xmath4 . set @xmath442 , the countably infinite dimensional vector space over @xmath4 , and @xmath443 .
it is shown in [ akl ] that @xmath444 is uniquely ergodic .
the unique measure is just the uniform measure : for any finite @xmath445 , there are @xmath446 admissible expansions , each of which has measure @xmath447 .
another way to see this is to notice the following : if @xmath264 is finite dimensional , @xmath448 have the same dimension , and @xmath449 are natural linear orderings on @xmath450 , then there is an automorphism of @xmath264 sending @xmath451 to @xmath452 .
given any vector space @xmath134 with any ordering @xmath453 , say that @xmath454 has the _ finite lex property _ ( flp ) if for any finite dimensional subspace @xmath455 , we have that @xmath456 is a natural ordering .
in particular , a finite dimensional , ordered vector space has the flp iff that ordering is natural ( see thomas [ th ] ) , and @xmath457 has the flp iff @xmath458 .
given @xmath135 , @xmath459 , say that @xmath460 is _ minimal in its line _ if @xmath461 for all @xmath462 .
given @xmath463 , define the relation @xmath464 for all @xmath465 . also define @xmath466 . for @xmath467 ,
we need to tweak these definitions a bit .
we define @xmath468 to hold when @xmath469 .
we have @xmath470 if @xmath471 and @xmath472 .
note that if @xmath468 , then @xmath473 for all @xmath474 .
also , if @xmath475 , then @xmath476 and @xmath477 .
when @xmath454 is finite dimensional , @xmath478 has a simple characterization .
say @xmath453 is given by ordered basis @xmath479 . for @xmath480 ,
write @xmath481 , and suppose @xmath482 is least with @xmath483 nonzero . then for @xmath484
, we have @xmath468 exactly when @xmath482 is also least with @xmath485 nonzero . for @xmath454 with the flp
, @xmath478 is an equivalence relation .
suppose @xmath468 and @xmath486 .
form @xmath487 , and let @xmath488 be the basis of @xmath264 inducing @xmath456 , @xmath489 .
write @xmath490 , @xmath491 ; likewise for @xmath460 , @xmath219 .
observe that one of @xmath492 , @xmath493 , @xmath494 is nonzero ; as @xmath468 and @xmath486 , all three must be nonzero .
it follows that @xmath495 .
observe that @xmath496 , the set of nonzero equivalence classes , is also linearly ordered by @xmath497 for all @xmath373 , @xmath498 .
we will denote by @xmath499 the order type of this linear ordering
. it will be useful to introduce a standard notation for the @xmath162 nonzero equivalence classes of an @xmath162-dimensional , naturally ordered vector space @xmath454 .
call these classes @xmath500_{n-1 } < ... < [ v,<]_0 $ ] ; note that @xmath501_l$ ] exactly when @xmath482 is least with @xmath483 nonzero .
fix @xmath502 a @xmath245-dimensional subspace .
given @xmath503 , observe that for any @xmath504 we have @xmath468 .
call this equivalence class @xmath505 $ ] .
define @xmath506 < [ u ] \text { for at most } k \text { equivalence classes } [ u]\in \mathbf{v}_\infty /\!\sim\}.\ ] ] note that @xmath507 when @xmath508 .
let @xmath509 .
for any @xmath510 , @xmath511 is not a frass structure
. we will show that the extension property does not hold for @xmath511 . to see this ,
let @xmath512 , where @xmath264 is @xmath513-dimensional , @xmath514 on @xmath264 is given by an ordered basis @xmath515 , and @xmath516 .
now given @xmath503 , let @xmath517 with @xmath518 .
there are at most @xmath364 equivalence classes greater than @xmath505 $ ] in @xmath3 , but @xmath519 equivalence classes greater than @xmath520 = [ u,\prec]_{k+1}$ ] in @xmath264 .
it follows that there is no @xmath521 extending @xmath10 .
somewhat conversely , if there are at least @xmath519 equivalence classes above @xmath505 $ ] in @xmath511 , then @xmath10 as in the proof of 6.2 does admit an extension @xmath522 .
we can see this as follows : choose representatives @xmath523 , @xmath524 from @xmath519 equivalence classes @xmath525>\cdots > [ w_k]$ ] , where @xmath526>[v_1 , < ] $ ] .
let @xmath527 .
then @xmath528 is a @xmath513 dimensional space ; let @xmath529 be the basis which gives the ordering @xmath530 .
then we see that @xmath531 , and an extension @xmath522 is given by letting @xmath532 .
let @xmath533 be a finite dimensional subspace with @xmath534 , and let @xmath535 .
define @xmath536 to be those @xmath537 for which @xmath538 \geq [ w,<\!\!|_w]_k$ ] , i.e. those orderings for which there are at most @xmath364 equivalence classes of @xmath539 greater that @xmath540 $ ] ; if @xmath541 , set @xmath542 .
note that @xmath543 for @xmath544 .
hence @xmath545 where @xmath117 denotes the set of finite substructures of @xmath59 .
in particular , if @xmath546 where each @xmath547 is @xmath162-dimensional and @xmath548 , we have @xmath549 . whether or not @xmath550 depends only on @xmath551 ; each @xmath536 is a union of basic open sets of the form @xmath552 .
let us compute @xmath553 .
we have : + @xmath554 \geq [ w,<_w]_k)}{\#(<_w\text { on } w)}.\ ] ] + fix @xmath555 . for @xmath556
not all zero , let + @xmath557 + first let us show that for any two nonzero @xmath558 , @xmath559 , we have @xmath560 .
let @xmath482 be least with @xmath483 nonzero .
select @xmath561 arbitrarily such that they are linearly independent and do nt span @xmath562 .
there are @xmath563 ways to do this .
now set + @xmath564 + this ordered basis gives some @xmath565 , and certainly each @xmath566 can be uniquely produced in this manner .
hence @xmath567 .
now we have @xmath568 now @xmath569 . letting @xmath570 , we have @xmath571 .
this proves theorem 1.2 . + as there are countably many @xmath245-dimensional subspaces of @xmath3 , we obtain the following immediate corollary .
. then @xmath573 .
we will show that @xmath574 . clearly @xmath575 .
to show the other inclusion , note that if @xmath537 with @xmath576 , then there are @xmath577 , [ v_i]\in \mathbf{v}_\infty/\!\sim$ ] , @xmath578 , with @xmath579\neq[v_j]$ ] if @xmath580 and @xmath577<[v_i]$ ] for each @xmath581 .
pick @xmath582 $ ] . then @xmath583 .
in this section we develop some of the tools we will need to prove theorem 8.3 below . define a _
chain _ to be a sequence of subspaces @xmath584 with @xmath585 @xmath162-dimensional and @xmath586 . given a chain and an ordering @xmath537 ,
we will write @xmath587 . we will write @xmath588 for the least basis of @xmath589 in @xmath585 , i.e. the basis which induces the antilexicographic ordering .
let @xmath590 be a chain and @xmath537 . for @xmath591
, we may represent the inclusion map @xmath592 via the change of basis matrix @xmath593 .
writing @xmath594 , @xmath595 , @xmath596 , we have @xmath597 we will call @xmath593 as above the _ matrix of ordered inclusion _ ; we will use the shorthand @xmath598 when there is no confusion .
we see that @xmath599 .
this leads us to ask the following : for which @xmath600 matrices @xmath601 is there @xmath537 such that @xmath602 ? first observe that if @xmath537 , then the following must hold : 1 . for @xmath603 , we have @xmath604 .
2 . for any @xmath605 , we have @xmath606 .
let @xmath607 , @xmath608 be any ordered bases of @xmath585 , @xmath609 inducing orderings @xmath589 , @xmath610 .
let @xmath601 be the matrix of inclusion with respect to these bases .
if ( 1 ) and ( 2 ) hold , then @xmath610 extends @xmath589 . let @xmath611 be the least basis of @xmath612 .
consider the following subset of @xmath585 : @xmath613 we see that @xmath614 and @xmath615 .
however , we also see that for @xmath616 , @xmath617 .
hence @xmath618 , and @xmath619 .
it follows that @xmath620 , @xmath621 .
call an @xmath622 matrix @xmath601 _ valid _ if @xmath602 for some @xmath537 and some chain .
write @xmath623 , @xmath595 , @xmath596 . for each column @xmath624 of @xmath601 ,
let @xmath625 denote the least row number with @xmath626 nonzero .
if @xmath601 is valid , witnessed by @xmath537 and some chain , then the following must hold : * for @xmath627 , we have @xmath628 since @xmath629 . * for each @xmath364 , we have @xmath630 , as each @xmath631 is minimal in its line . * for @xmath632 , we have @xmath633 . for @xmath634 ,
this is clear . for @xmath635 , it is because we have @xmath636 for all @xmath462 .
the necessary conditions amount to saying that @xmath601 is the transpose of a matrix in reduced row echelon form with rank @xmath162 .
these conditions are also sufficient ; let @xmath601 satisfy the above , and let @xmath590 be a chain .
fix an ordered basis @xmath607 of @xmath585 , and choose an ordered basis @xmath608 such that @xmath601 is the matrix of inclusion with respect to these bases ; we may do this as @xmath601 has rank @xmath162 .
we will show that the two conditions of proposition 7.2 are satisfied .
the first is clear .
now , for @xmath605 , suppose @xmath637 with @xmath638 .
write @xmath639 , and suppose @xmath364 is least with @xmath640 ; we are done if we can show @xmath641 .
in the basis @xmath608 , we have : @xmath642 we see that @xmath643 is least with @xmath644 , so we must have @xmath645 .
suppose @xmath646 .
then @xmath647 . let @xmath648 and let @xmath624 be least with @xmath649 .
we have that @xmath650 .
then @xmath625 is least with @xmath651 . but
@xmath652 , from which it follows that @xmath653 , a contradiction .
the case @xmath654 will be of special interest , so let us introduce some terminology specific to this case . for @xmath601 a valid matrix , the map @xmath655 has range which excludes a single number @xmath364 , @xmath656 .
call such a matrix type @xmath364 . denote the type of matrix @xmath601 by @xmath657 . a natural question to ask
is how many valid @xmath658 matrices @xmath601 have @xmath659 . by the necessary conditions above
, we see that @xmath660 is determined except for those pairs @xmath661 with both @xmath662 and @xmath663 ; for these values of @xmath581 and @xmath624 , any choice of @xmath664 gives us a valid @xmath601 .
hence there are @xmath665 valid matrices @xmath601 of type @xmath364 .
observe that @xmath666_k\subset [ v_{n+1},<_{n+1}]_k$ ] if @xmath667 , and @xmath666_k\subset [ v_{n+1 } , < _ { n+1}]_{k+1}$ ] if @xmath668 .
in particular , @xmath669 iff @xmath670_k\cap v_n = \emptyset$ ] .
let @xmath671 be the set of valid @xmath658 matrices , which we will equip with the uniform probability measure @xmath672 .
let @xmath673 , and let @xmath674 be the product measure .
fix a chain @xmath590 .
there is a surjection @xmath675 as follows ; for any @xmath537 , let @xmath676 , where @xmath677 is the matrix of inclusion for the chain @xmath590 and the ordering @xmath453 .
let @xmath141 be the unique measure on @xmath93
. then @xmath678 .
first we will show the following : let @xmath589 be a natural order on @xmath585 , and let @xmath679 .
first let us show that the number of natural @xmath680 on @xmath681 extending @xmath589 such that the matrix of ordered inclusion is @xmath601 does not depend on @xmath601 or @xmath589 . write @xmath623 , @xmath682 , @xmath596 . pick @xmath162 linearly independent rows of @xmath601 ; without loss of generality assume rows @xmath621 are independent .
we now have the following procedure for producing @xmath680 extending @xmath589 : first pick @xmath683 from among the @xmath684 vectors in @xmath685 .
now set the other @xmath686 to be the unique solution to the system of equations @xmath687 this procedure can produce any natural @xmath680 extending @xmath589 with matrix of ordered inclusion @xmath601 .
moreover , we see that there are @xmath684 such extensions .
let @xmath688 be a basic open set in @xmath689 .
note that whether or not an ordering @xmath453 is in @xmath690 depends only on @xmath691 .
now we have @xmath692
to conclude this paper , we provide a more concrete representation of the measure @xmath141 .
first , we need a few general lemmas . let @xmath454 have the flp , and fix @xmath135 .
suppose @xmath693 are minimal in their lines with @xmath495 .
let @xmath694 be such that @xmath695 for @xmath696 , and likewise for @xmath219
. then @xmath697 .
form @xmath698 , and let @xmath456 be induced by basis @xmath699 , @xmath489 .
write @xmath490 , etc .
let @xmath364 be the least number with @xmath700 nonzero .
then @xmath364 is also the least number with @xmath701 nonzero .
we have @xmath702 ; we see that @xmath703 .
it now makes sense to define @xmath704}$ ] as in lemma 8.1 for @xmath135 and @xmath577\in v/\sim$ ] . for any @xmath705 and @xmath706
, we have @xmath707}+v_{[w ] } = ( u+v)_{[w]}$ ] and @xmath708 } = d(u_{[w]})$ ] .
let @xmath709 .
form @xmath698 , and let @xmath456 be induced by basis @xmath699 , @xmath489 .
write @xmath490 , etc .
let @xmath364 be the least number with @xmath701 nonzero .
then @xmath710 ; we see that @xmath707 } = u_k$ ] , @xmath711 } = v_k$ ] , @xmath712 } = ( u+v)_k = u_k + v_k$ ] , and @xmath708 } = du_k = d(u_{[w]})$ ] .
we may now injectively map @xmath454 to a subspace of @xmath713 , where @xmath514 is a partial ordering with @xmath714 iff for some @xmath577 $ ] in @xmath496 , we have @xmath715 ) < \beta([u])$ ] and @xmath716 ) \leq \beta([v])$ ] for all @xmath717 > [ u]$ ] .
we will be most interested in the case @xmath718 ; in this case @xmath514 is a linear order , as @xmath496 has no infinite ascending chains . for a fixed @xmath719 , enumerate @xmath720 by @xmath721 . now for any @xmath722 and @xmath723 , we may identify @xmath724 . to make this identification explicit ,
pick @xmath725 minimal in their lines , and define @xmath726 via @xmath727 , where @xmath728}$ ] .
fix any basis @xmath729 of @xmath3 .
define @xmath730 by setting @xmath731 .
note that @xmath732 is a borel map .
equip @xmath733 with the product measure @xmath734 .
the map @xmath732 is injective and a.e .
moreover , @xmath735 , giving a mod zero isomorphism of @xmath736 and @xmath737 . to see injectivity , it suffices to note that for any @xmath738 and any @xmath739 , the map @xmath740 is completely determined , which in turn determines @xmath453 . to show that @xmath732 is a.e .
surjective , consider @xmath741 .
certainly @xmath742 if the following hold : 1
. the @xmath743 are linearly independent .
2 . for each @xmath744
, there is an @xmath581 with @xmath745 , @xmath746 .
the second condition is easily seen to be the countable intersection of measure @xmath245 conditions .
for the first condition , observe that this is the countable intersection of conditions @xmath747 , each of which is measure 1 . to show that @xmath735 , it suffices to prove the next lemma .
for @xmath748 and @xmath749 , define : @xmath750 for @xmath751 linearly independent , @xmath752 .
fix @xmath722 and a chain @xmath590 , and let @xmath10 be as in proposition 5 .
first we consider the case where the @xmath753 are linearly independent .
the probability that @xmath754 for @xmath621 is bounded below by the probability that the following both occur for some @xmath482 with @xmath755 : 1 .
let @xmath280 on @xmath756 be given by basis @xmath757 .
. then @xmath759 .
2 . now suppose @xmath760 .
then we have @xmath761 for @xmath762 .
call the first event @xmath763 and the second event @xmath764 ; these events are independent . for
linearly independent @xmath765 , @xmath621 , the number of ordered bases @xmath766 with @xmath767 , @xmath621 , does not depend on which particular linearly independent @xmath765 are being considered . therefore : @xmath768 now for @xmath762 , let @xmath769 be the event that @xmath761 .
we have @xmath770 .
the events @xmath769 are mutually independent , hence : @xmath771 now we have : @xmath772 it follows that @xmath773 .
equality follows since for @xmath774 , we have @xmath775 .
when the @xmath753 are not linearly independent , let : @xmath776 we have the lower bound : @xmath777 we have @xmath778 , giving us : @xmath779 for any @xmath350 . letting @xmath780 , we see that @xmath773 , and hence @xmath781 .
our investigations above lend themselves to a number of open questions : assume the frass class @xmath44 admits a companion @xmath80 .
let @xmath59 be the frass limit of @xmath44 . if @xmath92 is uniquely ergodic , what are necessary and sufficient conditions for the unique measure on any minimal flow to be supported on the generic orbit ?
pongrcz in [ p ] has given a partial answer to question 8.2 .
let @xmath782 , for @xmath453 a symbol for a linear ordering and @xmath23 * relational*. let @xmath85 be an excellent pair of frass classes in @xmath23 and @xmath783 .
suppose @xmath80 is order forgetful , i.e. for @xmath784 , we have @xmath785 .
we see that @xmath92 , if amenable , is uniquely ergodic , and the measure satisfies @xmath786 ( see the introduction ) .
pongrcz has shown that in this case , @xmath141 is supported generically .
note that this does not contradict theorem 1.2 ; every hypothesis of pongrcz s theorem is satisfied except that the language of vector spaces contains function symbols .
his calculations also shed some light on the role that functions play in my calculations for @xmath3 , and they also suggest that we may be able to find relational examples with @xmath787 with the measure @xmath141 as above not supported generically . to make question 9.3 more precise , consider what was shown in sections 7 and 8 .
though no single orbit has positive measure , it seems that by taking a suitable completion of @xmath3 , the unique measure concentrates on the isomorphism type of @xmath788 ordered lexicographically . in what sense
can this be made precise and generalized to other structures ?
j. auslander , _ minimal flows and their extensions _ , north holland , 1988 .
o. angel , a.s .
kechris , and r. lyons , random orderings and unique ergodicity of automorphism groups , _
_ , to appear .
kc border , alternative linear inequalities , updated 2012.http://www.hss.caltech.edu/ kcb / notes / alternative.pdf wilfrid hodges , _ model theory _ , cambridge univ .
press , 1993 .
j. jasiski , hrushovski and ramsey properties of classes of finite inner product structures , finite euclidean metric spaces , and boron trees , _ ph.d .
thesis _ , univ . of toronto ( 2011 ) .
kechris , v.g .
pestov , and s. todorcevic , frass limits , ramsey theory , and topological dynamics of automorphism groups , _ geometric and functional analysis _ , * 15 * ( 2005 ) , 106189 .
kechris , m. soki , dynamical properties of the automorphism groups of the random poset and random distributive lattice , _ fund .
_ , * 218 * , ( 2012 ) , 6994 . l. nguyen van th , more on the kechris - pestov - todorcevic correspondence : precompact expansions , _ fund . math .
_ , to appear .
andrs pongrcz , private communication .
e. stiemke , ber positive lsungen homogener linearer gleichungen , math .
* 76 * ( 1915 ) , 340342 .
s. thomas , groups acting on infinite dimensional projective spaces , _ j. london math . soc .
_ , * 34 ( 2 ) * , ( 1986 ) , 265273 . | in this paper we consider those frass classes which admit companion classes in the sense of [ kpt ] .
we find a necessary and sufficient condition for the automorphism group of the frass limit to be amenable and apply it to prove the non - amenability of the automorphism groups of the directed graph @xmath0 and the boron tree structure @xmath1 .
also , we provide a negative answer to the unique ergodicity - generic point problem of angel - kechris - lyons [ akl ] . by considering @xmath2 , where @xmath3 is the countably infinite dimensional vector space over a finite field @xmath4
, we show that the unique invariant measure on the universal minimal flow of @xmath2 is not supported on the generic orbit . |
the standard model ( sm ) explains current high energy experiments , but it offers no understanding of many parameters included in the sm ( e.g. the fermion masses and the flavor mixing angles ) . it is expected that there exist some fundamental theory which includes the sm as a low energy effective theory . this theory should offer a deeper understanding of these parameters .
if some reasonable relations among the sm parameters could be discovered phenomenologically , it would provide us with important clue toward the search for the fundamental theory .
recently the top quark has been discovered and its mass has been measured .
more precise data on elements of the kobayashi - maskawa ( km ) matrix @xcite will be available in the near future , e.g. from the experiments at the @xmath0-factories . thus it is timely to study the quark mass matrices and the km matrix through various phenomenological approaches . up to now , many attempt have been made to investigate the explicit connection between the quark mixing matrix and quark mass matrices @xcite-@xcite .
here we present exact relations between the general mass matrices and the flavor mixing matrix .
this approach is obviously advantageous over many specific @xmath1 and can model - independently shed light on the origin of quark masses and flavor mixing . in this work ,
we start from the nearest - neighbor interactions ( nni ) form of quark mass matrices .
have shown that any @xmath2 quark mass matrices can be transformed into the nni basis form both up and down part at the same time , so we do not lose any generality by using this basis @xcite .
the up and down mass matrices have twelve parameters , we fix six of them so that the six quark masses can be correctly reproduced .
further more we show that two real parameters can be chosen to be _ any _ value .
the remaining four parameters are obtained from the measurements of the km matrix .
this paper is organized as follows . in section
[ sec : nni ] , we review the nni form of the mass matrices . in section [ sec : free ] , we separate out the two arbitrary degrees of freedom of the nni basis .
these two degrees of freedom can be chosen implicitly and will not change any observable . in section [ sec : epsf ] , we show how to separate the elements of quark mass matrices in terms of quark masses and dimensionless parameters on the nni basis . in section [ sec :
exact_km ] , exact analytical calculation of the km matrix is presented . in section [ sec : sum ] , we specialize our result to the fritzsch @xmath1 and summarize our results .
_ nni basis _ it has been shown that the arbitrary up and down part @xmath2 quark mass matrices can be simultaneously transformed into the nni form without changing any observable quantities @xcite .
that is , this transformation itself does not change the values of quark masses and the km matrix elements .
hence the nni form includes all physical contents of quark mass matrices in the sm . in this basis
each mass matrix has four texture zeros , so it is very simple .
generally , @xmath2 mass matrices in the nni basis can be written as follows : @xmath3 where @xmath4 and @xmath5 are the top and bottom quark masses , respectively .
note that in the nni form , @xmath6 and @xmath7 are chosen real and non - negative and the phases are moved to a diagonal phase matrix @xmath8 shown below . @xmath9 and @xmath10 are diagonalized by orthogonal matrices @xmath11 and @xmath12 through @xmath13 where @xmath14 , @xmath15 , @xmath16 and @xmath17 are the up , charm , down and strange quark masses , respectively .
we can write the km matrix @xmath18 by reparametrizing the right handed quark field phases , @xmath19 where @xmath20 is the phase - difference matrix between the up and down quark sectors .
_ degrees of freedom _ there are twelve real parameters in the above nni form , although the number of observable parameters is ten , that is , the six quark masses and four measurements of the km matrix . remaining two degrees of freedom are those of the choice of the nni basis .
any mass matrices @xmath21 and @xmath22 can be transformed into the above nni basis mass matrices @xmath23 and @xmath24 without changing the values of the quark masses and the km matrix through @xmath25 where @xmath26 , @xmath27 and @xmath28 are unitary matrices .
( here , the matrix @xmath8 is displayed in ( [ eqn : phase ] ) . ) in order to construct @xmath28 , we choose some non - vanishing complex value @xmath29 at first .
then , using @xmath29 , we obtain @xmath30 @xmath31 as the eigenvector of the matrix @xmath32 as follows : @xmath33 where @xmath34 is the eigenvalue of the matrix @xmath32 and @xmath35 .
so @xmath29 is arbitrary degrees of freedom of the nni basis . using equations ( [ eqn : defn ] ) and ( [ eqn : mu+md ] ) ,
we can write @xmath29 and @xmath34 in terms of the elements of the nni basis mass matrices @xmath23 and @xmath24 as follows : @xmath36
therefore we can fix two arbitrary degrees of freedom of the nni basis by fixing the complex number @xmath29 .
obviously from the above calculations , this fixing does not require any constraint to the quark masses and the km matrix . in another way
, we can construct @xmath28 by starting from @xmath37 @xmath31 as follows .
@xmath37 is given as the eigenvector of the matrix @xmath38 .
here , @xmath39 is some arbitrary complex and non - vanishing value .
then , instead of ( [ eqn : mu+md ] ) , we use following equation : @xmath40 where @xmath41 is the eigenvalue of the matrix @xmath38 and @xmath35 . in this way
, @xmath42 is also arbitrary degrees of freedom . using equations ( [ eqn : defn ] ) and ( [ eqn : mu+md2 ] ) , we obtain the expressions for @xmath39 and @xmath41 as follows : @xmath43 also , fixing the value of @xmath39 , instead of fixing the value of @xmath29 , does not require any constraint to the km matrix and the quark masses . if we fix the arbitrary two degrees of freedom of basis , @xmath29 or @xmath39 , then ten degrees of freedom remain in the nni basis mass matrices .
in this section we present the parametrization of the quark mass matrices on the nni basis in terms of six eigenvalues and six dimensionless parameters without any approximation .
+ + _ parameter separating _
the characteristic equation of the matrix @xmath44 is @xmath45 where @xmath46 , @xmath47 , @xmath48 , @xmath49 and @xmath50 , and @xmath51 is an eigenvalue of @xmath44 . because the textures of up and down mass matrices are the same , we discuss them in a common manner by omitting the indexes @xmath52 and @xmath53 .
we know that @xmath51 in the equation ( [ eqn : char ] ) has three solutions : @xmath54 where @xmath55 are the quark masses of the @xmath56-th generation .
obviously @xmath51 are dimensionless , because @xmath57 and @xmath58 have been normalized by the third generation quark masses .
therefore , @xmath59 we introduce two real and positive mass parameters as follows : @xmath60 note that the equation ( [ rqn : ace ] ) is satisfied by introducing two real and positive dimensionless parameters , @xmath61 and @xmath62 : @xmath63 although @xmath61 and @xmath62 are independent of the quark masses , they enter the km matrix .
furthermore , equation ( [ eqn : a+b+ ] ) and equation ( [ eqn : ab+ac+ ] ) can be written as @xmath64 therefore , @xmath65 and @xmath66 are solutions of the following quadratic equation of @xmath67 : @xmath68 solving equation ( [ eqn : eqn_zeta ] ) analytically , we obtain @xmath69 as a consequence , we can write general quark mass matrices in the following simple form without any approximation : @xmath70 note that two possibilities of signs are included in equation ( [ eqn : defbd ] ) , so let us call the case that the plus - minus sign in @xmath0 ( @xmath71 ) is plus ( minus ) case ( i ) and the case that the plus - minus sign in @xmath0 ( @xmath71 ) is minus ( plus ) case ( ii ) .
+ + _ mathematically allowed regions for @xmath61 and @xmath62 _ next , we discuss mathematically allowed regions for @xmath61 and @xmath62 .
each element of the mass matrices can be taken to be non - negative values through redefinition of the quark field phases , as discussed at the beginning .
equation ( [ eqn : deface ] ) means that @xmath72 , @xmath73 and @xmath74 .
that is , @xmath75 .
about @xmath0 and @xmath71 , they must satisfy the constraints @xmath76 and @xmath77 .
this implies that values in the square roots in equation ( [ eqn : defbd ] ) are non - negative .
the mathematically allowed regions for @xmath61 and @xmath62 with the non - negativity constraints are illustrated in fig.1(a),(b ) and fig.2(a),(b ) .
we point out that @xmath0 and @xmath71 do not vanish at the same time due to the following restrictions : the cp - violating phase should exist and no element of the km matrix should be zero .
we can write down the km matrix without any approximation by using above form .
+ + _ orthogonal matrices _ when both @xmath0 and @xmath71 are not zero , eigenvectors of @xmath78@xmath79 can be written in terms of eigenvalues as follows : @xmath80 where @xmath81 are eigenvalues of @xmath44 given in ( [ eqn : eva ] ) and @xmath82 are normalization factors of the eigenvectors . the orthogonal matrix in equations ( [ eqn : ommo ] ) can be written in terms of the above eigenvectors as follows : @xmath83 + + _ normalization factors _ normalization factors @xmath84 can be written as follows : @xmath85 ( [ eqn : nf1 ] ) is the fourth - order in @xmath81 .
using the equation ( [ eqn : char ] ) , ( [ eqn : nf1 ] ) becomes the second - order in @xmath81 .
we introduce new variables @xmath86 , @xmath8 and @xmath87 which can be written in terms of quark masses as follows : @xmath88 normalization factors ( [ eqn : nf1 ] ) can be written by using @xmath89 , @xmath90 and above variables @xmath86 , @xmath8 and @xmath87 @xmath91 where @xmath92 .
+ + _ elements of the km matrix _
the elements of the km matrix become as follows : @xmath93 .
\label{eqn : dkm2}\end{aligned}\ ] ] the expressions of elements of the km matrix ( [ eqn : dkm2 ] ) is written by eigenvalues of the mass matrices , the parameters @xmath94 , @xmath95 , @xmath96 , and @xmath97 included in up and down part mass matrices , and two phases .
indexes @xmath52 and @xmath53 show the up and down part , respectively , in equation ( [ eqn : dkm2 ] ) .
it is written in terms of only the eigenvalues of the mass matrices and the other six dimensionless parameters .
+ + _ more texture zeros case _
equation ( [ eqn : dkm2 ] ) is satisfied only when elements @xmath98 , @xmath99 , @xmath100 , @xmath53 and @xmath101 of up and down part mass matrices are positive definite . though @xmath102 , @xmath103 and @xmath104 are not zero , there are cases that @xmath105 or @xmath106 @xmath107
. however , two values of @xmath108 , @xmath109 , @xmath110 and @xmath111 do not become zero at the same time by physical conditions that cp - violating phase should exist and that no element of the km matrix should be zero .
therefore , we restrict our discussion to the case where one of @xmath112 , @xmath113 , @xmath114 and @xmath115 is zero . at first , let us discuss about the case of @xmath116 . in this case ,
up and down part mass matrices are @xmath117 except phases .
one of the eigenvalues of @xmath57 @xmath118 is @xmath119 .
the eigenvector associated with the eigenvalue @xmath119 is @xmath120 the other two eigenvectors related to eigenvalues @xmath121 are @xmath122 where @xmath123
are normalization factors : @xmath124 here , we have used the characteristic equation of @xmath57 @xmath118 .
therefore , the km matrix can be written as follows : @xmath125 where @xmath126 , @xmath127 and @xmath128 are displayed in ( [ eqn : eve ] ) . in the same way ,
we can write down the km matrix in the case of @xmath129 , @xmath130 and @xmath131 as follows . + + case of @xmath129 : @xmath132 & ( i \neq k ) , \end{array } \right .
\nonumber \\ \left(f_{u}^{d}\right)_{i}^{2 } & = & ( c_u+b_u - e_u){\xi_u}_i - e_u(c_u - b_u - e_u ) .
\label{eqn : kmud}\end{aligned}\ ] ] + + case of @xmath130 : @xmath133 + + case of @xmath131 : @xmath134 & ( j \neq k ) , \end{array } \right .
\nonumber \\ \left(f_{d}^{d}\right)_{j}^{2 } & = & ( c_d+b_d - e_d){\xi_d}_j - e_d(c_d - b_d - e_d ) .
\label{eqn : kmdd}\end{aligned}\ ] ]
the fritzsch - type mass matrices @xcite are a special case of the nni form
. they can be reproduced by taking @xmath135 and @xmath136 in the above form .
explicitly , if the conditions @xmath137 are satisfied , then @xmath138 and @xmath139 become the fritzsch - type mass matrices .
+ + _ summary _ starting from general mass matrices in the nni form , we presented a parametrization which guarantees the six eigenvalues ( quark masses ) .
it is written in terms of six parameters .
we show that two of these parameters can be chosen implicitly .
these two parameters change neither the eigenvalues nor the km matrix elements .
we then presented an analytic form of the km matrix elements in terms of these parameters .
our formulation is very useful for generating a class of @xmath1 for the km matrix elements .
+ + 99 cdf collaboration , f. abe _
et al . _ ,
* 73 * ( 1994 ) 225 .
d0 collaboration , s. abachi _
et al . _ ,
* 74 * ( 1995 ) 2632 .
m. kobayashi and t. maskawa , prog .
* 49 * ( 1973 ) 652 .
h. fritzsch , phys . lett . * 73b * ( 1978 ) 317 ; nucl . phys .
* b115 * ( 1979 ) 189 .
f. wilczek and a. zee , phys .
* 42 * ( 1979 ) 421 .
h. georgi and c. jarlskog , phys .
* 86b * ( 1979 ) 297 .
b. stech , phys . lett . *
130b * ( 1983 ) 189 .
y. koide , phys . rev .
* d46 * ( 1992 ) 2121 .
m. leurer , y. nir and n. seiberg , nucl . phys . * b398 * ( 1993 ) 319 .
l. ibanez and g. g. ross , phys . lett .
* b332 * ( 1994 ) 100 .
g. c. branco and j. i. silva - marcos , phys .
331b * ( 1994 ) 390 .
m. carena , s. dimopoulos , c. e. m. wagner and s. raby , phys .
* d52 * ( 1995 ) 4133 .
g. c. branco , l. lavoiura and f. mota , phys . rev .
* d39 * ( 1989 ) 3443 .
t. ito and m. tanimoto , dpnu-96 - 07 , hep - ph/9603393 .
1(a ) : mathematically allowed region for @xmath95 and @xmath94 for the case ( i ) . we use following values of the quark mass ratios : @xmath140 + fig .
1(b ) : mathematically allowed region for @xmath95 and @xmath94 for the case ( ii ) .
we use the same values of the quark mass ratios as those in fig .
+ + fig .
2(a ) : mathematically allowed region for @xmath97 and @xmath96 for the case ( i ) .
we use following values of the quark mass ratios : @xmath141 + fig .
2(b ) : mathematically allowed region for @xmath97 and @xmath96 for the case ( ii ) .
we use the same values of the quark mass ratios as those in fig . | we analyze properties of general quark mass matrices .
the up and down part quark mass matrices are written in terms of six dimensionless parameters and six quark masses .
it is shown that two of the former six dimensionless parameters can be chosen to be any value .
once values for these two parameters are chosen , kobayashi - maskawa matrix is written in terms of the remaining four parameters .
our results are given analytically without any approximation .
= 0.0 in = 24.8 cm = 15.8 cm = 3ex = 6ex = 0.0 cm = 0.0 cm = 0.0 cm = 0.0 cm = 1.3em dpnu-96 - 24 +
hep - ph/9605215 + may 1996 * pacs * : 12.15.ff , 12.15.hh + keywords : km matrix , mass matrix , flavor mixing |
the stochastic block model is one of the oldest and most ubiquitous models for studying clustering and community detection .
it was first introduced by holland et al .
@xcite more than thirty years ago and since then it has received considerable attention within statistics , computer science , statistical physics and information theory .
the model defines a procedure to generate a random graph : first , each of the @xmath0 nodes is independently assigned to one of @xmath1 communities , where @xmath2 is the probability of being assigned to community @xmath3 .
next , edges are sampled independently based on the community assignments : if nodes @xmath4 and @xmath5 belong to communities @xmath3 and @xmath6 respectively , the edge @xmath7 occurs with probability @xmath8 independent of all other edges , where @xmath9 is an @xmath10 symmetric matrix .
the goal is to recover the community structure either exactly or approximately from the graph . since its introduction
, the stochastic block model has served as a testbed for the diverse range of algorithms that have been developed for clustering and community detection , including combinatorial methods @xcite , spectral methods @xcite , mcmc @xcite , semidefinite programs @xcite and belief propagation @xcite .
recently , the stochastic block model has been thrust back into the spotlight , the goal being to establish tight thresholds for when community detection is possible , and to find algorithms that achieve them . towards this end ,
decelle et al .
@xcite made some fascinating conjectures ( which have since been resolved ) in the case of two equal - sized communities with constant average degree , that we describe next . throughout this paper , we will focus on the two - community case and use @xmath11 and @xmath12 to denote the within - community and between - community connection probabilities respectively .
we will be interested in the sparse setting where @xmath13 .
moreover we set @xmath14 , so that the two communities are roughly equal - sized , and we assume @xmath15 ( although we discuss the case @xmath16 in appendix [ sec : dissortative ] ) .
we will use @xmath17 to denote the corresponding random graph model .
the setting of parameters above is particularly well - motivated in practice , where a wide variety of networks have been observed @xcite to have average degree that is bounded by a small constant .
it is important to point out that when the average degree is constant , it is impossible to recover the community structure exactly in the stochastic block model because a constant fraction of nodes will be isolated .
instead , our goal is to recover a partition of the nodes into two communities that has non - trivial agreement ( better than random guessing ) with the true communities as @xmath18 , and we refer to this as _ partial recovery_. @xcite if @xmath19 then partial recovery in @xmath17 is possible , and if @xmath20 then partial recovery is information - theoretically impossible .
this conjecture was based on deep but non - rigorous ideas originating from statistical physics , and was first derived heuristically as a stability criterion for belief propagation .
this threshold also bears a close connection to known thresholds in the broadcast tree model , which is another setting in which to study partial recovery .
we formally define the broadcast tree model in section [ sec : prelim ] , but it is a stochastic process described by two parameters @xmath21 and @xmath22 .
kesten and stigum @xcite showed that partial recovery is possible if @xmath19 , and much later evans et al .
@xcite showed that it is impossible if @xmath23 .
the connection between the two models is that in the stochastic block model , the local neighborhood of a node resembles the broadcast tree model , and this was another compelling motivation for the conjecture . in an exciting sequence of developments , mossel , neeman and
sly @xcite proved a lower bound that even distinguishing @xmath17 from an erds rnyi graph @xmath24 is information - theoretically impossible if @xmath23 , by a careful coupling to the broadcast tree model .
subsequently mossel , neeman and sly @xcite and massouli @xcite independently gave matching algorithms that achieve partial recovery up to this threshold , thus resolving the conjecture !
mossel , neeman and sly @xcite later showed that for some constant @xmath25 , if @xmath26 then belief propagation works and moreover the agreement of the clustering it finds is the best possible .
in fact , many other sorts of threshold phenomena have been found in different parameter regimes .
abbe , bandeira and hall @xcite studied exact recovery in the logarithmic degree setting @xmath27 and showed that it is efficiently possible to recover the two communities exactly if @xmath28 and information - theoretically impossible if @xmath29 .
abbe and sandon @xcite gave a precise characterization of when exact recovery is possible in the general stochastic block model for more than two communities with arbitrary relative sizes and arbitrary connection probabilities .
this abundance of sharp thresholds begs a natural question : how robust are these reconstruction thresholds ?
it is clear that if one substantially changes the distributional model , the thresholds themselves are likely to change .
however there is a subtler issue .
the algorithms that achieve these thresholds may in principle be over - fitting to a particular distributional model .
random graphs are well - known @xcite to have rigid properties , such as sharp laws for the distribution of subgraph counts and a predictable distribution of eigenvalues .
real - world graphs do not have such properties . in a remarkable paper , blum and spencer
@xcite introduced the semirandom model as an intermediary between average - case and worst - case analysis , to address such issues .
the details of the model vary depending on the particular optimization problem , and since we will focus on clustering we will be most interested in the variant used by feige and kilian @xcite .
the adversary above is called ` monotone ' because it is restricted to making changes that seem to be helpful .
it can only strengthen ties within each community , and break ties between them .. if we had @xmath30 we could define the adversary in the opposite way ( which we analyze in appendix [ sec : dissortative ] ) . ]
the key is that a monotone adversary can break the sorts of rigid structures that arise in random graphs , such as predictable degree distributions and subgraph counts .
an algorithm that works in a semirandom model can no longer rely on these properties . instead of a graph containing
_ only _ random edges , all we are assured of is that it contains _ some _ random edges . in this paper
, we use semirandom models as our notion of ` robustness . '
many forms of robustness exist in the literature for example , the independent work @xcite gives algorithms that are robust to @xmath31 non - monotone changes in addition to monotone changes . with most robustness models , any algorithm will break down after enough errors , and one can compare algorithms based on how many errors they can tolerate . in contrast , semirandom models distinguish between algorithms _
qualitatively _ : as we will see , entire classes of algorithms continue to work under any number of monotone changes , while others do not .
feige and kilian @xcite showed that semidefinite programs for _ exact recovery _ in the stochastic block model continue to work in the semirandom model in fact they succeed up to the threshold for the random model @xcite .
since then , there have been many further developments including algorithms that work in semirandom models for planted clique @xcite , unique games @xcite , various graph partitioning problems @xcite , correlation clustering @xcite and the general planted partition model for all @xmath3 and @xmath32 for all @xmath33 , with @xmath15 , but the number of communities and their relative sizes are arbitrary constants . ] @xcite ( and in some cases for even more powerful adversaries ) .
a common theme is that if you have a semidefinite program that works in some stochastic setting , then it often extends almost automatically to the semirandom setting .
so are there semidefinite programs that achieve the same sharp partial recovery results as mossel , neeman and sly @xcite and massouli @xcite , and that extend to the semirandom model too ? or is there a genuine gap between what is achievable in the random vs. the semirandom setting ? recall that in the semirandom block model , a monotone adversary observes a sample from the stochastic block model and is then allowed to add edges within each community and delete edges that cross between communities .
we will design a particularly simple adversary to prevent an algorithm from utilizing the paths of length two that go from a ` @xmath34 ' node to a ` @xmath35 ' node and back to a ` @xmath34 ' node , where the middle node has degree two .
our adversary will delete any such path it finds ( with some additional technical conditions to locally coordinate these modifications ) , and our main result is that , surprisingly , this simple modification strictly changes the partial recovery threshold .
we will state our bounds in terms of the average degree @xmath36 and ` noise ' @xmath37 , in which case the threshold @xmath38 becomes @xmath39 .
note that this threshold requires @xmath40 .
then : [ thm : intro - main ] for any @xmath40 , there exists @xmath41 so that @xmath42 and hence partial recovery in the stochastic block model @xmath17 is possible , and yet there is a monotone adversary so that partial recovery in the semirandom model is information - theoretically impossible .
a common tool in the algorithms of mossel , neeman and sly @xcite and of massouli @xcite is the use of non - backtracking walks and spectral bounds for their transition matrices .
our adversary explicitly deletes a significant number of these walks .
this simple modification not only defeats these particular algorithms , but we can show that _ no _ algorithm can achieve partial recovery up to the threshold @xmath43 . to the best of our knowledge ,
this is the first explicit separation between what is possible in the random model vs. the semirandom model , for _ any _ problem with a monotone adversary .
we show a complementary result , that no monotone adversary can make the problem too much harder .
various semidefinite programs have been designed for partial recovery .
these algorithms work in the constant - degree regime , but are not known to work all the way down to the information - theoretic threshold . in particular , we follow gudon and vershynin @xcite and show that their analysis ( with a simple modification ) works as is for the semirandom model .
this shows that semidefinite programs not only give rise to algorithms that work for the semirandom model in the exact recovery setting , but also for partial recovery too under some fairly general conditions .
[ thm : intro - robust - sdp ] let @xmath36 and @xmath44 .
there is a constant @xmath45 so that if @xmath46 and @xmath47 then partial recovery is possible in the semirandom block model .
moreover it can be solved in polynomial time .
our robustness proof only applies to a particular form of sdp analysis .
given a different proof that the sdp succeeds in the random model for a larger range of parameters @xmath48 than above , there would be no guarantee that the sdp also succeeds in the semirandom model for that range of parameters .
hence we can not formally conclude that it is impossible for the sdp to reach the information - theoretic threshold in the random model , though our results are suggestive in this direction .
we remark that each possible monotone adversary yields a new distribution on planted community detection problems .
hence , we can think of the algorithm in the theorem above as one that performs almost as well as information - theoretically possible across an entire family of distributions , simultaneously .
this is a major advantage to algorithms based on semidefinite programming , and points to an interesting new direction in statistics .
can we move beyond average - case analysis ?
can we find robust , semirandom analogues to some of the classical , average - case thresholds in statistics ?
the above two theorems establish upper and lower bounds for this semirandom threshold , that show that it is genuinely different from the average - case threshold .
the usual notion of a threshold makes sense when we have exact knowledge of the process generating the instances we wish to solve .
but when we lack this knowledge , semirandom models offer an avenue for exploration that can lead to new algorithmic and statistical questions . along the way to proving our main theorem
, we show a random vs. semirandom separation for the broadcast tree model too .
we define this model in section [ sec : prelim ] . in short ,
it is a stochastic process in which each node is given one of two labels and gives birth to @xmath49 nodes of the same label and @xmath50 nodes of the different label , with the goal being to guess the label of the root given the labels of the leaves .
there are many ways we could define a monotone adversary , and we focus on a particularly weak one that is only allowed to cut edges between nodes with different labels .
we call this the cutting adversary , and we prove : [ thm : intro - treehard ] for any @xmath40 , there exists @xmath41 so that @xmath42 and hence partial recovery in the broadcast tree model is possible , and yet there is a monotone cutting adversary for which partial recovery in the semirandom broadcast tree model is information - theoretically impossible .
furthermore we analyze the recursive majority algorithm and show that it is robust to an even more powerful class of monotone adversaries , which are allowed to entirely control the subtree at a node whose label is different than its parent .
we call this a strong adversary , and we prove : [ thm : intro - recmaj ] if @xmath51 and @xmath52 then partial recovery in the broadcast tree model is possible , even with respect to a strong monotone adversary , where ` @xmath53 ' is taken as @xmath54 .
these results highlight another well - studied model where the introduction of a monotone adversary strictly changes what is possible .
nevertheless there is an algorithm that succeeds across the entire range of distributions that arise from the action of a monotone adversary , simultaneously .
interestingly , our robustness results can also be seen as a justification for why practitioners use recursive majority at all .
it has been known for some time that recursive majority does not achieve the kesten stigum bound @xcite the threshold for reconstruction in the broadcast tree model although taking the majority vote of the leaves does .
the advantage of recursive majority is that it is robust to very powerful adversaries while majority is not , and this only becomes clear when studying these algorithms through semirandom models !
here we formally define the models we will be interested in , as well as the notion of partial recovery .
recall that @xmath55 denotes the stochastic block model on two communities with @xmath14 so that the communities are roughly equal sized .
we will encode community membership as a _ label _ @xmath56 on each node @xmath5 .
we will also refer to this as a _ spin _
, following the convention in statistical physics .
this numeric representation has the advantage that we can ` add ' spins in order to take the majority vote , and ` multiply ' them to compute the relative spin between a pair of nodes
. we will be interested in the sparse regime @xmath13 where the graph has constant average degree , and we will assume @xmath15 .
next , we formally define partial recovery in the stochastic block model . throughout this paper will will be interested in how our algorithms perform as @xmath0 ( number of nodes ) goes to @xmath57 .
we say that an event holds a.a.s .
( asymptotically almost surely ) if it holds with probability @xmath58 as @xmath59 .
similarly , we say that an event happens for a.a.e .
( asymptotically almost every ) @xmath60 if it holds with probability @xmath61 over a random choice of @xmath60 .
we say that an assignment of @xmath62 spins to the nodes achieves * @xmath63-partial recovery * if at least @xmath64 of these spins match the true spins , or at least @xmath64 match after a global flip of all the spins .
moreover , an algorithm that outputs a vector of spins @xmath65 ( indexed by nodes ) is said to _ achieve partial recovery _ and there exists @xmath66 such that @xmath67 achieves @xmath63-partial recovery a.a.s .
in the limit @xmath59 . next , we define the broadcast tree model ( which we introduced informally earlier ) . the broadcast tree model is a stochastic process that starts with a single root node @xmath68 at level @xmath69 whose spin @xmath70 is chosen uniformly at random . each node in turn gives birth to @xmath49 same - spin children and @xmath50 opposite - spin children , where @xmath71 is the poisson distribution with expectation @xmath72 .
this process continues until level @xmath73 at which point it stops , and the nodes at level @xmath73 are called the _ leaves_. ( the nodes on level @xmath74 that by chance do not give birth to any children are not considered leaves , even though they are leaves in the graph - theoretic sense . )
an algorithm observes the spins at the leaf nodes and the topology of the tree , and the goal is to recover the root spin : an algorithm that outputs a spin @xmath75 is said to _ achieve partial recovery _ on the tree if there exists @xmath66 such that @xmath76 with probability at least @xmath63 , as @xmath77 .
the reparameterization in terms of @xmath78 becomes particularly convenient here : each node gives birth to @xmath79 children , so @xmath80 is the average branching factor .
moreover , each child has probability @xmath81 of having spin opposite to that of its parent .
it is known that taking the majority vote of the leaves is optimal in theory , in the sense that it achieves partial recovery for @xmath43 @xcite and that for @xmath82 , partial recovery is information - theoretically impossible @xcite .
this is called the kesten stigum bound , and it can also be interpreted as a condition on the second - largest eigenvalue of an appropriately defined transition matrix .
there are many other natural variants of the broadcast tree model , that are more general instances of multi - type branching processes . however , even for simple extensions , the precise information - theoretic threshold is still unknown . in the potts model , where nodes are labeled with one of @xmath83 labels , sly @xcite showed that the kesten stigum bound is not tight , as predicted by mzard and montanari @xcite . and for an asymmetric extension of the binary model above , borgs et al .
@xcite showed that the kesten stigum bound is tight for some settings of the parameters . in our setting
, this historical context presents a substantial complication because if we apply a monotone adversary to a broadcast tree model and it results in a complex propagation rule , there may not be good tools to prove that partial recovery is impossible .
there is a close connection between the stochastic block model and the broadcast tree model , since the local neighborhood of any vertex in the graph is locally tree - like .
hence , if our goal is to prove a random vs. semirandom gap for community detection , a natural starting point is to establish such a gap for the broadcast tree model .
it turns out that there is more than one natural way to define a monotone adversary for the broadcast tree model , but it will not be too difficult to establish lower bounds for either of them .
the more difficult task is in finding an adversary that can plausibly be coupled to a corresponding adversary in the stochastic block model , and this will require us to put many sorts of constraints on the type of adversary that we should use to obtain a separation for the broadcast tree model . in the broadcast tree model , we will work with two notions of a monotone adversary .
one is weak and will be used to show our separation results : our other adversary is stronger , and we will establish upper bounds against this adversary ( with the recursive majority algorithm ) to give a strong recoverability guarantee : an upper bound against this latter model amounts to a recovery guarantee without any assumptions as to what happens after a ` mutation ' of labels for example , a genetic mutation might affect reproductive fitness and change the birth rule for the topology
. other variants of monotone adversaries could also be justified .
[ [ majority - fails ] ] majority fails + + + + + + + + + + + + + + it is helpful to first see how adversaries in these models might break existing algorithms .
recall that in the broadcast tree model , taking the majority vote of the leaves yields an algorithm that works up to the kesten stigum bound , and this is optimal since reconstruction is impossible beyond this . in the language of @xmath84 and @xmath85 ,
each node gives birth to @xmath86 nodes of the same label and @xmath87 nodes of the opposite label .
hence in a depth @xmath73 tree we expect @xmath88 leaves , but the total bias of the spins can be recursively computed as @xmath89 . the fact that majority works can be proven by applying the second moment method and comparing the bias to its variance .
however , an overwhelming number of the leaves are on some path that has a flip in the label at some point : we only expect @xmath90 nodes with all - root - spin ancestry , a vanishing proportion as @xmath77 .
the strong monotone adversary has control over all the rest , and can easily break the majority vote .
even the monotone cutting adversary can control the majority vote , by cutting leaves whose spin matches the root but whose parents have the opposite label .
this happens for a constant fraction of the leaf nodes , and this change overwhelms the majority vote .
so majority vote fails against the semirandom model , for _ all _ nontrivial @xmath84 and @xmath85 .
this is an instructive example , but we emphasize that breaking one algorithm does not yield a lower bound .
for example , if the algorithm knew what the adversary were doing , it could potentially infer information about where the adversary has cut edges based on the degree profile , and could use this to guess the label of the root .
[ [ the - problem - of - orientation ] ] the problem of orientation + + + + + + + + + + + + + + + + + + + + + + + + + + many first attempts at a separation in the broadcast tree model ( which work ! ) rely on knowing the label of the root . however , such adversaries present a major difficulty in coupling them to a corresponding adversary in the graph .
each graph gives rise to many overlapping broadcast trees ( the local neighborhood of each vertex ) and a graph adversary needs to simultaneously make all of these tree reconstruction problems harder .
this means a graph adversary can not focus on trying to hide the spin of a specific tree root ; rather , it should act in a local , orientation - free way that inhibits the propagation of information in all directions .
a promising approach is to look for nodes @xmath5 whose neighbors in @xmath91 all have the opposite label , and cut all of these edges .
such nodes serve only to further inter - connect each community , and cutting their edges would seem to make community detection strictly harder . in the corresponding broadcast tree model
, these nodes @xmath5 represent flips in the label that are entirely corrected back , and deleting them penalizes any sort of over - reliance on distributional flukes in how errors propagate .
for example , majority reconstruction in the tree fully relies on predictable tail events whereby nodes with label different from the root may lead to subtrees voting in the correct direction nonetheless .
[ [ the - problem - of - dependence ] ] the problem of dependence + + + + + + + + + + + + + + + + + + + + + + + + + now , however , a different sort of problem arises : if we were to naively apply the adversary described above to a broadcast tree , this would introduce complicated distance-@xmath92 dependencies in the distribution of observed spins , as certain diameter-@xmath92 spin patterns are banned in the observed tree ( as they would have been cut ) . in particular , the resulting distribution is no longer a markov random field .
this is not inherently a problem , in that we could still hope to prove stronger lower bounds for such models beyond the kesten stigum bound .
however , the difficulty is that even for quite simple models on a tree ( e.g. the potts model @xcite , asymmetric binary channels @xcite ) the threshold is not known , and the lower bound techniques that establish the kesten stigum bound seem to break down .
an alternative is to specifically look for degree-@xmath92 nodes @xmath5 whose neighbors in @xmath91 have the opposite label , and cut both incident edges .
although there are still issues about making this a markov random field , we can alleviate them by adding a @xmath93-clique potential on each degree-@xmath92 node and its two neighbors .
then after we marginalize out the label of the degree-@xmath92 node , the @xmath93-clique potential becomes a @xmath92-clique potential , and we return to a markov random field over a tree ! in other words , if we ignore the spin on a degree-2 node and treat its two incident edges like a single edge , we return to a well - behaved spin propagation rule .
we are now ready to describe the adversary that we will use to prove theorems [ thm : intro - main ] and [ thm : intro - treehard ] .
we only need two additional adjustments beyond the discussion above .
instead of making every possible degree-@xmath92 cutting move as described earlier , we will only cut with probability @xmath94 .
we will tune @xmath94 to ensure that the monotone changes we make do not overshoot and accidentally reveal more information about the underlying communities by cutting in too predictable a fashion .
finally , our adversary adds local restrictions to where and how it cuts , to ensure that the changes it makes do not overlap or interfere with each other ( e.g. chains of degree-@xmath92 nodes ) .
these details will not be especially relevant until section [ sec : no - lrc ] , where they simplify the combinatorics of guessing the original precursor graph from the observed graph .
[ dist : graph ] let @xmath95 be given .
write @xmath80 and @xmath81 .
we sample a ` precursor ' graph @xmath96 , and apply the following adversary : we can now outline the proof of our main theorem . in order to show that partial recovery is impossible , it suffices to show that it is impossible to reconstruct the relative spin ( same or different ) of two random nodes @xmath4 and @xmath5 , better than random guessing . before applying the adversary , an @xmath97-radius neighborhood @xmath98 around @xmath4
resembles the broadcast tree model rooted at @xmath4 . after the adversary
is applied to the graph , @xmath98 resembles a broadcast tree model with a corresponding cutting adversary applied .
this resemblance will be made precise by the coupling argument of section [ sec : coupling ] ; there will be some complications in this , and our tree will not be uniformly the same depth but will have a serrated boundary .
we show in section [ sec : treehard ] that when the cutting adversary is applied to the tree , the tree reconstruction problem ( guess the spin of @xmath4 from the spins on the boundary of @xmath98 ) becomes strictly harder : the average branching factor becomes lower due to sufficient cutting , while the new spin propagation rule resembles classical noisy propagation with at least as much noise ( so long as we marginalize out the spins of nodes ) .
we then apply the proof technique of evans et al .
@xcite to complete the tree lower bound .
the final step in the proof is to show that reconstructing the relative spin of @xmath4 and @xmath5 is at least as hard as reconstructing the spin of @xmath4 given the spins on the boundary of @xmath98 ( which separates @xmath4 and @xmath5 with high probability ) .
this is one of the most technically involved steps in the proof . in the lower bound for the random model @xcite , this step called `` no long - range correlations '' was already an involved calculation on a closed - form expression for @xmath99 . in our setting , in order to get a closed - form expression for the conditional probability , we will sum over all the possible precursor graphs @xmath91 that could have yielded the observed graph @xmath100 .
we characterize these precursors in lemma [ lemma : unsurgery ] , and the main reason for the and nodes in our adversary construction is to simplify this process .
a natural open question is whether one can find the optimal monotone adversary .
for instance , we have not even used the power to add edges within communities .
note however , that our current adversary is delicately constructed in order to make each step of the proof tractable .
it is not too hard to propose alternative adversaries that seem stronger , but it is likely that one of the major steps in the proof ( tree lower bound or no long - range correlations ) will become prohibitively complicated .
recent predictions on the performance of sdp methods @xcite could be suggestive of the true semirandom threshold .
it is well - known that sparse , random graphs are locally tree - like with very few short cycles .
this is the basis of mossel neeman
sly s approach in coupling between a local neighborhood of the stochastic block model and the broadcast tree model @xcite .
hence , our first order of business will be to couple a typical neighborhood from our graph distribution ( distribution [ dist : graph ] ) to the following tree model , against which we can hope to show lower bounds . given the parameters @xmath101 , generate a tree @xmath102 with spins @xmath103 as follows : [ dist : tree ] * start with a root vertex @xmath68 , with spin @xmath104 chosen uniformly at random from @xmath105 .
* until a uniform depth @xmath106 ( where the root is considered depth 0 ) , let each vertex @xmath5 give birth to @xmath49 children of the same spin and @xmath50 children of opposite spin .
* apply the graph adversary ( from distribution [ dist : graph ] ) to this tree ; this involves assigning markings and , and cutting some nodes .
keep only the connected component of the root . *
remove all nodes of depth greater than @xmath73 ( the bottom @xmath93 levels ) . *
remove any node at depth @xmath73 along with its siblings , exposing the parent as a leaf .
the reason we trim @xmath93 levels at the bottom is to ensure that the markings and cuttings in @xmath102 match those in @xmath100 , since these depend on the radius-@xmath93 neighborhood of each node and edge , respectively . removing nodes at level @xmath73
will ensure that the more complicated spin interactions of nodes and their neighbors do not span across the boundary of a tree neighborhood in the graph we want to cleanly separate out a tree recovery problem .
we use a slightly non - conventional definition of ` leaves ' : [ def : leaves ] when we refer to the * leaves * of a tree sampled from distribution [ dist : tree ] we mean the nodes at depth @xmath73 plus any nodes at depth @xmath107 that are revealed during the last step .
nodes at depth @xmath74 that happen to give birth to no children are not considered leaves ; if the root is and gets cut by the adversary so that the tree is a single node , this single node is not considered a leaf .
we can couple the above tree model to neighborhoods of the graph : [ prop : coupling ] let @xmath95 be given , and let @xmath108 be any vertex of the graph .
let @xmath109 there exists a joint distribution @xmath110 such that the marginals on @xmath111 and @xmath112 match the graph and tree models , respectively , while with probability @xmath58 : * there exists an isomorphism ( preserving edges , spins , and the markings and ) between the tree @xmath102 and a vertex - subset @xmath98 of @xmath100 , * the tree root corresponds to the vertex @xmath108 in the graph , and the leaves correspond to a vertex set @xmath113 that separates the interior @xmath114 from the rest of the graph @xmath115 .
* we have @xmath116 . as proven in @xcite
, there exists a coupling between the precursor graph @xmath117 and the precursor tree @xmath118 , such that @xmath119 matches the radius @xmath106 neighborhood @xmath98 of @xmath108 in @xmath91 , which has size @xmath120 ; this fails with probability @xmath53 .
next the adversary assigns vertex markings ( and ) to both models .
the marking of each vertex is deterministic and only depends on the topology of the radius-@xmath93 neighborhood of the vertex .
thus the markings in @xmath91 and @xmath119 match up to radius @xmath73 .
some nodes are _ cuttable _ , i.e. both their neighbors have opposite spin ; the cuttable nodes in @xmath91 and @xmath119 also match up to radius @xmath73 .
the adversary now cuts the edges incident to a random subset of cuttable vertices ; we can trivially couple the random choices made on @xmath91 with those on @xmath119 up to radius @xmath73 .
we keep only the connected component of the root in @xmath102 ; likewise let us keep only the corresponding vertices in @xmath98 , i.e. only those still connected to @xmath108 by a path in @xmath98 .
after removing nodes of depth greater than @xmath73 , we have removed the subset of @xmath102 for which the markings and the action of the adversary differ from those in @xmath100 .
thus at this stage , the tree exactly matches the radius @xmath73 neighborhood of @xmath108 in @xmath100 , along the isomorphism given by the coupling from @xcite .
any boundary vertex of this neighborhood must have distance exactly @xmath73 from @xmath108 , thus its corresponding vertex in the tree has depth @xmath73 and is a leaf . passing to the final step of removing leaves in @xmath102 and their siblings ,
we remove their corresponding nodes from @xmath98 ; this does not change the boundary - to - leaves correspondence .
in this section we show that our tree distribution ( distribution [ dist : tree ] ) evidences a random vs. semirandom separation in the broadcast tree model .
recall that the goal is to recover the spin of the root from the spins on the leaves , given full knowledge of the tree topology and the node markings ( and ) .
recall the non - conventional definition of ` leaves ' ( definition [ def : leaves ] ) .
let @xmath121 be the advantage over random guessing , defined such that @xmath122 is the probability that the optimal estimator ( maximum likelihood ) succeeds at the above task .
we will show that our tree model is asymptotically infeasible @xmath123 as @xmath77 ) for a strictly larger range of parameters @xmath124 than that of the corresponding random model .
[ prop : treehard ] for every real number @xmath40 , there exists @xmath41 such that @xmath43 , yet for a tree sampled from distribution [ dist : tree ] with parameters @xmath125 , @xmath126 and depth @xmath73 , we have @xmath127 as @xmath77 .
recall that the condition @xmath128 is the classical kesten
stigum bound , which is sufficient to beat random guessing in the random model @xcite .
several decades later , this bound was found to be tight for the random model @xcite .
there remain many open questions regarding the hardness of tree models , and some care was required in crafting an adversary for this problem that keeps the proof of this lower bound tractable . broadly , the kesten stigum bound asserts that recoverability depends on the average branching factor ( a contribution from the tree topology ) and the amount of noise ( a contribution from the spin propagation rule ) .
the first step of our proof is to distinguish these in our distribution : we can first generate a tree topology from the appropriate marginal distribution , and then sample spins from the conditional distribution given this tree .
we will show how to view this distribution on spins within the lower bound framework of @xcite .
moreover , the new spin propagation rule is at least as noisy as the original , while our cutting adversary has strictly decreased the average branching factor .
we first re - state our tree model in terms of topology generation followed by spin propagation . instead of letting each node
give birth to @xmath49 same - spin children and @xmath50 opposite - spin children , we can equivalently let each node @xmath5 give birth to @xmath79 children and then independently choose the spin of each child to be the same as @xmath5 with probability @xmath129 and opposite to @xmath5 otherwise .
( the equivalence of these steps is often known as ` poissonization ' . ) here the correspondence between @xmath124 and @xmath78 is as usual : @xmath80 and @xmath81 .
this allows us to first sample the entire tree topology without spins
. then we can add markings ( and ) , since these depend only on the topology .
next , we sample the spins as above , by generating an appropriate independent @xmath105 value @xmath130 on each edge , indicating whether or not a sign flip occurs across that edge .
finally , we cut edges according to the adversary s rule .
given the parameters @xmath101 , generate a tree @xmath102 with spins @xmath103 as follows : [ dist : tree-1 ] * start with a root vertex @xmath68 . until a uniform depth @xmath106 ,
let each vertex @xmath5 give birth to @xmath79 nodes . * mark nodes as and according to the rules of the graph adversary .
* for each edge @xmath131 , generate an independent flip value @xmath130 which is @xmath132 with probability @xmath129 and @xmath133 otherwise . *
choose the root spin @xmath104 uniformly from @xmath105 .
propagate spins down the tree , letting @xmath134 where @xmath131 is the edge connecting @xmath5 to its parent @xmath4 .
* cut edges according to the adversary s rule , keeping only the connected component of the root . *
trim the bottom of the tree ( according to the last two steps of distribution [ dist : tree ] ) .
it is clear that this tree distribution ( distribution [ dist : tree-1 ] ) is identical to the original tree distribution ( distribution [ dist : tree ] ) .
our next step will be to re - state this model in yet another equivalent way .
the goal now is to sample the final tree topology ( including which edges get cut ) before sampling the spins . consider a node @xmath5 , its parent @xmath4 , and its single child @xmath135 . instead of writing the spin propagation as independent flips @xmath136 and @xmath137 , we will write it as random variables @xmath138 and @xmath139 . here
@xmath138 is equal to 1 if the adversary decides to cut @xmath5 ( and 0 otherwise ) , and @xmath139 is equal to 1 if @xmath140 ( and @xmath133 otherwise ) .
this means if @xmath141 then @xmath138 is 1 with probability @xmath94 ( and 0 otherwise ) ; and if we do not have @xmath141 then @xmath142 .
hence @xmath143 with probability @xmath144 . if @xmath143 then @xmath139 is irrelevant because the adversary will cut @xmath5 from the tree . conditioned on @xmath142 , @xmath139 takes the value @xmath132 with probability @xmath145 and @xmath133 otherwise .
this means that for a ( but not cut ) node @xmath5 , the joint distribution of @xmath146 obeys a propagation rule that is equivalent to putting noise @xmath147 ( instead of @xmath85 ) on edges @xmath7 and @xmath148 , where ( using the definition of @xmath94 ) @xmath149 one can verify that @xmath150 for all @xmath151 . for most nodes in the tree , we are simply going to replace @xmath85 by @xmath147 on the incident edges , which gives the correct joint distribution of @xmath146 but not the correct joint distribution of @xmath152 .
this is acceptable because the distribution of leaf spins ( given the root spin ) is still correct ; for this reason we have made sure that none of the leaves are nodes .
the only time when we actually care about the spin of a node is in the case where the root @xmath68 is . in this case
, the root might be cut by the adversary , yielding a 1-node tree ( with no revealed leaves ) .
otherwise , if the root is but not cut , our spin propagation rule needs to sample the spins of the root s two children ( a node that is not cut must have degree @xmath92 ) from the appropriate joint distribution over @xmath153 ; let @xmath154 denote this distribution , conditioned on @xmath155 .
it will not be important to compute @xmath154 explicitly ( although it is straightforward to do so ) .
we are now ready to state the next tree model .
this model is equivalent to the previous ones in that the joint distribution of the root spin , topology , markings , and leaf spins is the same .
the spins on the nodes ( other than the root ) do not have the same distribution as before , but this is irrelevant to the question of recovering the root from the leaves . given the parameters @xmath101 ,
generate a tree @xmath102 with spins @xmath103 as follows : [ dist : tree-2 ] * start with a root vertex @xmath68 . until a uniform depth @xmath106 ,
let each vertex @xmath5 give birth to @xmath79 nodes . * mark nodes as and according to the rules of the graph adversary .
* decide which nodes the adversary should cut : cut each node independently with probability @xmath144 . *
let @xmath156 for edges @xmath131 that are incident to a node , and let @xmath157 for all other edges . * for each edge @xmath131 ,
generate an independent flip value @xmath130 which is @xmath132 with probability @xmath158 and @xmath133 otherwise .
* choose the root spin @xmath104 uniformly from @xmath105 .
if the root is but not isolated : let @xmath159 be its children , draw @xmath160 , and let @xmath161 , @xmath162 .
for all other nodes , propagate spins down the tree as usual , letting @xmath134 where @xmath131 is the edge connecting @xmath5 to its parent @xmath4 . *
trim the bottom of the tree ( according to the last two steps of distribution [ dist : tree ] ) .
our next step is to further modify this tree model in ways that only make it easier , in the sense that the advantage @xmath163 can only increase .
first , we will address the issue of the complicated propagation rule @xmath154 in the case that the root is .
suppose the root is ( but not cut ) and consider deterministically setting @xmath164 where @xmath159 are the root s two children .
from there , spin propagation continues as usual .
we claim that this new model can only be easier than the original one . to see this , note that the new model can ` simulate ' the original one : upon observing leaf spins drawn from the new model , drawn @xmath165 and then , for each @xmath166 and for each leaf @xmath135 descended from @xmath167 , replace @xmath168 by @xmath169 . for convenience
we will also replace the first level of the tree by deterministic zero - noise propagation in the case where the root is not .
we can similarly argue that the model only gets easier , since one can simulate the old model using the new model by sampling the noise on the first level .
note that we now have a tree model such that once the topology is chosen , the spin - propagation rule is very simple : at each edge @xmath131 , a sign flip occurs independently with some probability @xmath170 .
hardness results for such a tree model were studied by evans et al .
, who established the following bound on the advantage @xmath171 for a fixed tree topology @xmath102 ( @xcite theorem 1.3 ) : @xmath172 and where @xmath173 , and @xmath174 denotes the unique path from the root to leaf @xmath5 . in our case ,
the tree @xmath102 ( including both the tree topology and markings ) is random , so the advantage is @xmath175 $ ] , by the law of total probability . by jensen s inequality ,
@xmath176\right)^2 \le \ee_t[\delta_t^2 ] \le \ee_t\left [ 2 \sum_{\text{leaves } v } \theta_v^2 \right ] \defeq w.\ ] ] our goal is to show @xmath177 ( as @xmath77 ) , so it is sufficient to show @xmath178 .
any leaf in our tree is at height @xmath179 .
furthermore we have @xmath180 for edges incident to the root and @xmath181 elsewhere .
therefore , for any leaf @xmath5 , we can bound @xmath182 . note that we have not actually used the fact that @xmath147 strictly exceeds @xmath85 ; we will in fact be able to prove our result by leveraging only the decrease in branching factor and not the noise increase .
now we have @xmath183 \le 2 \ , \ee_t[\text{\ # leaves } ] ( 1 - 2 \eps)^{2(r-2)}\ ] ] and so we need only to bound the expected number of leaves . for @xmath184 ,
level @xmath3 _ denote the set of vertices at distance @xmath3 from the root ( so the root is on level 0 ) .
call levels @xmath185 the _ base _ levels and call levels @xmath186 the _ cutting _ levels .
imagine growing the tree using @xmath79 births as usual , and marking nodes as and as usual .
now allow the adversary to cut vertices as usual , except refuse to cut any node that is not on a cutting level .
note that this can only result in more leaves , and so this new process will give an upper bound on the expected number of leaves in our actual model .
to analyze the new process , suppose we have a node @xmath5 on a base level .
the number of descendants @xmath187 of @xmath5 on the next base level is a random variable whose distribution does not depend on the ancestors of @xmath5 ; this independence is the reason for defining base levels and cutting levels . define @xmath188 by @xmath189 $ ] ; then @xmath190 , since there is some nonzero probability that one of @xmath5 s descendants will be cut at the subsequent cutting level .
the expected number of leaves in the entire tree is now at most @xmath191 , where the factor @xmath192 accounts for the first level and the last @xmath193 levels that are not followed by a base level .
now @xmath194 , which goes to 0 as @xmath77 provided that @xmath195 hence this inequality suffices for impossibility of recovery . however , @xmath188 depends on the topology of the final tree model , which depends on @xmath84 , @xmath85 , and @xmath94 , so this inequality is slightly more complicated than it looks .
we write @xmath196 to make this dependence explicit ; recall that @xmath94 is a ( continuous ) function of @xmath85 .
we argued above that @xmath197 for all @xmath198 .
in particular , at the critical value @xmath199 for the _ random _ model , we have @xmath200 but @xmath201 is a continuous function of @xmath85 , so if we take @xmath85 to be slightly less than @xmath202 , we must still have @xmath203 , while @xmath128 , as desired .
this completes the proof of theorem [ thm : intro - treehard ] ( semirandom vs. random separation on trees ) . in appendix
[ sec : lower - bound - explicit ] , we explicitly compute a lower bound on the separation @xmath204 .
the lower bound of the previous section will form the core of a lower bound for the graph : we have already established that , for a large range of parameters , it is impossible to learn the spin of a node @xmath4 purely from the spins at the boundary @xmath113 of its tree - like local neighborhood .
we will now see why this makes recovery in the graph impossible : there is almost nothing else to learn about @xmath205 from beyond its local neighborhood once we know the spins on @xmath113 .
[ lemma : no - lrc ] let a graph @xmath100 ( including markings and ) and spins @xmath103 be drawn from distribution [ dist : graph ] .
let @xmath206 , @xmath207 , @xmath208 be a vertex - partition of @xmath100 such that * @xmath113 separates @xmath209 from @xmath210 in @xmath100 ( no edges between @xmath209 and @xmath210 ) , * @xmath211 , * @xmath113 contains no nodes .
then @xmath212 for asymptotically almost every ( a.a.e . ) @xmath100 and @xmath103 .
here @xmath213 denotes the spins on @xmath209 and @xmath214 denotes the spins on @xmath215 . to clarify , when we refer to or nodes in @xmath100 we are referring to the original markings that they were assigned in @xmath91 .
for instance , if a node is cut by the adversary , it is still considered in @xmath100 even though it no longer has degree @xmath92 .
recall that the markings ( and ) in @xmath100 are revealed to the reconstruction algorithm . in lemma [ lemma : no - lrc ] , we do crucially use that @xmath113 does not contain nodes : if a node in @xmath113 were with spin @xmath132 , and we then revealed that its neighbor in @xmath210 has spin @xmath133 , this would strengthen our belief that the neighbor in @xmath209 has spin @xmath132 , as otherwise the node would have some substantial probability of having been cut , which is observed not to be the case .
so lemma [ lemma : no - lrc ] would be false if we allowed nodes in @xmath113
. the proof of lemma [ lemma : no - lrc ] will require a thorough understanding of the distribution of spins given @xmath100 , which we can only obtain by understanding the possible precursors of @xmath100 under the adversary .
the reason for the and nodes in our adversary construction is to make these precursors well - behaved .
we start with some simple observations .
let @xmath91 be a graph that yields @xmath100 ( with nonzero probability ) by application of the adversary .
[ obs : good ] a node has degree at least @xmath93 in @xmath100 .
a node has at least three non - degree-@xmath92 neighbors in @xmath91 and the adversary will not remove these .
[ obs : create ] the adversary does not create any new degree-@xmath92 nodes .
in other words , if a node has degree @xmath92 in @xmath100 then it has degree @xmath92 in @xmath91 ( with the same two neighbors ) . in order to create a degree-@xmath92 node @xmath5
, the adversary must cut at least one edge incident to @xmath5 .
this can only happen if @xmath5 is either or .
it can not be because it does not have degree @xmath92 in @xmath91 .
but it also can not be or else it has degree at least 3 in @xmath100 ( by observation [ obs : good ] ) .
the key property of the /construction is that the adversary can not change whether or not a node has the following ` goodness ' property .
say that a node has the * goodness * property with respect to a graph if at least three of its neighbors do not have degree @xmath92 .
[ lemma : goodness ] the nodes are precisely the nodes that have the goodness property with respect to @xmath100 . note that this is not tautological because the nodes are defined as the nodes that have the goodness property with respect to @xmath91 , not @xmath100 .
we need to show that the goodness property is invariant under the adversary s action .
first we assume @xmath5 has goodness in @xmath91 and show that it also has goodness in @xmath100 . since @xmath5 has the goodness property in @xmath91 , it has three neighbors @xmath216 in @xmath91 that do not have degree @xmath92 .
each @xmath167 remains connected to @xmath5 in @xmath100 since the adversary only cuts edges that are incident to a degree-@xmath92 node .
furthermore , each @xmath167 does not have degree @xmath92 in @xmath100 because degree-@xmath92 nodes can not be created ( observation [ obs : create ] ) .
now we show the converse : assume @xmath5 does not have goodness in @xmath91 and show that it still does not have goodness in @xmath100 .
the only way that @xmath5 could obtain goodness is if at least one of its degree-@xmath92 neighbors @xmath4 becomes non - degree-@xmath92 ( while remaining connected to @xmath5 ) .
but this is impossible because whenever the adversary cuts an edge incident to a degree-2 vertex @xmath4 , it causes @xmath4 to become isolated in @xmath100 .
next we state an easy fact about the structure of nodes in @xmath100 .
[ lemma : marked ] the nodes in @xmath100 are precisely the degree-@xmath92 nodes with two neighbors , plus some isolated nodes .
every node in @xmath91 has degree-@xmath92 with two neighbors .
if the node is cut , it becomes isolated in @xmath100 ; otherwise it remains degree-@xmath92 with two neighbors .
conversely , let @xmath5 be degree-@xmath92 in @xmath100 with two neighbors ; we will show @xmath5 is .
since degree-@xmath92 nodes can not be created ( observation [ obs : create ] ) , @xmath5 has degree-@xmath92 in @xmath91 with the same two neighbors , and is therefore .
now we are ready to characterize the possible precursors @xmath91 of a given @xmath100 .
[ lemma : unsurgery ] suppose we have a graph @xmath100 ( including node markings ) and spins @xmath103 , drawn from distribution [ dist : graph ] .
the probability that a graph @xmath91 yields @xmath100 under the action of the adversary is zero unless @xmath91 can be obtained from @xmath100 by connecting each isolated node @xmath5 of @xmath100 to exactly two nodes of opposite spin to @xmath5 . in this case , if @xmath100 has @xmath217 isolated nodes and @xmath218 nodes with two opposite - spin neighbors , then @xmath219 where we define @xmath220 in the case @xmath221 .
recall that @xmath94 is the probability with which the adversary cuts each @xmath222 node that has two opposite - spin neighbors .
first suppose @xmath91 can yield @xmath100 via the adversary .
we will show that @xmath91 takes the desired form .
the nodes cut by the adversary are precisely the isolated nodes of @xmath100 .
every such node was originally connected to two opposite - spin nodes in @xmath91 .
therefore @xmath91 can be obtained from @xmath100 by connecting each isolated node to exactly two opposite - spin nodes .
conversely , let @xmath91 be obtained from @xmath100 by connecting each isolated node to exactly two opposite - spin nodes
. the nodes in @xmath100 are ( by lemma [ lemma : goodness ] ) precisely the nodes that have the goodness property in @xmath100 .
these are also precisely the nodes that have the goodness property in @xmath91 , since the process of connecting each isolated node to two nodes does not change goodness .
consider running the adversary on @xmath91 .
the nodes that it marks will be precisely the nodes in @xmath100 .
also , the nodes that it marks will be precisely the nodes in @xmath100 ; this is clear from lemma [ lemma : marked ] .
this means the adversary will output @xmath100 iff it chooses to cut the @xmath223 nodes that are isolated in @xmath100 and chooses not to cut the @xmath135 nodes that have two opposite - spin neighbors in @xmath100 .
this happens with probability @xmath224 .
we proceed as follows . in the ordinary stochastic block model ,
given an observed graph , the probability of any set of spins @xmath103 factorizes as a product of pairwise interactions . in our model , by summing over all possible precursors @xmath91 that could have lead to @xmath100 via the adversary , we find the same pairwise interactions together with a further global , combinatorial interaction .
we show that the neighborhood @xmath209 is too small to make a significant impact on this global interaction , while the pairwise interactions between @xmath209 and @xmath210 are weak , so that the only factors relevant to @xmath209 are the pairwise interactions within @xmath225 , which are independent of the spins in @xmath210 .
let @xmath91 denote the graph before the action of the adversary .
then @xmath226 factors into the following potentials on unordered pairs : @xmath227 with @xmath228 ranging over unordered pairs of distinct vertices , where @xmath229 where @xmath230 denotes adjacency in the graph @xmath231 . to leverage this description ,
let us sum over all possible precursors @xmath91 of @xmath100 under the adversary .
let @xmath232 denote the set of possible precursors @xmath91 of @xmath100 , as described by lemma [ lemma : unsurgery ] : @xmath91 is obtained from @xmath100 by connecting each of the @xmath223 isolated nodes to exactly two nodes of the opposite spin . @xmath233
the proportionality constant hidden by ` @xmath234 ' depends on @xmath100 but not on @xmath103 .
as @xmath91 is obtained from @xmath100 by replacing @xmath235 opposite - spin non - edges by opposite - spin edges , we have for every @xmath236 , @xmath237 thus none of the terms depend on the precise choice of @xmath91 : @xmath238 where we have dropped the constants that only depend on @xmath100 ( and not @xmath103 ) .
now we compute @xmath239 .
suppose there are @xmath240 isolated nodes of positive spin and @xmath241 of negative spin .
suppose there are @xmath242 nodes of positive spin , and @xmath243 of negative spin .
then the number of possible @xmath91 is @xmath244 we can establish that this global @xmath245 factor only barely depends on the spins of @xmath209 : [ lemma : binom - ratio ] for a.a.e .
@xmath246 , it holds for all @xmath213 , @xmath247 that @xmath248 the proof proceeds via concentration of measure and taylor expansion , and is deferred to appendix [ sec : binom - ratio - proof ] . paraphrasing this lemma , a.a.s . over @xmath249 , there exists a ` good ' subset of a.a.e .
@xmath250 such that @xmath239 is independent of @xmath213 up to a @xmath251 factor .
let us also require of ` good ' @xmath250 that the census ( sum of spins ) is @xmath252 ; as @xmath210 consists of all but @xmath120 nodes of @xmath100 , it is equivalent to ask for the same concentration over all spins of @xmath100 , and this occurs a.a.s . by hoeffding .
let @xmath253 be this set of ` good ' @xmath250 values . since @xmath254 a.a.s .
we have for any set @xmath255 ( which will be taken as either @xmath209 or @xmath225 below ) , @xmath256 we include a rigorous proof of this statement in appendix [ sec : omega - fact - proof ] .
it will be useful to factor the product @xmath257 of classical pairwise interactions into subsets of these interactions as @xmath258 here , for instance , @xmath259 denotes the product of @xmath260 over unordered pairs @xmath228 consisting of one vertex from @xmath215 and one from @xmath210 . the corresponding `` no long - range correlations '' proof in @xcite established that for ` good ' values @xmath254 , @xmath261 for a quantity @xmath262 depending only on @xmath263 and @xmath264 .
their proof of this fact holds verbatim in our setting : it only requires that @xmath265 and that the number of @xmath132 spins in the graph is distributed as @xmath266 .
similarly , we can factor the @xmath267 term as @xmath268 where , for instance , @xmath269 counts the number of nodes in @xmath209 with two opposite - spin neighbors .
this factorization holds as there are no @xmath209@xmath210 edges and no nodes in @xmath113 .
we can now absorb these terms into the @xmath9 terms : define @xmath270 and similarly @xmath271 , and @xmath272 , so that for @xmath254 , @xmath273 at this point we roughly adapt the proof of conditional independence from factorization in a markov random field , following the corresponding proof in @xcite .
we compute that for a.a.e .
@xmath274 , @xmath275 as desired .
we can now assemble all of the pieces to prove our main result .
we first prove that it is impossible to estimate the relative spin of any fixed pair of nodes , in a strictly larger parameter range than for the random model .
the impossibility of partial recovery will then easily follow .
[ prop : rel - spin - hard ] for all @xmath40 , there exists @xmath276 such that @xmath43 , yet given a graph @xmath100 ( including markings / ) from distribution [ dist : graph ] with parameters @xmath125 , @xmath126 , we have that for any fixed vertices @xmath228 , @xmath277 for a.a.e .
@xmath100 as @xmath59 . by ` fixed ' vertices @xmath228
we mean that the vertices are fixed before @xmath103 and @xmath100 are chosen ; so by symmetry it does nt matter which @xmath228 pair we fix .
given @xmath84 , choose @xmath85 as in proposition [ prop : treehard ] ( tree separation ) .
with probability @xmath58 , the tree coupling of proposition [ prop : coupling ] centered at vertex @xmath4 succeeds . in this case ,
the neighborhood @xmath98 of @xmath4 coupled to the tree is of size @xmath120 , and so with probability @xmath61 , @xmath5 lies outside this neighborhood .
let @xmath113 be the boundary vertices of @xmath98 , i.e. those vertices in @xmath98 corresponding to the leaves of the tree .
let @xmath209 be the interior @xmath114 , and let @xmath210 be the complement of @xmath98 in @xmath100 . by the law of total variance , @xmath278.\ ] ] with further probability @xmath58 ,
lemma [ lemma : no - lrc ] ( no long - range correlations ) succeeds , and we have @xmath279 so that @xmath280 but it follows from the coupling in proposition [ prop : coupling ] ( which includes spins and markings ) that @xmath281 since @xmath77 as @xmath59 , non - reconstruction in the tree model ( proposition [ prop : treehard ] ) implies that this latter variance converges to @xmath282 : the variance of a @xmath283-valued random variable with expectation @xmath53 is @xmath58 .
so we know that , with probability @xmath58 , a @xmath58 proportion of @xmath214 contribute a value @xmath58 to the expectation in ( [ eq : totvar ] ) .
the remaining @xmath53 proportion of @xmath214 contribute a value bounded in magnitude by @xmath282 , as the variance of a @xmath284-valued variable must be .
it follows that , a.a.s .
, @xmath285 = 1 - o(1),\ ] ] and the only way that this is possible in the @xmath284-valued distribution @xmath286 is if @xmath287 as desired
. it will follow immediately from proposition [ prop : rel - spin - hard ] that it is hard to find a partition that is correlated ( better than random guessing ) with the true one . for
if it was possible to find such a partition then one could also guess the relative spins of @xmath4 and @xmath5 .
we now complete the proof of our main theorem , restating it slightly more precisely .
for any @xmath40 , there exists @xmath276 so that @xmath42 and hence partial recovery in the stochastic block model @xmath17 is possible , and yet against the monotone adversary given in section [ sec : models ] , for any @xmath288 , no estimator of the spins achieves @xmath63-partial recovery with probability greater than @xmath53 as @xmath59 .
let @xmath289 be some assignment of spins to vertices that achieves @xmath63-partial recovery : @xmath289 agrees with the true spins on at least @xmath64 vertices , possibly after a global spin flip .
consider the relative spins @xmath290 , where @xmath4 and @xmath5 are distinct vertices ; these match the true relative spins on at least @xmath291 ordered pairs of distinct vertices .
if we choose two distinct vertices at random , the chance of correctly estimating their relative spin from @xmath289 is at least @xmath292 suppose for a contradiction that some estimator achieves @xmath63-partial recovery , for some @xmath66 , with probability @xmath293 not converging to @xmath69 as @xmath59 .
when @xmath63-partial recovery succeeds , the process above recovers the relative spin of two random vertices with probability at least @xmath294 , and note @xmath295 .
when partial recovery does not succeed , the process still recovers the relative spin of two random vertices with probability at least @xmath296 , as can be seen by plugging in @xmath297 to ( [ eq : eta ] ) .
it follows that we can recover the relative spin of two random vertices with probability at least @xmath298 $ ] which remains bounded above @xmath299 as @xmath59 , contradicting proposition [ prop : rel - spin - hard ] .
( proposition [ prop : rel - spin - hard ] assumes the markings on @xmath100 are known , but clearly the problem is only harder when they are unknown , since an estimator can choose to ignore them . )
we now turn to giving algorithms for partial recovery in the semirandom setting . in the existing literature on _
exact _ recovery , it has been observed that algorithms obtained through semidefinite programming extend almost automatically to semirandom models @xcite , and moreover many of these results match the information - theoretic thresholds for exact recovery .
in contrast , semidefinite programs for partial recovery , such as those of gudon and vershynin @xcite , come within a constant of the threshold but have been unable to close this gap .
in fact gudon and vershynin @xcite developed a general framework for using sdps to solve problems on sparse graphs , including partial recovery for the stochastic block model .
we define a notion of semirandom model for any problem in this framework , generalizing the semirandom model for community detection .
we show that any analysis that follows their framework carries over automatically to this semirandom model .
in particular , after minor modifications , the sdp analyzed in @xcite achieves partial recovery in the semirandom block model , up to within a constant factor of the classical threshold .
semirandom vs. random gaps offer an explanation for why it seems hard to find semidefinite programs for partial recovery that reach the information - theoretic threshold : the analysis often extends equally well to the semirandom model , where we know the threshold is strictly harder ! in the framework of @xcite , there are several key terms .
the first is a matrix @xmath113 which is usually computed in a simple manner from the adjacency matrix of the observed graph . in our case , we compute @xmath300 , where @xmath301 is the all - ones matrix , @xmath302 is a regularization constant , and @xmath209 is the observed adjacency matrix ( albeit with the non - standard convention that there are ones along the diagonal instead of zeros ) .
the goal is to show that @xmath113 which is a random matrix is near some fixed reference matrix @xmath73 . in our case
this will be @xmath303 , which is nearly equal to @xmath304 $ ] except on the diagonal .
then , one solves an sdp to maximize @xmath305 over @xmath306 , where @xmath253 is the space of symmetric psd @xmath307 matrices @xmath308 satisfying @xmath309 ( i.e. the diagonal entries of @xmath308 are at most @xmath282 ) . to have additional constraints , but we will not need this . ]
the goal of the analysis is to show that the sdp outputs some @xmath310 that is close to a `` ground truth '' @xmath311 , which in our case is the @xmath284-valued matrix @xmath312 of true relative spins .
the following result outlines the steps of the analysis .
[ prop : three ] suppose we have @xmath313 such that the following three conditions hold for some value @xmath314 , some function @xmath315 , and some matrix norm @xmath316 : 1 .
@xmath311 is a maximizer of the reference objective @xmath317 over @xmath253 + ( the reference sdp recovers the truth ) , 2 .
@xmath318 + ( the observed objective is close to the reference one in cut norm ) , 3 . if @xmath306 and @xmath319 , then @xmath320 + ( good solutions to the reference sdp are close to the ground truth ) .
then @xmath321 where @xmath310 is any maximizer of the empirical objective @xmath322 over @xmath253 , and @xmath323 is the grothendieck constant . here ,
the cut norm ( or @xmath57-to-@xmath282 norm ) of a matrix is defined as @xmath324 the proof of proposition [ prop : three ] follows from lemma 3.3 in @xcite , and uses grothendieck s inequality .
the partial recovery results of @xcite proceed by verifying the three conditions of proposition [ prop : three ] for a particular choice of parameters : [ prop : check - three ] assume @xmath46 .
let @xmath300 with @xmath302 , @xmath325 , and @xmath326 .
conditions ( 13 ) of proposition [ prop : three ] hold a.a.s . with @xmath327 , and @xmath328 as the frobenius norm @xmath329 . concretely , this means that we solve the following sdp : [ sdp : gv ] maximize @xmath330 subject to @xmath331 and @xmath332 , where @xmath302 .
note that the regularization constant @xmath333 is necessary because if we simply take @xmath334 then condition 1 of proposition [ prop : three ] fails . in @xcite
they estimate @xmath333 from the empirical average degree , but their arguments also apply to the case where we deterministically take @xmath302 .
( this requires knowledge of the parameters @xmath335 but we will address this issue later . ) in @xcite , condition 1 is shown in lemma 5.1 , and conditions 2 and 3 are implicit in the proof of lemma 5.2 .
these require a technical condition : @xmath336 , which for large @xmath0 simply amounts to @xmath46 . by propositions
[ prop : three ] and [ prop : check - three ] , we now attain the result @xmath337 a.a.s .
it is also shown in @xcite how to translate this to a precise partial recovery result : [ prop : rounding ] let @xmath326 .
for any @xmath338 , @xmath63-partial recovery succeeds a.a.s . in the stochastic block model ,
provided that @xmath339 , by taking the signs in the top eigenvector of @xmath310 .
[ cor : random - recovery ] there exists a constant @xmath210 such that @xmath63-partial recovery succeeds a.a.s . in the stochastic block model as @xmath59
whenever @xmath46 and @xmath340 .
the constant @xmath210 is quite large : @xcite sets @xmath341 , although no attempt is made to optimize this constant .
nevertheless this result is only off by a constant from the threshold , which is @xmath38 .
our random vs. semirandom separation becomes small as @xmath84 grows , so it remains plausible that semidefinite programs can achieve partial recovery when @xmath342 as @xmath54 .
in fact , sdps are known to distinguish random block model graphs from erds rnyi graphs in such a range @xcite , and it would be interesting to determine whether this carries through to partial recovery in the semirandom model .
we now turn to a semirandom view of the general gudon
vershynin framework . in the semirandom block model
, a monotone adversary is allowed to make changes aligned with the ground truth @xmath311 ; more formally , it can add a matrix @xmath343 to the observed adjacency @xmath209 , where @xmath343 is symmetric and has @xmath282 s in some same - spin entries where @xmath209 is @xmath69 , and @xmath133 s in some opposite - spin entries where @xmath209 is @xmath282 .
it is easily verified that @xmath311 maximizes @xmath344 over @xmath253 : every matrix in @xmath253 has entries in @xmath345 $ ] , and @xmath311 has @xmath105 entries that match the sign pattern of @xmath343 . following this observation , we propose a precise definition of semirandom models for gudon vershynin problems : a matrix @xmath343 is a * monotone change * if @xmath311 is a maximizer for @xmath346 over @xmath253 . in the * semirandom model * ,
a preliminary objective @xmath113 is generated from the random model , and then an adversary modifies the objective by any monotone change @xmath343 , yielding an observed matrix @xmath347 . note that the set of such monotone changes @xmath343 forms a cone , which matches the intuition that semirandom models can make unbounded changes , but only in certain directions aligned with the ground truth .
any analysis following the framework of proposition [ prop : three ] generalizes in a formal sense to the semirandom model : [ prop : s ] suppose that @xmath313 satisfy conditions ( 13 ) of proposition [ prop : three ] for some @xmath348 .
then it remains the case that @xmath349 , where @xmath350 is any maximizer of @xmath351 over @xmath253 , and @xmath343 is any monotone change ( in the sense above ) , which is allowed to depend adversarially on @xmath113 .
it suffices to establish conditions ( 13 ) of proposition [ prop : three ] with @xmath352 playing the role of @xmath73 and @xmath353 playing the role of @xmath113 : 1 .
as @xmath311 maximizes both @xmath317 and @xmath344 over @xmath253 , it certainly maximizes their sum @xmath354 .
we have @xmath355 , which by hypothesis is at most @xmath314 with high probability .
3 . suppose that @xmath306 and @xmath356 .
then @xmath357 since @xmath311 is a maximizer of @xmath346 over @xmath253 .
it follows from the original condition ( 3 ) that @xmath320 in the appropriate norm .
we have shown that any _ analysis _ of an sdp following the gudon
vershynin framework transfers to the semirandom model .
a recovery guarantee for the same sdp by different means would not automatically yield such a guarantee in the semirandom model .
this differs from exact recovery problems , where semirandom guarantees can take the success of the sdp as a black box , and certify that any successful instance remains successful under monotone changes @xcite .
let @xmath209 be sampled from the semirandom block model with parameters @xmath358 .
suppose that we know these parameters , so that we can run sdp [ sdp : gv ] with the deterministic choice @xmath302 . if @xmath46 and @xmath359 then this sdp achieves @xmath63-partial recovery a.a.s .
( after taking the sign of the top eigenvector of the sdp output @xmath310 ) .
this dependence on parameter knowledge is slightly unwieldy ; on the other hand , estimating the correct regularization parameter @xmath333 from @xmath360 is not immediately easy , since model - specific estimators are precisely the sort of techniques that semirandom models aim to penalize . instead , we show that deterministically taking @xmath361 in place of @xmath333 avoids this dependence
. indeed , the same sdp appears in the exact recovery literature @xcite with deterministic regularization parameter @xmath362 , which is comparatively very large , but comes at the cost of requiring precisely balanced communities .
our approach here is a middle ground that allows for natural @xmath363-scale deviations in the balance of the two communities .
there is nothing special about the value @xmath364 and it can in fact be taken to be anything of order strictly between @xmath365 and @xmath366 .
[ prop : new - sdp ] the gudon vershynin sdp ( sdp [ sdp : gv ] ) with regularization parameter @xmath367 achieves @xmath63-partial recovery against the semi - random model a.a.s .
so long as @xmath46 and @xmath359 . in light of proposition
[ prop : s ] , it is sufficient to show that this sdp works in the random model .
we check the three conditions for @xmath368 where @xmath369 , @xmath370 , and @xmath326 where @xmath302 and @xmath371 .
we will use the fact that the three conditions are satisfied for @xmath313 , and we will achieve nearly the same parameters @xmath372 as in that case . 1 .
we certify that @xmath311 maximizes @xmath373 by sdp duality .
the dual sdp reads @xmath374 where @xmath5 runs over all vertices in the graph . to certify that @xmath311 is optimal
, it suffices by complementary slackness to find @xmath375 such that @xmath376 and @xmath377 . since @xmath378 and @xmath311 are psd , the second condition is equivalent to @xmath379 .
as the columns of @xmath311 are spanned by @xmath103 , this second condition reads @xmath380 ; expanding , @xmath381 where @xmath382 is the set of nodes with the same spin as node @xmath5 , and @xmath383 is the complement of @xmath382 . by hoeffding we have @xmath384 a.a.s . and so @xmath385 in particular , @xmath386 for all @xmath5 a.a.s .
this choice of @xmath375 guarantees @xmath380 so it remains to show that @xmath376 .
since @xmath380 , it suffices to show that @xmath387 for all @xmath388 .
but note that @xmath389 @xmath390 v \geq 0,\ ] ] since @xmath391 a.a.s . so that both @xmath301 and @xmath392 are psd .
this establishes optimality of @xmath311 .
we have @xmath393 and so this step follows from the corresponding step for @xmath313 .
3 . suppose that @xmath306 satisfies @xmath394 .
letting @xmath395 we have @xmath396 and @xmath397 .
notice that @xmath398 since both matrices are psd , while @xmath399 a.a.s . where @xmath400 are the community sizes .
thus , @xmath401 applying condition 3 for @xmath313 , we obtain @xmath402 . by proposition
[ prop : rounding ] , we obtain partial recovery by rounding the leading eigenvector , under the same conditions as in corollary [ cor : random - recovery ] , and with the same constant @xmath210 .
this analysis completes the proof of theorem [ thm : intro - robust - sdp ] from the introduction .
independently , @xcite proved a result on sdp robustness that matches ours .
additionally , they consider robustness to a different adversary that can only add or remove @xmath31 edges but is not required to be monotone .
our methods also extend to this case ( and match their results ) because @xmath31 changes can only change the cut norm by @xmath31 .
here we establish that recursive majority achieves reconstruction in the semirandom broadcast tree model , even with respect to the strong adversary that has total control over entire subtrees .
as one would imagine , in recursive majority the root spin is estimated as the majority vote of its children s estimated spins , which in turn are estimated as the majority vote of their children , and so on down to the known leaf spins . in the random model , this algorithm falls short of the kesten stigum bound by a factor of @xmath403 as @xmath54 @xcite .
nevertheless , it remains the algorithm of choice in practice for tree reconstruction problems , often by the name of ` parsimony ' @xcite .
it seems that the advantages of recursive majority over majority only become clear when studying these algorithms through semirandom models ! to keep things simple , we will first consider trees in which each non - leaf node has exactly @xmath84 children , matching the setting of @xcite and much of the tree reconstruction literature . at the end , we show how to adapt our proof in a straightforward manner to the case where the number of children of each non - leaf node is a poisson random variable .
[ prop : eps - star ] let @xmath404 be the probability of a majority ` yes ' vote of @xmath84 voters each voting ` yes ' with probability @xmath293 : @xmath405 let @xmath406 be drawn from the strong semirandom model on the regular tree with braching factor @xmath84 , height @xmath73 , and noise @xmath85 .
let @xmath407 be defined by @xmath408 } \frac{m_k(q)}{q},\ ] ] and let @xmath409 be the maximizer .
so long as @xmath410 , the recursive majority algorithm correctly recovers the root spin from the topology and leaf spins , with probability at least @xmath411 .
conversely , if @xmath412 , there exists an adversary forcing recursive majority to fail with probability @xmath58 as @xmath77 .
let @xmath406 be drawn from the strong semirandom model , starting with a @xmath84-regular tree of height @xmath73 .
let @xmath413 denote the success probability on such a tree . as a base case , we have @xmath414 .
for each child of the root , the spin matches that of the root with probability @xmath415 , in which case the child will be correctly labeled by recursive majority with probability @xmath416 , by induction .
the other children are under adversary control .
then the number of recovered child labels agreeing with the root label stochastically dominates @xmath417 , for any choice of adversary .
the recovered root label is the majority vote of the recovered child labels , so we obtain that @xmath418 moreover this inequality is exactly tight for a certain adversary , which replaces each replaceable subtree by a path down to a single leaf of spin opposite to the root .
we are interested in the limiting recovery probability @xmath419 ; this is the greatest fixed point in @xmath420 $ ] of the function @xmath421 .
equivalently , @xmath422 is the greatest solution in @xmath423 $ ] of the fixed - point equation @xmath424 .
geometrically , this is the greatest intersection point of the graph of @xmath425 with the line of slope @xmath426 through the origin . before proceeding ,
it is worth noting the geometry of the graph of @xmath425 . from the definition ,
@xmath425 is monotone on @xmath420 $ ] , with @xmath427 , @xmath428 , and @xmath429 . by expanding the probability mass function of the binomial distribution ,
@xmath425 is a degree @xmath84 polynomial , and thus smooth .
less obviously , @xmath425 is strictly convex on @xmath430 $ ] and strictly concave on @xmath431 $ ] @xcite .
the line of slope @xmath92 ( @xmath432 ) does not intersect the graph of @xmath425 nontrivially , since @xmath433 on @xmath430 $ ] and @xmath434 on @xmath431 $ ] .
the line of slope @xmath282 ( @xmath435 ) certainly intersects the graph of @xmath425 at @xmath362 and @xmath282 .
hence there exists some maximal @xmath436 , at which the line of slope @xmath437 intersects the graph of @xmath425 tangentially at some @xmath409 .
equivalently , we can characterize this slope as the maximum slope of the line defined by the origin and any point on the graph of @xmath425 : @xmath408 } \frac{m_k(q)}{q}.\ ] ] by the concavity and convexity properties of @xmath425 , the graph of @xmath425 lies below the line of slope @xmath282 on @xmath438 , and above on @xmath439 .
it follows that @xmath440 . as we sweep @xmath83 from @xmath409 to @xmath282 , the slope @xmath441 passes continuously from @xmath437 to @xmath282 , granting an intersection @xmath442 satisfying @xmath443 for every @xmath444 $ ] , by the intermediate value theorem .
thus , whenever @xmath445 , recursive majority achieves recovery with probability @xmath446 whenever @xmath412 , the two curves do not intersect on @xmath420 $ ] except at @xmath69 , so the limiting success probability @xmath447 is @xmath69 .
it is interesting that , as the noise @xmath85 varies , we see the success probability jump discontinuously from @xmath448 to zero .
this contrasts with the behavior of recursive majority in the ordinary broadcast tree model , in which the error probability transitions continuously to @xmath362 at a threshold @xcite . for any fixed @xmath84 , computing the critical value @xmath407 amounts to maximizing the polynomial @xmath449 , or equivalently finding the unique root of its derivative in @xmath450 $ ] .
for example it is easily computed that @xmath451 .
however , we would like some asymptotic understanding of the values @xmath407 as @xmath54 .
the following asymptotic expression holds for @xmath407 as @xmath54 : @xmath452 by the maximality property defining @xmath407 and @xmath409 , we must have @xmath453 from which it follows that @xmath454 thus , as @xmath455 $ ] , we know that @xmath456 in analyzing recursive majority in the random model , mossel @xcite computed this derivative using russo s formula .
let us assume for the moment that @xmath84 is odd ; the even case is similar . then @xmath457
if we evaluate this at @xmath458 , we obtain : @xmath459 in the second - to - last step , we have used the asymptotic identity @xmath460 as @xmath54 , which still holds when @xmath60 depends on @xmath84 so long as @xmath461 .
now for @xmath462 , the derivative @xmath463 tends to @xmath69 as @xmath54 , while for @xmath464 , this tends to @xmath57 as @xmath54 .
thus , writing @xmath465 , we must have @xmath466 .
we immediately obtain a lower bound on @xmath407 : @xmath467 for a matching upper bound , apply hoeffding s inequality for a binomial tail bound , to find @xmath468 and use the maximality of @xmath409 to find @xmath469 so that @xmath470 is an asymptotic upper bound for every @xmath471 , which completes the proof .
note that @xmath472 is the probability of success at the threshold @xmath407 , so this proof also provides some sense of the critical success probability : @xmath473 so we observe a very strong threshold : as we vary @xmath85 , there is a discrete jump from very likely success of recursive majority to almost - sure failure ! this result does not change if we pass back to the poisson - birth tree used throughout the rest of this paper , rather than the @xmath84-regular tree . here
@xmath474 $ ] , which leads to a poisson - averaged derivative , and the same threshold in @xmath83 for when the derivative passes from large to small .
by poissonization , we can also write @xmath475 and the standard chernoff bound for `` poisson races '' yields , for @xmath476 , @xmath477 which is as strong as the hoeffding bound used above .
the results above constitute a proof for theorem [ thm : intro - recmaj ] : recursive majority achieves recovery against the strong semirandom model , so long as @xmath478 or , rearranging , @xmath479 with asymptotics holding as @xmath54 .
it is interesting to see how the semirandom model can yield recovery results across an entire family of previously studied distributions .
the asymmetric broadcast trees studied in @xcite may be described as markov chains on the tree with transition matrix @xmath480 where the ` asymmetry ' @xmath94 may be positive or negative .
the strong semirandom adversary can simulate these models : starting from the broadcast tree model with noise @xmath481 , an adversary searches the tree top - down for either @xmath132-to-@xmath133 or @xmath133-to-@xmath132 transitions , depending on the sign of @xmath94 , and flips the entire resulting subtree with probability @xmath482 , recursing into subtrees .
thus , the semirandom recovery results above guarantee for free that recursive majority is competitive against all of these models simultaneously , as are all algorithms robust to the strong semirandom model .
by contrast , these models each admit a simple but brittle `` weighted majority '' reconstruction algorithm , where the weighting must be perfectly tuned to the specific model a tiny deviation in the parameters @xmath483 will cause those census algorithms to fail as @xmath77 .
in revisiting the stochastic block model from the perspective of semirandom models , we showed that there is a tension between establishing sharp thresholds and obtaining algorithms with natural robustness guarantees .
there are many more classical problems in statistics and machine learning where semirandom models could offer a promising way to move beyond average - case analysis and explore issues related to robustness .
in particular , belief propagation is one of the most far - reaching heuristics , but at high ` temperature ' it can result in algorithms such as majority that we have shown are not robust . at low ` temperature ' it leads to more robust algorithms , like recursive majority , that are connected to convex optimization .
can exploring ` temperature ' further lead to interesting , provable tradeoffs between robustness and statistical power that result in a richer understanding of how belief propagation performs in practice ?
the authors would like to thank philippe rigollet and the mit learning theory group for helpful discussion , and roxane sayde for reading a draft of this document .
here we compute an explicit lower bound on the separation @xmath204 between the average branching factor in the random model and that of the six - level - periodic branching rule described at the end of section [ sec : treehard ] .
we will need to assume @xmath484 .
fix @xmath485 .
we will further modify the six - level - periodic rule by considering a node to be if it has three _ children _ of degree not equal to @xmath92 and we disregard the contribution of the parent to this rule , to simplify the computation ; this only further reduces cutting . in this one - period tree ,
the ` base ' levels are @xmath69 ( the root ) and @xmath486 ( the leaves ) , and the ` marking ' level is level @xmath93 . in expectation , there are @xmath487 nodes on level @xmath92 .
let us focus on the subtree descending from one of these nodes .
let @xmath488 ; then by poissonization , before cutting , each node @xmath5 on level @xmath92 gives birth to @xmath489 degree-@xmath92 children and @xmath490 degree - not-@xmath92 children .
in particular , @xmath5 is with probability @xmath491 . supposing that this is the case ,
let us see how many leaves get cut . in expectation
, @xmath5 has @xmath492 degree-@xmath92 children on level @xmath93 , each with exactly one child on level @xmath493 , which is with probability @xmath491 . in that case
, it gives rise to @xmath494 \ , \ee[\pois(k ) \given\ ; \ne 1]$ ] leaves , in expectation ; but the level-@xmath93 degree-@xmath92 node and its subtree gets cut with probability @xmath495 .
the expected number of leaves cut in the subtree descending from a level-@xmath69 node is : @xmath496 \ , \ee[\pois(k ) \given\ ; \ne 1 ] ) \\ & = k^3 p
\delta \eps^2 \cdot \prob{\pois(k(1-p ) ) \geq 3}^2 \cdot ( kp + \ee[\pois(k(1-p ) ) \given\ ; \geq 3 ] \ , \ee[\pois(k ) \given\ ; \ne 1]).\end{aligned}\ ] ] whenever @xmath484 and @xmath39 , we must have @xmath497 and @xmath498 ( from the definition ( [ eq : delta ] ) of @xmath94 ) . in this range
we have @xmath499 \ , \ee[\pois(k ) \given\ ; \ne 1 ] ) \defeq \mathcal{k}(k).\ ] ] now , choosing any @xmath85 such that @xmath500 the semirandom model will be impossible while the random model is possible .
this concludes our explicit computation of a separation .
we have not made any attempt to optimize it , and that remains an interesting open question for future work .
recall that @xmath223 is the number of isolated nodes in @xmath100 , and @xmath501 is the number of nodes .
let @xmath314 be the number of excess @xmath132 spins among the isolated nodes , i.e. there are @xmath502 isolated nodes of spin @xmath132 and @xmath503 of spin @xmath133 .
similarly let @xmath504 be the number of excess @xmath132 spins among the nodes .
write @xmath505 and @xmath506 , where @xmath507 denotes the number of excess @xmath132 spins among the isolated nodes of @xmath209 , and likewise for the others
. we will need a number of results on the sizes of these values .
[ lemma : conc ] we have the following results a.a.s .
* @xmath508 , * @xmath509 , * @xmath510 .
the first result is easy : @xmath507 and @xmath511 are bounded above by @xmath263 , which is @xmath120 by assumption .
we will establish the remaining concentration results through the bounded differences method ; however , this bounded difference property will require controlling the maximum degree of a node .
note that a.a.s .
every node of @xmath91 ( and thus @xmath100 ) has degree at most @xmath512 , by a chernoff - and - union argument .
let @xmath513 denote the graph obtained from @xmath91 by simultaneously removing all edges incident to nodes of degree larger than @xmath512 .
note now that the radius-@xmath493 neighborhood of any node in @xmath513 has size at most @xmath514 .
let @xmath515 denote the census ( number of @xmath132 spins minus number of @xmath133 spins ) among those nodes that are ` cuttable ' in @xmath513 , i.e. nodes @xmath5 that are degree-@xmath92 in @xmath513 , with two opposite - sign ( to @xmath5 ) neighbors that each have at least @xmath93 non - degree-@xmath92 neighbors .
this property of @xmath5 depends only on the radius-@xmath93 neighborhood in @xmath513 .
given any vertex @xmath4 in @xmath91 , if we add or remove any number of edges incident to @xmath4 in @xmath91 , we can only change @xmath515 by at most @xmath514 , as we can only change the ` cuttable ' status of the previous and new radius-@xmath493 neighborhoods of @xmath5 in @xmath513 .
( here we need radius @xmath493 instead of @xmath93 because by changing edges incident to @xmath5 we can push the neighbors of @xmath5 over the @xmath516-degree cutoff . )
this constitutes a bounded differences property for the function @xmath517 .
we now apply the following concentration inequality : let @xmath518 be a function of a random graph of size @xmath0 with independent edges ( not necessarily identically distributed ) .
suppose that , when we add or remove any number of edges incident to any given vertex @xmath5 , the value of @xmath518 changes by at most @xmath72
. then @xmath519 \given \geq \lambda c \sqrt{n } } \leq 2 \exp(-\lambda^2/2).\ ] ] this result is classical , and is based on applying azuma
hoeffding to a _
vertex exposure martingale _ ; see for example chapter 7 of @xcite .
note that @xmath520 = 0 $ ] due to spin - reversal symmetry .
it follows that @xmath521 a.a.s .
moreover , @xmath522 a.a.s .
, so that the true census of cuttable nodes in @xmath91 is at most @xmath523 a.a.s .
our adversary turns a subset of these cuttable nodes into isolated nodes , choosing each independently with probability @xmath94 ; applying hoeffding s inequality separately to the @xmath132 and @xmath133 cuttable nodes , we see that the census of isolated nodes in @xmath100 is @xmath524 a.a.s . but this census equals @xmath525 , and @xmath526 , so we have @xmath527 a.a.s .
the same argument applies ( with a strict subset of the technical mess ) to show that @xmath528 , and that @xmath223 and @xmath501 concentrate within @xmath524 of their expectations .
it is also straightforward to see that @xmath529 $ ] and @xmath530 $ ] are @xmath531 ; this amounts to showing that the probability of any vertex being isolated in @xmath100 , or in @xmath100 , is bounded above zero as @xmath59 .
but these two properties are local , depending only on a neighborhood in the graph , so this is clear .
this completes the proof of the concentration results ( lemma [ lemma : conc ] ) .
now we proceed to the proof of lemma [ lemma : binom - ratio ] .
we need to show that for a.a.e .
@xmath532 , it holds for all @xmath533 that @xmath534 we will need sufficiently tight asymptotics for powers of binomials in this regime : @xmath535 we now apply these asymptotics to the task at hand : @xmath536 as every non - error term cancels .
the result now follows : if the logarithm of an expression is @xmath53 then the expression itself is @xmath537 .
enumerate all possible values @xmath538 for @xmath539 .
let @xmath540 and let @xmath541 .
we know @xmath542 and @xmath543 for some @xmath544 .
we want to show that for some @xmath545 we have @xmath546 with probability @xmath547 ; here the probability is over @xmath3 drawn proportional to @xmath548 . if @xmath549 is the set of @xmath3 for which @xmath550 , we have @xmath551 .
this means @xmath552 and so @xmath553 . we need @xmath547 and @xmath545 so it suffices to take any @xmath554 such that @xmath555 and @xmath556 , i.e. @xmath554 goes to 0 slower than @xmath557 does .
throughout this paper , we have assumed @xmath558 ; such block models are often called ` assortative ' . however , everything does carry through equally in the ` dissortative ' case @xmath30 . here
we briefly sketch the relevant changes .
our semirandom models are certainly designed for an assortative model , and are entirely too powerful in the dissortative case for example , by randomly adding and removing edges appropriately , the current semirandom block model can simulate the erds
rnyi distribution @xmath559 when starting from any dissortative model , which clearly reveals no community structure . instead
, the semirandom model in the dissortative case should add edges between communities , and remove edges within communities , so that these monotone changes are aligned with the latent structure to be recovered .
similarly , the semirandom tree models are able to cut or replace any subtree that follows a same - spin edge .
this leads to many sign changes throughout the paper .
we can still couple the resulting graph neighborhoods to a tree distribution ; although this tree distribution might look quite different at a glance , it couples perfectly with an assortative tree model , corresponding via @xmath560 , by flipping all spins at odd levels of the tree .
thus we obtain a random vs. semirandom separation for the dissortative tree model .
our recovery results for recursive majority in section [ sec : recmaj ] carry through this coupling also , guaranteeing robust recovery by recursive anti - majority in the dissortative tree model . much of the `` no long - range correlations '' argument of section [ sec : no - lrc ] carries through unchanged , but precursors @xmath91 of @xmath100 now reconnect isolated nodes with two _ same - spin _ nodes .
hence the new formula for @xmath239 is @xmath561 which still satisfies lemma [ lemma : binom - ratio ] , by negating each @xmath504 variable everywhere in the proof in appendix [ sec : binom - ratio - proof ] .
so we also obtain a random vs. semirandom separation in the dissortative block model .
the sdp upper bounds in section [ sec : upper - bound ] carry through with very few changes of sign .
one verifies that the unchanged reference objective @xmath303 does maximize @xmath346 where @xmath343 is the matrix form of any semirandom change as redefined above .
a sign does flip in the proof of proposition [ prop : new - sdp ] : we now have @xmath562 overall we obtain the same guarantees on semirandom partial recovery as in the assortative model , requiring @xmath563 rather than @xmath46 . | the stochastic block model is one of the oldest and most ubiquitous models for studying clustering and community detection . in an exciting sequence of developments , motivated by deep but non - rigorous ideas from statistical physics , decelle et al .
@xcite conjectured a sharp threshold for when community detection is possible in the sparse regime .
mossel , neeman and sly @xcite and massouli @xcite proved the conjecture and gave matching algorithms and lower bounds . here
we revisit the stochastic block model from the perspective of semirandom models where we allow an adversary to make ` helpful ' changes that strengthen ties within each community and break ties between them .
we show a surprising result that these ` helpful ' changes can shift the information - theoretic threshold , making the community detection problem strictly harder .
we complement this by showing that an algorithm based on semidefinite programming ( which was known to get close to the threshold ) continues to work in the semirandom model ( even for partial recovery ) .
this suggests that algorithms based on semidefinite programming are robust in ways that _ any _ algorithm meeting the information - theoretic threshold can not be .
these results point to an interesting new direction : can we find robust , semirandom analogues to some of the classical , average - case thresholds in statistics ? we also explore this question in the broadcast tree model , and we show that the viewpoint of semirandom models can help explain why some algorithms are preferred to others in practice , in spite of the gaps in their statistical performance on random models . |
the ability to extract weak signals from noise is generally enhanced by having advanced knowledge of the expected signals @xcite . for low - mass compact binary inspiral ,
the expected waveform can be accurately computed in the band of earth - based detectors , and matched filtering is used to search the detector output for the signals @xcite .
these searches for gravitational - wave chirp signals are the archetypical example of using advanced knowledge of a waveform @xcite in gravitational - wave astronomy .
once the signals are detected , the extraction of information about the astrophysical sources responsible for their generation will be of great interest . for waves from the inspiral of low - mass compact binary systems ,
the problem of information extraction is reasonably straightforward .
a well developed formalism , based on matched filtering , exists to characterize the accuracy of parameter estimation for sufficiently strong signals @xcite .
in general , the strongest sources of gravitational waves are composed of high density material moving at relativistic speeds in extreme gravitational fields and detailed computations of the expected waveforms are not always possible . among the most plausible examples of such sources
are supernova explosions @xcite or binary black hole mergers @xcite .
efforts to model these systems , and to determine the gravitational waveforms emitted by them , are under way @xcite .
as the simulations become more sophisticated , the gravitational waveforms become better approximations to the waves that will arrive from real astrophysical systems .
a challenge for gravitational - wave astronomy is to use the partial information from simulations to enhance signal detection and information extraction .
if robust features of the waves can be identified using the simulations and those features can be associated with definite physical processes , a great deal of information might be obtained about the structure of dense matter or of spacetime itself .
searches for gravitational - wave bursts have been made by a number of groups @xcite .
these searches generally take an unbiased approach by using time - frequency methods to identify excess power in the data stream .
plausible physical waveforms are only used to determine the efficiency of the search and report an upper limit . in this note
, we outline a method to use information from numerical simulations in searches for gravitational - wave bursts .
this method is based on the excess - power statistic @xcite and provides an algorithmic approach to incorporating the source - modeling information into searches .
moreover , it allows for the incorporation of new information as it becomes available .
the theoretical formulation of the method is presented in sec .
[ s : theory ] with an example of the construction in sec .
[ s : example ] .
finally , other possible applications are mentioned in sec .
[ s : summary ] .
the signal @xmath0 from a gravitational wave detector is sampled to produce a time series @xmath1 , where @xmath2 and @xmath3 is the time between samples .
this signal can be written as @xmath4 where @xmath5 is the detector noise and @xmath6 is the ( possibly absent ) gravitational - wave signal .
for the purposes of our presentation , we assume that the detector noise is gaussian with joint probability distribution @xmath7 where @xmath8 and @xmath9 . this assumption is not critical to the approach , but makes for easier presentation and motivates the inner product used below . for stationary noise , the kernel @xmath10
is determined by the power - spectral density .
our approach is based on the excess - power search method described in refs .
@xcite although extensions of the present approach which might be more powerful will be described elsewhere @xcite .
the gravitational wave signals form a subspace @xmath11 of the vector space @xmath12 spanned by all @xmath13-dimensional vectors given by eq .
( [ e : signal - vector ] ) .
if the subspace @xmath11 is also a vector space , then the optimal statistic for the detection of these gravitational waves in gaussian noise was derived by anderson _
one first decomposes the signal into @xmath14 which lies in @xmath11 and @xmath15 which is orthogonal to @xmath11 , so that @xmath16 then , the optimal statistic by which to identify gravitational - wave signals which are in the vector space @xmath11 is @xmath17 where the inner product is defined by @xmath18 the key to this approach is the determination of the signal subspace @xmath19 and the associated operator which projects vectors from @xmath20 into @xmath11 .
this was already pointed out by anderson et al @xcite where an example based on a fourier basis was described in detail .
suppose numerical simulations yield a set of @xmath21 gravitational - wave signals @xmath22 where @xmath23 .
each waveform arises from a simulation with some parameters @xmath24 .
if the simulations are costly and only some representative sample can be completed , it is desirable to search for all signals with similar characteristics to the sample .
a restricted search method might use the known waveforms as matched filters to search for waveforms which match the simulations well .
this approach is extremely limited when there are only a few simulations or the simulations are known to be only qualitatively correct
. the alternative approach , which we advocate here , is to use the excess power method by constructing a vector space @xmath19 of signals which capture the essential features of the waveforms given by the simulations .
this approach also has its limitations since the space of signals generated by some class of astrophysical sources is not , in general , expected to be a vector space .
nevertheless , the projection of the detector output signal into a vector space of sufficiently small dimensions can still be a powerful method to distinguish signal from noise when combined with coincidence between detectors .
the simplest approach to determine a vector subspace would use the gram - schmidt method to determine an orthonormal set of vectors from the waveforms @xmath22 and hence identify @xmath19 as the vector space spanned by all the simulation waveforms .
the orthonormal basis of this space is therefore @xmath25 where @xmath26 now , any of the original simulation waveforms can be re - constructed as a sum over the basis @xmath27 . in general , however , we are interested not in the exact re - construction of the original waveform , but the identification of robust features of the waveforms .
this is because the simulations may not be completely accurate representations of the physical sources .
for this reason , we would like to use a smaller subspace which captures the essential features of the simulated waveforms _
well enough_. our approach is similar to that used when determining the grid - spacing in a matched filtering search and is most easily explained by describing the algorithmic selection of basis vectors . first , normalize all the simulated waveforms using the inner product ( [ e : inner - prod ] ) ; denote the normalized waveforms by @xmath28 .
suppose then , that one has used @xmath29 of the simulated waveforms to construct an orthonormal set @xmath30 where @xmath31 ; this is a basis for a vector space @xmath32 .
compute the norm @xmath33 for all waveforms which have not been used to construct the basis of @xmath34 .
the quantity @xmath35 represents the _ match _ of the @xmath36th remaining waveform with the nearest vector in the space @xmath34 . select the signal with @xmath37 .
the number of basis vectors is incremented by one if @xmath38 where @xmath39 is some minimal match that will be tolerated in the search .
this procedure is repeated until a set of @xmath40 orthonormal vectors has been identified and @xmath41 for all remaining waveforms which have not been used in the construction .
this approach will tend to select out a subspace of signals @xmath19 which capture the essential features of the original set of simulated waveforms although the basis vectors may be unphysical . for typical source modeling simulations
, one expects that @xmath42 since the waveforms produced for different parameter values often share similar qualitative behavior . as described above
, the signal used to start the gram - schmidt construction is chosen arbitrarily and might not result in the smallest vector space of signals .
thus , the construction can be carried out repeatedly selecting a different seed signal each time . the seed signal which leads to the smallest @xmath43
is deemed the best choice for the algorithm since the vector space of smallest dimension has been identified .
it should be noted that waveforms sometimes need to be shifted in time relative to each other in order to obtain the best match . in each step of the construction
then , it may be necessary to shift each of the remaining waveforms in order to obtain the maximum match with the subspace @xmath34 .
this is done in the examples given below .
the simulation of supernova explosions is extremely complicated because it relies on detailed knowledge of physics at very high densities in addition to strong field gravitational effects . to date
, only simple models of gravitational - wave production in core collapse and supernova explosions have been considered @xcite .
these efforts provide guidance to the gravitational wave astronomy community in the form of waveform catalogs which represent a variety of observed behavior depending on the parameters used in a particular simulation .
the catalog of waveforms produced by zwerger and mller @xcite have become a reference point to calibrate searches for gravitational wave bursts @xcite .
the catalog provides the mass quadrupole wave amplitudes @xmath44 from axi - symmetric simulations of core - collapse .
the strain at the detector is given in terms of these quadrupolar amplitudes by @xmath45 where @xmath46 is the distance between the detector and the source and @xmath47 is the azimuthal angle .
the catalog consists of 78 different waveforms arising from representative simulations across the parameter space of the models . following the method outlined in the previous section , we have examined the number of basis vectors required to cover all 78 zwerger - mller ( zm ) waveforms at various levels of minimal match @xmath48 .
the results are summarized in fig .
[ f : basis - v - mismatch ] for a flat noise spectrum and the ligo - i design spectrum @xcite . at fixed @xmath39 , more basis vectors
are generally needed for the white noise case .
this is expected since high and low frequency components of the waveforms are suppressed in the inner - product by the ligo - i noise spectrum .
there is some variation in the number of basis vectors depending on the seed waveform used in the construction .
it is interesting to see that only 13 waveforms are needed to cover the space of waveforms at @xmath49 .
this confirms what can be seen by visually inspecting the zm waveforms , that is , these waveforms have dominant features which are robust across many examples in the catalog .
the number of basis vectors as a function of minimal match @xmath39 for the zm family of waveforms .
the dashed line shows the results for the inner product appropriate to white noise .
the triangles indicate the range of vlaues for the inner product appropriate to the ligo - i design using different seed waveforms . , title="fig:",scaledwidth=90.0% ] figure [ f : waveform - recon ] shows waveform a3b3g2 from the zm catalog plotted with the reconstructed waveform using different numbers @xmath50 of basis vectors .
the match , as defined in eq .
( [ e : match ] ) , between the waveform and the vector space @xmath51 is also shown . as indicated in fig .
[ f : basis - v - mismatch ] , the match is already greater than @xmath52 for only @xmath53 basis vectors . for completeness
, the exactly reconstructed waveform is also shown when waveform a3b3g2 has been absorbed into the vector space @xmath51 .
waveform a3b3g2 from the zm catalog ( light - gray line ) shown along with reconstructed waveforms for a different numbers @xmath50 of basis vectors .
the value of the match between waveform a3b3g2 and the corresponding vector space @xmath51 is also indicated .
the fourth plot on the bottom right shows the exactly reconstructed waveform once the a3b3g2 is absorbed into the vector space @xmath51 .
, scaledwidth=90.0% ] a nice feature of this approach to using source simulation information in searches for gravitational waves is the ease with which it can be updated using new information .
moreover , the method provides an algorithmic way to identify features of the simulated waveforms which are robust from simulation to simulation . as a concrete example of this
, we consider the set of waveforms produced by ott et al .
( oblw ) @xcite . as with the zm simulations ,
these waveforms also represent possible wave emission from core collapse supernova explosions .
the simulations explore alternate simulation methods in the numerical implementation and in accounting for the complicated physics of supernovae . when seeking guidance for the detection of gravitational waves , it is interesting to understand how different the resulting waveform family is from the zm family .
this is important for two reasons : first , it tests the robustness of any search method that is tuned to search for waveforms in the zm family , and second , it tests how much the waveform depends on changes in the underlying physics of the source .
we determined the number of basis vectors needed to cover the 149 waveforms from both zwerger - mller and ott et al .
in the same way as we determined the basis for only the zm waveforms . at @xmath54 ,
one needs 33 basis vectors . in this example
, we used the seed waveform that minimized the number of basis vectors for the zm family alone .
the resulting basis was constructed out of 12 waveforms from the zm family and 21 from the oblw family @xcite ; only 6 of the 12 zm waveforms were used in the zm - only basis set computed at this value of @xmath39 . for the purpose of this note , the exact numbers are not as important as the algorithmic approach which has been taken .
searches for gravitational waves using broad - band detectors and the extraction of physical insight from future detections will benefit greatly from close interaction with scientists who model sources of gravitational waves . in the case where complete and accurate information is available , it is well known that matched filtering provides the best approach to both detection and information extraction . for many physically complicated sources ,
however , there is a dearth of information available and even the best simulations are not accurate enough to support matched filtering as the main approach to detect the associated gravitational waves . in this note , we have sketched an algorithmic approach to determining a vector space which includes waveforms generated by the limited simulations and incorporates waveforms of similar character .
the benefit of this approach is that it is algorithmic and is easily updated using newer information provided by simulations .
the supernova simulations of zwerger - mller @xcite and ott et al .
@xcite have been used as an example to demonstrate the method . in a future publication , we will present more detailed investigations of these and other supernova simulations along with estimates of the performance of the approach in the presence of noise .
this is particularly important since the spiky nature of the basis vectors could increase the false alarm probability beyond a tolerable level in real detector noise one of the more interesting avenues to explore with this approach is , however , the detection of the late stages of binary black hole inspiral and merger @xcite .
current efforts in this direction have focused on identifying a detection family of waveforms for use in the search @xcite ; given the large number of templates required to span known waveform models , it seems worthwhile to explore the method outlined here as an alternative .
we would like to thank duncan brown , jolien creighton and alan wiseman for useful discussions .
this work has been supported by the national science foundation grant phy-0200852 .
patrick brady is also grateful to the alfred p sloan foundation and the research corporation cottrell scholars program for support .
99 h. tagoshi _ et al . _
( tama collaboration ) 2001 _ phys .
d _ * 63 * 062001 [ arxiv : gr - qc/0012010 ] b. abbott _ et al . _ ( the ligo scientific collaboration ) analysis of ligo data for gravitational waves from binary neutron stars _ preprint _ [ arxiv : gr - qc/0308069 ] . c. cutler and e. e. flanagan , phys .
d * 49 * ( 1994 ) 2658 [ arxiv : gr - qc/9402014 ] .
l. s. finn and d. f. chernoff , phys .
d * 47 * ( 1993 ) 2198 [ arxiv : gr - qc/9301003 ] .
h. dimmelmeier , j. a. font and e. mller , astrophys .
j. * 560 * ( 2001 ) l163 [ arxiv : astro - ph/0103088 ] . c. d. ott , a. burrows , e. livne and r. walder , astrophys .
j. * 600 * ( 2004 ) 834 [ arxiv : astro - ph/0307472 ] . l. lehner , class .
* 18 * ( 2001 ) r25 [ arxiv : gr - qc/0106072 ] .
j. g. baker , m. campanelli , c. o. lousto and r. takahashi , phys .
d * 65 * ( 2002 ) 124012 [ arxiv : astro - ph/0202469 ] .
b. bruegmann , w. tichy and n. jansen , arxiv : gr - qc/0312112 .
b. abbott _ et al .
_ [ ligo scientific collaboration ] , `` first upper limits from ligo on gravitational wave bursts , '' to appear in phys .
d [ arxiv : gr - qc/0312056 ] .
astone et al . , ( international gravitational event collaboration ) , phys . rev .
d * 68 * , 022001 ( 2003 ) , [ arxiv : astro - ph/0302482 ] .
w. g. anderson , p. r. brady , j. d. e. creighton and e. e. flanagan , phys .
d * 63 * ( 2001 ) 042003 [ arxiv : gr - qc/0008066 ] .
r. brady and s. ray - majumdar , `` interfacing gravitational - wave source modelling and observations '' , in preparation ( 2004 ) .
, phys . today * 52 ( oct ) * , 44 ( 1999 ) . | the detection of gravitational waves from astrophysical sources of gravitational waves is a realistic goal for the current generation of interferometric gravitational - wave detectors .
short duration bursts of gravitational waves from core - collapse supernovae or mergers of binary black holes may bring a wealth of astronomical and astrophysical information .
the weakness of the waves and the rarity of the events urges the development of optimal methods to detect the waves .
the waves from these sources are not generally known well enough to use matched filtering however ; this drives the need to develop new ways to exploit source simulation information in both detections and information extraction .
we present an algorithmic approach to using catalogs of gravitational - wave signals developed through numerical simulation , or otherwise , to enhance our ability to detect these waves .
as more detailed simulations become available , it is straightforward to incorporate the new information into the search method .
this approach may also be useful when trying to extract information from a gravitational - wave observation by allowing direct comparison between the observation and simulations . |
a polygon is _ simple _ if it has distinct vertices and interior - disjoint edges that do not pass through vertices .
geometric algorithms are often designed for simple polygons , but many also work for degenerate polygons that do not `` self - cross . '' a polygon with at least three vertices is _ weakly simple _ if for every @xmath3 , the vertices can be perturbed by at most @xmath4 to obtain a simple polygon .
such polygons arise naturally in numerous applications , e.g. , for modeling planar networks or as the geodesic hull of points within a simple polygon ( fig . [ fig : intro ] ) . with 16 vertices .
( b ) eight points in the interior of @xmath5 ( solid dots ) ; their geodesic hull is a weakly simple polygon @xmath6 with 14 vertices .
( c ) a perturbation of @xmath6 into a simple polygon .
[ fig : intro ] ] several alternative definitions have been proposed for weakly simple polygons , formalizing the intuition that such polygon does not cross itself .
some of these definitions were unnecessarily restrictive or incorrect ; see @xcite for a detailed discussion .
rib mor @xcite[theorem 3.1 ] proved that a weakly simple polygon with at least three vertices can be perturbed into a simple polygon continuously while preserving the lengths of its edges , and maintaining that no two edges properly cross .
chang et al .
@xcite gave an equivalent definition for simple polygons in terms of the frchet distance ( see section [ sec : preliminaries ] ) , in which a polygon is perturbed into a simple closed curve .
the latter definition is particularly useful for recognizing weakly simple polygons , since it allows transforming edges into polylines ( by subdividing the edges with steiner points which may be perturbed ) .
the perturbation of a vertex incurs only local changes , and need not affect the neighborhood of adjacent vertices ( in other words , we do not need to worry about stretchability of the perturbed configuration ) . we can decide whether an @xmath1-gon is simple in @xmath0 time by a sweepline algorithm @xcite .
chazelle s polygon triangulation algorithm also recognizes simple polygons ( in @xmath7 time ) , because it only produces a triangulation if the input is simple @xcite . recognizing weakly simple polygons , however , is more subtle .
cortese et al .
@xcite achieved this in @xmath8-time .
chang et al .
@xcite improved this to @xmath2 in general ; and to @xmath0 in several special cases .
they identified two features that are difficult to handle : a _ spur _ is a vertex whose incident edges overlap , and a _ fork _ is a vertex that lies in the interior of an edge ( a vertex may be both a fork and a spur ) . for polygons with neither forks nor spurs , chang et al .
@xcite gave an @xmath0-time algorithm .
in the presence of both forks and spurs , their solution is to eliminate forks by subdividing all edges that contain vertices in their interiors , potentially creating a quadratic number of vertices .
we show how to manage this situation efficiently , while building on ideas from @xcite and from arkin et al .
@xcite , and obtain the following main result .
[ thm : main ] deciding whether a given @xmath1-gon in the plane is weakly simple takes @xmath0 time .
our algorithm is detailed in sections [ sec : preprocess][sec : tree - exp ] .
it consists of three phases , simplifying the input polygon by a sequence of reduction steps .
first , the _ preprocessing _ phase applies known methods such as _ crimp reductions _ and _ node expansions _ ( section [ sec : preprocess ] ) .
second , the _ bar simplification _ phase successively eliminates all forks ( section [ sec : bars ] ) .
third , the _ spur elimination _ phase eliminates all spurs ( section [ sec : tree - exp ] ) .
when neither forks not spurs are present , we can decide weak simplicity in @xmath7 time @xcite .
we can also perturb any weakly simple polygon into a simple polygon , in @xmath0 time , by reversing the sequence of operations .
here , we review previously established definitions and known methods from @xcite and @xcite .
* polygons and weak simplicity . *
an _ arc _ in @xmath9 is a continuous function @xmath10\rightarrow \mathbb{r}^2 $ ] .
a _ closed curve _ is a continuous function @xmath11 .
a closed curve @xmath12 is _ simple _ ( also known as a _
jordan curve _ ) if it is injective . a ( _ simple _ ) _ polygon _ is the image of a piecewise linear ( _ simple _ ) closed curve . thus a polygon @xmath5 can be represented by a cyclic sequence of points @xmath13 , called _ vertices _ , where the image of @xmath12 consists of line segments @xmath14 , and @xmath15 in this cyclic order .
similarly , a _ polygonal chain _
( alternatively , _ path _ ) is the image of a piecewise linear arc , and can be represented by a sequence of points @xmath16 $ ] .
a polygon @xmath17 is _ weakly simple _ if @xmath18 , or if @xmath19 and for every @xmath3 there is a simple polygon @xmath20 such that @xmath21 for all @xmath22 .
this definition is difficult to work with because a small perturbation of a vertex modifies the two incident edges , which may be long , and the effect of a perturbation is not localized .
chang et al .
@xcite gave an equivalent definition in terms of the frchet distance : a polygon given by @xmath23 is weakly simple if for every @xmath3 there is a simple closed curve @xmath24 such that @xmath25 , where @xmath26 denotes the frchet distance between two closed curves .
the curve @xmath27 can approximate an edge of the polygon by a polyline , and any perturbation of a vertex can be restricted to a small neighborhood . with this definition , recognizing weakly simple polygons becomes a combinatorial problem independent of @xmath4 , as explained below . * bar decomposition and image graph . * two edges of a polygon @xmath5 _ cross _ if their interiors intersect at precisely one point , we call this an _ edge crossing_. weakly simple polygons can not have edge crossings . in the following ,
we assume that such crossings have been ruled out .
two edges of @xmath5 _ overlap _ if their intersection is a ( nondegenerate ) line segment .
the transitive closure of the overlap relation is an equivalence relation on the edges of @xmath5 ; see fig .
[ fig : bars](a ) where equivalence classes are represented by purple regions .
the union of all edges in an equivalence class is called a _
bar_. all bars of a polygon can be computed in @xmath0 time @xcite .
the bars are line segments that are pairwise noncrossing and nonoverlapping , and the number of bars is @xmath7 .
the vertices and bars of @xmath5 define a planar straight - line graph @xmath28 , called the _ image graph _ of @xmath5 .
we call the vertices and edges of @xmath28 _ nodes _ and _ _ segments__to distinguish them from the vertices and edges of @xmath5 .
every node that is not in the interior of a bar is called _
sober_. the set of nodes in @xmath28 is @xmath29 ( note that @xmath5 may have repeated vertices that correspond to the same node ) ; two nodes are connected by an edge in @xmath28 if they are consecutive nodes along a bar ; see fig .
[ fig : bars](b ) .
hence @xmath28 has @xmath7 vertices and edges , and it can be computed in @xmath0 time @xcite .
note , however , that up to @xmath7 edges of @xmath5 may pass through a node of @xmath28 , and there may be @xmath30 edge - node pairs such that an edge of @xmath5 passes through a node of @xmath28 .
an @xmath31-time algorithm can not afford to compute these pairs explicitly .
+ with 16 vertices ( @xmath5 is perturbed into a simple polygon for clarity ) .
( b ) the image graph of @xmath5 .
( c ) the combinatorial representation of @xmath5 .
[ fig : bars ] ] * operations .
* we use certain elementary operations that successively modify a polygon and ultimately eliminate forks and spurs .
suppose that an operation produces a weakly simple polygon if and only if it is performed on a weakly simple polygon
. then the operation is called _ ws - equivalent_. several such operations are already known ( e.g. , crimp reduction , node expansion , bar expansion ) .
we shall use these and introduce several new operations in sections [ ssec : bar - expansion][sec : tree - exp ] .
* combinatorial characterization of weak simplicity . * to show that an operation is ws - equivalent , it suffices to show the existence of @xmath4-perturbations .
we will use perfect matchings to combinatorially represent @xmath4-perturbations ( independent of @xmath4 or any specific embedding ) .
this representation is a variation of the `` strip system '' introduced in @xcite[appendix b ] .
let @xmath5 be a polygon and @xmath28 its image graph .
we construct a family of simple polygons as follows .
let @xmath32 , to be specified shortly .
for every node @xmath33 of @xmath28 , draw a disk @xmath34 of radius @xmath4 centered at @xmath33 .
choose @xmath3 sufficiently small so that the disks are pairwise disjoint , and no disk intersects a nonincident segment of @xmath28 .
let the _ corridor
_ @xmath35 of a segment @xmath36 be the set of points at distance at most @xmath37 from @xmath36 , outside of the disks @xmath34 and @xmath38 , that is , @xmath39 .
reduce @xmath4 further , so that all corridors are pairwise disjoint , and also disjoint from any disk @xmath40 , @xmath41 .
for every segment @xmath36 of @xmath28 , let the _ volume
_ @xmath42 be the number of edges of @xmath5 that contain both @xmath33 and @xmath43 ( either as an endpoint or in the interior ) .
for every segment @xmath36 , draw @xmath42 parallel line segments between @xmath44 and @xmath45 within @xmath35 ( see fig .
[ fig : bars](c ) ) .
finally , for every disk @xmath34 , construct a noncrossing straight - line perfect matching between the segment endpoints on the boundary @xmath44 ( in fig .
[ fig : bars](c ) , the matchings are drawn with circular arcs for clarity ) .
the line segments in the corridors and the perfect matchings in the disks produce a plane 2-regular graph @xmath46 .
denote by @xmath47 the family of all plane graphs constructed in this way such that @xmath46 is connected and visits the disks in the same cyclic order as @xmath5 . by theorem b.2 in @xcite
, @xmath5 is weakly simple iff @xmath48 .
every 2-regular graph @xmath49 defines ( and is defined by ) a linear order on overlapping edges of @xmath5 , hence on all edges containing an segment of the image graph of @xmath5 .
we say that two graphs in @xmath47 are _ combinatorially equivalent _ if they define the same order over the edges containing each segment of the image graph .
we denote by @xmath50)$ ] the path in @xmath46 corresponding to a path @xmath51 $ ] in @xmath5 including the matchings in @xmath52 and @xmath53 . in the absence of spurs ,
a weakly simple polygon @xmath5 defines a unique crossing - free perfect matching in each disk @xmath34 @xcite[section 3 ] which defines a 2-regular graph @xmath46 .
consequently , to decide whether @xmath5 is weakly simple it is enough to check whether @xmath49 .
this is no longer the case in the presence of spurs .
in fact , it is not difficult to construct weakly simple @xmath1-gons that admit @xmath54 combinatorially different perturbations into simple polygons ; see fig .
[ fig : ambig ] . ] note that the above combinatorial representation may have @xmath55 size , since each edge passing through a node @xmath33 contributes one edge to a matching in @xmath34 .
we use this simple combinatorial representation in our proofs of correctness , but our algorithm will not maintain it explicitly .
by a standard line sweep @xcite , we can detect and halt if any two edges properly cross .
we then simplify the polygon , using some known steps from @xcite , and some new ones .
all of this takes @xmath0 time .
arkin et al .
@xcite gave an @xmath7-time algorithm for recognizing weakly simple @xmath1-gons in the special case where all edges are collinear ( in the context of flat foldability of a polygonal linkage ) .
they use the ws - equivalent crimp - reduction operation .
a _ crimp _ is a chain of three consecutive collinear edges @xmath56 $ ] such that both the first edge @xmath57 $ ] and the last edge @xmath58 $ ] contain the middle edge @xmath59 $ ] ( the containment need not be strict ) .
the operation crimp - reduction@xmath60 replaces the crimp @xmath56 $ ] with edge @xmath61 $ ] ; see fig .
[ fig : crimp ] . $ ] with @xmath61 $ ] .
top : image graph .
bottom : polygon .
[ fig : crimp ] ] [ lem : crimp ] the crimp - reduction operation is ws - equivalent .
let @xmath62 and @xmath63 be two polygons such that @xmath63 is obtained from @xmath62 by the operation crimp - reduction@xmath60 . without loss of generality , assume that @xmath64 is horizontal with @xmath65 on the left and @xmath66 on the right .
first assume that @xmath62 is weakly simple .
then there exists a simple polygon @xmath67 .
we modify @xmath68 to obtain a combinatorial representation @xmath69 . without loss of generality , assume that edge @xmath57 $ ] is above @xmath59 $ ] ( consequently , @xmath58 $ ] is below @xmath59 $ ] ) in @xmath68 .
the modification involves the perfect matchings at the disks @xmath70 and @xmath71 , and all disks and corridors along the line segment @xmath72 .
denote by @xmath73 the set of maximal paths that lie in the convex hull of @xmath74 , below @xmath57 $ ] and above @xmath59 $ ] ; similarly , let @xmath75 be the set of maximal paths that lie in the convex hull of @xmath74 , below @xmath59 $ ] and above @xmath58 $ ] .
we proceed in two steps ; refer to fig .
[ fig : crimp1 ] .
first , replace the path @xmath76 $ ] with the edge @xmath61 $ ] such that the new edge follows @xmath77 $ ] from @xmath78 to @xmath70 , then @xmath59 $ ] from @xmath70 to @xmath71 , and finally @xmath79 $ ] from @xmath71 to @xmath80 .
second , exchange @xmath73 and @xmath75 such that the above - below order within each set of paths remains the same . since the above - below order within @xmath73 and @xmath75 is preserved , and the paths in @xmath73 ( resp . , @xmath75 ) lie below ( resp .
, above ) the new edge @xmath61 $ ] , no edge crossings have been introduced .
we obtain a simple polygon @xmath69 , which shows that @xmath63 is weakly simple . $ ] with an edge @xmath81 $ ] .
[ fig : crimp1 ] ] $ ] with a crimp @xmath56 $ ] .
[ fig : crimp2 ] ] next assume that @xmath63 is weakly simple .
then , there exists a simple polygon @xmath69 .
we modify @xmath82 to obtain a simple polygon @xmath67 ; refer to fig .
[ fig : crimp2 ] .
note that edge @xmath61 $ ] passes through all segments of @xmath28 along the line segment @xmath72 . in each of these segments @xmath83 , as well as in all disks along the line segment @xmath72 ,
replace the part of @xmath61 $ ] with three parallel line segments . in @xmath71 , connect the right endpoints of the top and middle segments by a new edge of the perfect matching ; and similarly in @xmath70 , connect the left endpoints of the middle and the bottom edges .
we obtain a simple polygon @xmath67 , which shows that @xmath62 is weakly simple . given a chain of two edges @xmath84 $ ] such that @xmath57 $ ] and @xmath59 $ ] are collinear but do not overlap , the merge operation replaces @xmath84 $ ] with a single edge @xmath77 $ ] .
the merge operation ( as well as its inverse , subdivision ) is ws - equivalent by the definition of weak simplicity in terms of frchet distance @xcite .
if we greedily apply crimp - reductions and merge operations , in linear time we obtain a polygon with the following two properties : 1 .
[ cond : a1 ] every two consecutive collinear edges overlap ( i.e. , form a spur ) .
[ cond : a2 ] no three consecutive collinear edges form a crimp . assuming properties [ cond : a1 ] and [ cond : a2 ] , we can characterize a chain of collinear edges with the sequence of their edge lengths .
[ lem : irreducible ] let @xmath85 $ ] be a chain of collinear edges in a polygon with properties [ cond : a1 ] and [ cond : a2 ]
. then the sequence of edge lengths @xmath86 is unimodal ( all local maxima are consecutive ) ; and no two consecutive edges have the same length , except possibly the maximal edge length that can occur at most twice . for any @xmath87 such that @xmath88 , consider @xmath89 . if @xmath90 and @xmath91 are at least as large , then the three edges form a crimp , by [ cond : a1 ]
however , this contradicts [ cond : a2 ] .
this proves unimodality , and that no three consecutive edges can have the same length .
in fact if @xmath89 is not maximal , one neighbor must be strictly smaller , to avoid the same contradiction .
compute the bar decomposition of @xmath5 and its image graph @xmath28 ( defined in section [ sec : preliminaries ] , see fig .
[ fig : bars ] ) . for every sober node of the image graph
, we perform the ws - equivalent node - expansion operation , described in @xcite[section 3 ] ( cortese et al .
@xcite call this a _
cluster expansion _ ) .
let @xmath33 be a sober node of the image graph .
let @xmath34 be the disk centered at @xmath33 with radius @xmath92 sufficiently small so that @xmath34 intersects only the segments incident to @xmath33 .
for each segment @xmath93 incident to @xmath33 , create a new node @xmath94 at the intersection point @xmath95 .
then modify @xmath5 by replacing each subpath @xmath96 $ ] passing through @xmath33 by @xmath97 $ ] ; see fig .
[ fig : node - expansion ] . if a node expansion produces an edge crossing , report that @xmath5 is not weakly simple .
( the vertices are perturbed for clarity ) .
new nodes are shown as squares . ]
chang et al .
@xcite[section 4 ] define a bar expansion operation . in this paper
, we will refer to it as old - bar - expansion . for a bar @xmath98 of the image graph ,
draw a long and narrow ellipse @xmath70 around the interior nodes of @xmath98 , create subdivision vertices at the intersection of @xmath99 with the edges , and replace each maximal path in @xmath70 by a straight - line edge .
if @xmath98 contains no spurs , old - bar - expansion is known to be ws - equivalent @xcite .
otherwise , it can produce false positives , hence it is not ws - equivalent ; see fig .
[ fig : normal - expansion ] for an example . * new bar expansion operation .
* let @xmath98 be a bar in the image graph with at least one interior node ; see fig .
[ fig : bar - expansion ] . without loss of generality , assume that @xmath98 is horizontal .
let @xmath70 be an ellipse whose major axis is in @xmath98 such that @xmath70 contains all interior nodes of @xmath98 ( nodes in @xmath98 except its endpoints ) , but does not contain any other node of the image graph and does not intersect any segment that is not incident to some node inside @xmath70 .
similar to old - bar - expansion , the operation new - bar - expansion introduces subdivision vertices on @xmath99 , however we keep all interior vertices of a bar at their original positions . in section
[ sec : bars ] , we apply a sequence of new operations to eliminate all vertices on @xmath98 sequentially while creating new nodes in the vicinity of @xmath70 .
our bar expansion operation can be considered as a preprocessing step for this subroutine . for each segment @xmath93 between a node @xmath100 and a node @xmath101 , create a new node @xmath94 at the intersection point @xmath102 and subdivide every edge @xmath103 $ ] to a path @xmath104 $ ] .
for each endpoint @xmath43 of @xmath98 , create two new nodes , @xmath105 and @xmath106 , as follows .
node @xmath43 is adjacent to a unique segment @xmath107 , where @xmath108 .
create a new node @xmath109 sufficiently close to the intersection point @xmath110 , but strictly above @xmath98 ; and create a new node @xmath106 in the interior of segment @xmath111 .
subdivide every edge @xmath112 $ ] , where @xmath113 , into a path @xmath114 $ ] .
since the new - bar - expansion operation consists of only subdivisions ( and slight perturbations of the edges passing through the end - segments of the bars ) , it is ws - equivalent . * crossings paths . *
apart from node - expansion and old - bar - expansion , none of our operations will create edge crossings .
our bar simplification algorithm ( section [ sec : bars ] ) will detect in some cases whether two subpaths cross .
crossings between overlapping paths is not easy to identify ( see @xcite[section 2 ] for a discussion ) .
we rely on the following simple condition to detect some ( but not all ) crossings .
[ lem : cross ] let @xmath5 be a weakly simple polygon parameterized by a curve @xmath115 ; and let @xmath116 be a closed jordan curve that does not pass through any vertices of @xmath5 and intersects every edge of @xmath5 transversely .
if @xmath117 and @xmath118 intersect in four distinct points , then the cyclic order of the intersection points along @xmath119 and @xmath120 is either the same or reverse .
if @xmath5 is weakly simple , then @xmath119 can be perturbed to a closed jordan curve @xmath121 with the same properties .
denote the intersection points by @xmath122 , @xmath123 , in counterclockwise order along @xmath120 . by the jordan curve theorem , @xmath124 has two connected components , the interior and the exterior of @xmath120 .
since @xmath121 and @xmath120 intersect transversely , the intersection points partition @xmath121 into four jordan arcs that lie alternately in the interior and exterior of @xmath120 .
the two jordan arcs of @xmath121 lying in the interior ( resp . , exterior ) of @xmath120 can not cross . without loss of generality ,
the two arcs of @xmath121 in the interior of @xmath120 connect the pairs @xmath125 and @xmath126 .
then the two arcs of @xmath121 in the interior of @xmath120 must connect the pairs @xmath127 and @xmath128 .
that is , the cyclic order of the intersection points along @xmath121 ( hence , along @xmath129 is also @xmath130 , as calimed .
specifically , a weakly simple polygon can not contain the certain configurations .
[ cor : forbidden ] a weakly simple polygon can not contain a pair of paths @xmath62 and @xmath62 such that 1 .
[ cross:1 ] @xmath131 $ ] and @xmath132 $ ] , where @xmath133 , @xmath134 , @xmath135 , and @xmath136 are nonoveralpping segments in this cyclic order about @xmath137 ( _ node crossing _
; see fig .
[ fig : forbidden](a ) ) .
[ cross:3 ] @xmath138 $ ] and @xmath139 $ ] , where @xmath140 , @xmath137 , @xmath141 , and @xmath142 are on a line in this order , and segments @xmath143 and @xmath144 lie in a closed halfplane bounded by this line ( fig . [ fig : forbidden](b ) ) .
[ cross:4 ] @xmath145 $ ] and @xmath146 $ ] where @xmath147 , and segments @xmath148 and @xmath149 lie in two different halfplanes bounded by line containing @xmath150 ( fig .
[ fig : forbidden](c ) ) . in all four cases ,
lemma [ lem : cross ] with a suitable jordan curve @xmath120 completes the proof . in case
[ cross:1 ] , let @xmath120 be a small circle around @xmath137 . in case [ cross:3 ] , let @xmath120 be a small neighborhood of segment @xmath150 . in case [ cross:4 ] ,
let @xmath120 be a small neighborhood of the convex hull of @xmath151 .
, @xmath137 , @xmath141 , and @xmath142 are collinear.,scaledwidth=75.0% ] * terminology . * here , we classify each path in @xmath70 .
all nodes @xmath152 lie either above or below @xmath98 .
we call them _ top _ and _ bottom _ nodes , respectively
. let @xmath153 denote the set of maximal paths @xmath154 $ ] in @xmath70 .
the paths in @xmath153 can be classified based on the position of their endpoints .
a path @xmath155 is called a * _ cross chain _
if @xmath156 and @xmath157 are top and bottom nodes respectively ; * _ top chain _
( resp . , _ bottom chain _ ) if both @xmath156 and @xmath157 are top nodes ( resp .
, bottom nodes ) ; * _ pin _ if @xmath158 $ ] ( note that every pin is a top or a bottom chain ) ; * _ v - chain _ if @xmath159 $ ] , where @xmath160 and @xmath155 is a top or a bottom chain .
let @xmath161 be the set of pins , and @xmath162 the set of v - chains .
let @xmath163 be the set of longest edges of _ cross chains _ in @xmath153 ( by lemma [ lem : irreducible ] , each cross chain contributes with one or two edges ) .
every weakly simple polygon has the following properties . 1 .
[ cond : a4 ] polygon @xmath5 has no node - crossings ( cf .
[ fig : forbidden](a ) ) .
[ cond : a3 ] no edge in @xmath163 lies in the interior of any other edge of @xmath5 ( cf .
[ fig : forbidden](b ) ) .
we can test properties [ cond : a3][cond : a4 ] in @xmath0 time at preprocessing : for each bar , sort all edges by their endpoints , and compute @xmath163 ; and for each node , temporarily compute a node - expansion .
if property [ cond : a3 ] or [ cond : a4 ] fails , we report that @xmath5 is not weakly simple .
similarly , the operations introduced in section [ sec : bars ] maintain properties [ cond : a1][cond : a4 ] for all maximal paths inside an elliptical disk @xmath70 . as a preprocessing for spur elimination ( section [ sec : tree - exp ] ) , we group all nodes that do not lie inside a bar into _ clusters_. after node - expansion and new - bar - expansion , all such nodes lie on a boundary of a disk ( circular or elliptical ) . for every sober node @xmath33
, we create @xmath164 clusters as follows .
refer to fig .
[ fig : newcluster ] .
the node expansion has replaced @xmath33 with new nodes on @xmath44 .
subdivide each segment in @xmath34 with two new nodes .
for each node @xmath165 , form a cluster @xmath166 that consists of @xmath43 and all adjacent ( subdivision ) nodes inside @xmath34 . for each node @xmath33 on the boundary of an elliptical disk @xmath70 , subdivide the unique edge outside @xmath70 incident to @xmath33 with a node @xmath167 .
form a cluster @xmath168 containing @xmath33 and @xmath167 .
every cluster will maintain the following invariants . 1 .
* cluster invariants . * for every cluster @xmath169 : 2 .
[ inv : tree ] @xmath169 induces a tree @xmath170 $ ] in the image graph rooted at @xmath33 .
[ inv : max - path]every maximal path of @xmath5 in @xmath169 is of one of the following two types : 1 .
both endpoints are at the root of @xmath170 $ ] and the path contains a single spur ; 2 .
one endpoint is at the root , the other is at a leaf , and the path contains no spurs .
additionally , each leaf node @xmath171 satisfies the following : 1 .
[ inv : deg ] @xmath171 has degree one or two in the image graph of @xmath5 ; 2 .
[ inv : no - spur ] there is no spur at @xmath171 ; 3 .
[ inv : subdiv ] no edge passes through @xmath171 ( i.e. , there is no edge @xmath57 $ ] such that @xmath172 but @xmath173 ) .
initially , every cluster trivially satisfies [ inv : tree ] and [ inv : max - path]a and every leaf node satisfies [ inv : deg][inv : subdiv ] since it originated from a subdivision .
* dummy vertices . *
although the operations described in sections [ sec : bars ] and [ sec : tree - exp ] introduce new nodes in the clusters , the image graph will always have @xmath7 nodes and segments .
a vertex at a cluster node is called _ benchmark _ if it is a spur or if it is at a leaf node ; and _ dummy _ otherwise .
paths traversing clusters may contain @xmath55 dummy vertices in the worst case , however we do not store these explicitly . by [ inv : tree ] ,
[ inv : max - path ] and [ inv : no - spur ] a maximal path in a cluster can be uniquely encoded by one benchmark vertex : if it goes from a root to a spur at an interior node @xmath83 and back , we record only @xmath174 $ ] ; and if traverses @xmath170 $ ] from the root to a leaf @xmath171 , we record only @xmath175 $ ] .
in this section we introduce three new ws - equivalent operations and show that they can eliminate all vertices from each bar independently ( thus eliminating all forks ) .
the bar decomposition is pre - computed , and the bars remain fixed during this phase ( even though all edges along each bar are eliminated ) .
we give an overview of the overall effect of the operations ( section [ ssec : overview ] ) , define them and show that they are ws - equivalent ( sections [ ssec : primitives][ssec : operations ] ) , and then show how to use these operations to eliminate all vertices from a bar ( section [ ssec : phases ] ) .
after preprocessing in section [ sec : preprocess ] , we may assume that @xmath5 has no edge crossings , no node crossings , and satisfies [ cond : a1][cond : a3 ] .
we summarize the overall effect of the bar simplification subroutine for a given expanded bar . * changes in the image graph _
g_. * refer to fig .
[ fig : overview - g ] . all nodes in the interior of the ellipse @xmath70
are eliminated .
some spurs on @xmath98 are moved to new nodes in the clusters along @xmath99 .
segments inside @xmath70 connect two leaves of trees induced by clusters . * changes in the polygon _
p_. * refer to fig .
[ fig : overview - p ] . consider a maximal path @xmath155 in @xmath5 that lies in @xmath70 .
the bar simplification will replace @xmath176 $ ] with a new path @xmath177 . by [ inv : no - spur]-[inv : subdiv ] , only nodes @xmath33 and @xmath43 in @xmath155 lie on @xmath99 .
if @xmath155 is the concatenation of @xmath178 and @xmath179 , then @xmath177 will be a spur in the cluster containing @xmath33 ( fig .
[ fig : overview - p ] ( a ) ) .
if @xmath155 has no such decomposition , but its two endpoints are at the same node , @xmath180 , then @xmath177 will be a single edge connecting two leaves in the cluster containing @xmath33 ( fig .
[ fig : overview - p ] ( b ) ) .
if the endpoints of @xmath155 are at two different nodes , @xmath177 is an edge between two leaves of the clusters containing @xmath33 and @xmath43 resp .
[ fig : overview - p ] ( c ) , ( d ) ) .
the operations in section [ ssec : operations ] rely on two basic steps , spur - reduction and node - split ( see fig .
[ fig : primitives ] ) . together with merge and subdivision , these operations are called _
primitives_. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * spur - reduction@xmath181 . *
assume satisfying this assumption was called a _ base _ of segment @xmath36 . ]
that every vertex at node @xmath33 has at least one incident edge @xmath182 $ ] .
replace any path @xmath183 $ ] , with a single - vertex path @xmath184 $ ] .
( see fig .
[ fig : primitives ] , left . ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * node - split@xmath185 . * assume segments @xmath36 and @xmath186 are consecutive in radial order around @xmath43 , and not collinear with any adjacent segment ; and @xmath5 contains no spurs of the form @xmath183 $ ] or @xmath187 $ ] .
create node @xmath188 in the interior of the wedge @xmath189 sufficiently close to @xmath43 ; and replace every path @xmath190 $ ] with @xmath191 $ ] .
( see fig . [
fig : primitives ] , right . )
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .
right : node - split@xmath185 . ]
the following two lemmas are generalization of the results in @xcite[section 5 ] .
[ lem : primitives1 ] operation spur - reduction is ws - equivalent .
let @xmath6 be obtained from applying spur - reduction@xmath181 to @xmath5 .
first suppose that @xmath5 is weakly simple .
then , there exists a simple polygon @xmath192 . by [ cond : a2 ]
, @xmath5 contains no path @xmath193 $ ] .
then , every path @xmath194)$ ] has endpoints @xmath195 and @xmath196 in @xmath197 .
replace each path of this form by a matching @xmath198 $ ] .
let @xmath199 be the resulting polygon ; refer to fig .
[ fig : primitives](left ) .
since @xmath46 is simple , the convex hull of any path @xmath194)$ ] can only contain another path of the form @xmath194)$ ] , so @xmath199 is simple . by construction ,
@xmath200 , and @xmath6 is weakly simple .
now suppose @xmath6 is weakly simple .
then there exists a simple polygon @xmath201 .
the maximal paths of @xmath199 in the disk @xmath34 form a matching @xmath202 .
let @xmath203 be the set of edges @xmath198 $ ] such that both endpoints are in @xmath204 .
we successively `` regrow '' the edges in @xmath205 into a spur in the corridor @xmath35 to obtain a weakly simply polygon @xmath192 .
the edges in @xmath206 each have an endpoint in @xmath207 , and they partition @xmath34 into _ faces _ that are each adjacent to @xmath35 .
therefore , there exists a face @xmath208 whose border intersects @xmath207 and contains an edge @xmath198\in m_u$ ] .
choose an arbitrary connected component of @xmath209 , create two parallel edges in @xmath35 that have one endpoint in this connected component , and replace @xmath198 $ ] with a simple path containing the new edges .
this is possible since the new edges appear sequentially in the linear order in @xmath35 . by construction
, the resulting polygon is simple . by repeating this step until all edges in @xmath205
are replaced by a spur in @xmath35 .
we obtain a simple polygon @xmath49 , hence @xmath5 is weakly simple .
[ lem : primitives2 ] operation node - split is ws - equivalent .
let @xmath6 be obtained from @xmath5 via node - split@xmath185 .
first assume that @xmath5 is weakly simple .
then there is a simple polygon @xmath192 .
since @xmath5 has no spurs @xmath183 $ ] or @xmath187 $ ] , every path @xmath210)$ ] has endpoints in @xmath204 and @xmath211 , resp .
; see fig .
[ fig : primitives](right ) .
since @xmath46 is simple , the endpoints of the paths of this form have matching linear orders along @xmath204 and @xmath211 .
replace the paths of this form @xmath210)$ ] by parallel paths via @xmath212 .
since @xmath46 is simple , the convex hull of any path @xmath210)$ ] can only contain another path of the form @xmath210)$ ] , so @xmath199 is simple . by construction ,
@xmath200 , and @xmath6 is weakly simple .
now , assume that @xmath6 is weakly simple . since the face in the image graph bounded by @xmath213 is empty
, we can change the embedding of the graph by bringing @xmath188 arbitrarily close to @xmath43 , maintaining weak simplicity .
let @xmath214 be the distance between @xmath188 and @xmath43 .
let @xmath201 be a simple polygon defined on disks of radius @xmath4 .
then , @xmath199 is within @xmath215 frchet distance from @xmath5 and therefore @xmath5 is weakly simple . we describe three operations : pin - extraction , v - shortcut , and l - shortcut . in section [ ssec : phases ] , we show how to use them to eliminate spurs along any given bar @xmath98 .
the pin - extraction and v - shortcut operations will eliminate pins and v - chains .
chains in @xmath153 with two or more vertices in the interior of @xmath70 will be simplified incrementally , removing one vertex at a time , by the l - shortcut operation .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * pin - extraction@xmath181 . *
assumes that @xmath5 satisfies [ inv : tree][inv : subdiv ] and contains a pin @xmath216\in { { \mathcal pin}}$ ] . by [ inv : deg ] , node @xmath43 is adjacent to a unique node @xmath217 outside of @xmath70 . perform the following three primitives : ( 1 ) subdivision of every path @xmath218 $ ] into @xmath219 $ ] ; ( 2 ) spur - reduction@xmath220 .
( 3 ) spur - reduction@xmath221 .
see fig .
[ fig : pin - extraction ] for an example . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * v - shortcut@xmath222 .
* assumes that @xmath5 satisfies [ inv : tree][inv : subdiv ] and @xmath223\in { { \mathcal v}}$ ] .
furthermore , @xmath5 contains no pin of the form @xmath224 $ ] or @xmath225 $ ] , and no edge @xmath226 $ ] such that segment @xmath227 is in the interior of the wedge @xmath228 . by [ inv : deg ] , nodes @xmath229 and @xmath230 are each adjacent to unique nodes @xmath231 and @xmath232 outside of @xmath70 , respectively .
the operation executes the following primitives sequentially : ( 1 ) node - split@xmath222 , which creates @xmath167 ; ( 2 ) node - split@xmath233 and node - split@xmath234 ; which create @xmath235 , respectively ; ( 3 ) merge every path @xmath236 $ ] to @xmath237 $ ] .
see fig .
[ fig : shortcut ] for an example . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ lem : pin - and - v ] pin - extraction and v - shortcut are ws - equivalent and maintain [ inv : tree][inv : subdiv ] . * pin - extraction . * by construction , the operation maintains [ inv : tree][inv : subdiv ] . also , [ inv : deg][inv : subdiv ] ensure that spur - reduction@xmath220 in step ( 2 ) satisfies its preconditions .
consequently , all three primitives are ws - equivalent
. * v - shortcut .
* by construction , the operation maintains [ inv : tree][inv : subdiv ] .
the first two primitives are ws - equivalent by lemma [ lem : primitives2 ] .
the third step is ws - equivalent because triangle @xmath238 is empty of nodes and segments , by assumption . *
l - shortcut operation . *
the purpose of this operation is to eliminate a vertex of a path that has an edge along a given bar . before describing the operation
, we introduce some notation ; refer to fig .
[ fig : classify ] . for a node @xmath239 ,
let @xmath240 be the set of paths @xmath241 $ ] in @xmath5 such that @xmath242 .
each path in @xmath153 is either in @xmath243 , in @xmath244 or has two subpaths in some @xmath240 .
recall that @xmath163 is the set of longest edges of cross chains in @xmath153 .
denote by @xmath245 the set of paths @xmath241 $ ] , where @xmath145 $ ] is _ not _ in @xmath163 . , @xmath244 , @xmath246 , @xmath247 , @xmath248 , and @xmath249 .
] we partition @xmath240 into four subsets ( refer to fig . [
fig : classify ] ) : a path @xmath241\in l_v$ ] is in 1 .
@xmath246 ( _ top - right _ ) if @xmath43 is a _
vertex and @xmath250 ; 2 .
@xmath247 ( _ top - left _ ) if @xmath43 is a _
vertex and @xmath251 ; 3 .
@xmath248 ( _ bottom - right _ ) if @xmath43 is a _
vertex and @xmath250 ; 4 .
@xmath249 ( _ bottom - left _ ) if @xmath43 is a _ bottom _
vertex and @xmath250 .
we partition @xmath252 into four subsets analogously .
we define the operation l - shortcut for paths in @xmath246 ; the definition for the other subsets can be obtained by suitable reflections . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * l - shortcut@xmath253 . *
assume that @xmath5 satisfies [ inv : tree][inv : subdiv ] , @xmath239 and @xmath254 .
by [ inv : deg ] , @xmath43 is adjacent to a unique node @xmath255 and to a unique node @xmath256 .
let @xmath257 denote the set of all nodes @xmath137 for which @xmath241\in l_v^{tr}$ ] .
let @xmath258 and @xmath259 be the leftmost and rightmost node in @xmath257 , respectively .
further assume that @xmath5 satisfies : 1 .
[ cond : b1 ] no pins of the form @xmath260 $ ] ; 2 .
[ cond : b2 ] no edge @xmath261 $ ] such that segment @xmath262 is in the interior of the wedge @xmath263 ; 3 .
[ cond : b3 ] no edge @xmath264 $ ] such that @xmath265 is a top vertex and @xmath266 , @xmath267 . do the following ( see fig .
[ fig : spur - simplification ] for an example ) . 1 .
create a new node @xmath268 to the right of @xmath43 sufficiently close to @xmath43 .
[ phase1 ] for every path @xmath241\in l_v^{tr}$ ] where @xmath150 is the _ only _ longest edge of a cross chain , create a crimp by replacing @xmath145 $ ] with @xmath269 $ ] .
[ phase2 ] replace every path @xmath270 $ ] by @xmath271 $ ] .
[ phase3 ] replace every path @xmath272 $ ] , where @xmath273 , by @xmath274 $ ] . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ see fig . [
fig : spur - simplification - diff ] for an explanation of why l - shortcut requires conditions [ cond : b2][cond : b3 ] and phase [ phase1 ] of the operation .
if we omit any of these conditions , l - shortcut would not be ws - equivalent .
has a pin @xmath260 $ ] not satisfying [ cond : b2 ] .
top right : @xmath5 does not satisfy [ cond : b3 ] .
bottom : the operation skips phase [ phase1 ] . ]
[ rem : sequence]note that l - shortcut@xmath253 may create a pin @xmath275 $ ] or a v - chain @xmath276 $ ] , which can immediately be extracted and shortcut , respectively . if @xmath277 , then every path @xmath278\in l_v^{tr}$ ] , @xmath279 becomes @xmath280\in l_{v^*}^{tl}$ ] .
we can perform l - shortcut@xmath281 , since conditions ( s2)-(s3 ) are fulfilled .
consequently , a sequence of pin - extraction , v - shortcut , and l - shortcut operations eliminates _ all _ top - right paths in @xmath246 .
[ lem : spur - simplification ] l - shortcut is ws - equivalent and it maintains [ inv : tree][inv : subdiv ] .
let @xmath62 be the polygon obtained from @xmath5 after phase [ phase1 ] of l - shortcut@xmath253 .
similarly , let @xmath63 and @xmath282 be the polygons obtained after phase [ phase2 ] and [ phase3 ] , respectively .
note that phase [ phase1 ] of the operation only creates crimps , and it is ws - equivalent by @xcite .
let @xmath283 be the set of edges @xmath145 $ ] where @xmath273 .
phases [ phase2][phase3 ] are equivalent to the concatenation of the primitives : subdivision , node - split , and merge .
specifically , phase [ phase2 ] is equivalent to subdividing every edge in @xmath283 into @xmath284 $ ] , where @xmath285 , and applying node - split@xmath286 ( which creates @xmath287 ) to @xmath63 .
phase [ phase3 ] consists of node - split@xmath288 ( which creates @xmath188 ) , and merging every path @xmath289 $ ] to @xmath290 $ ] .
the only primitive that may not satisfy its preconditions is node - split@xmath286 : segment @xmath291 may be collinear with several segments of @xmath98 , and @xmath63 may contain spurs that overlap with @xmath291 . we show that the spurs that may overlap with @xmath291 do not pose a problem , and we can essentially repeat the proof of lemma [ lem : primitives2 ] .
it remains to show that @xmath62 is weakly simple iff @xmath63 is weakly simple .
assume that @xmath62 is weakly simple and @xmath292 .
the above - below relationship among edges that contain segment @xmath291 determines a linear order .
we claim that there exists a simple polygon @xmath293 such that every edge in @xmath283 lies above all edges that are not in @xmath283 and overlap with @xmath291 .
let @xmath68 be an arbitrary simple polygon in @xmath294 .
let @xmath295 be the set of edges that are not in @xmath283 but lie above some edge in @xmath283 . due to [ cond : b2 ] and [ cond : b3 ] , @xmath295 can only contain edges of the form @xmath296 $ ] where @xmath297 , otherwise @xmath68 would not be simple .
let @xmath298 $ ] , @xmath299 , be the lowest edge in @xmath98 that is adjacent to some edge in @xmath283 ( shown in blue in fig .
[ fig : reshuffling ] ) . due to
[ cond : a1 ] , the right endpoint @xmath142 is incident to an edge in @xmath283 , and we have @xmath300 .
we have @xmath301 either by property [ cond : a2 ] or by a crimp introduced in phase [ phase1 ] .
subdivide every edge @xmath296\in z$ ] into a path @xmath302 $ ] if @xmath303 .
leave the right edge of the path , @xmath304 $ ] , in the original position ; and reroute the left edge @xmath305 $ ] to a new position below @xmath306 , maintaining the linear order among these edges .
we obtain a polygon @xmath293 .
we need to show that this modification does not introduce crossings in the perfect matchings in the disks @xmath34 .
the modification affects matching only in the nodes @xmath307 ( inclusive ) . in the disk at @xmath308
, we modify the matching such that the new edges do not cross any surviving edges by the choice of @xmath306 . in the disk at all the other affected nodes , the new edges are immediately below @xmath306 in the original linear order .
therefore , @xmath309 is a simple polygon , which proves our claim .
we can proceed as in the proof of lemma [ lem : primitives2 ] to show that @xmath63 is a weakly simple iff @xmath62 is weakly simple , that is , phase [ phase2 ] is ws - equivalent .
note that the intermediate polygons , @xmath62 and @xmath63 , may violate conditions [ cond : a1 ] and [ cond : a2 ] , since phase [ phase1 ] introduces crimps and phase [ phase2 ] subdivides edges .
however , after phase [ phase3 ] , conditions [ cond : a1 ] and [ cond : a2 ] are restored , and operation l - shortcut maintains [ cond : a1][cond : a3 ] in the ellipse @xmath70 .
is weakly simple , we can change the linear order of the edges as shown . ] in this section , we show that the three operations ( pin - extraction , v - shortcut , and l - shortcut ) can successively remove all spurs of the polygon @xmath5 from a bar @xmath98 .
algorithm bar - simplification@xmath310 . +
while @xmath5 has an edge along @xmath98 , perform one operation as follows . 1
. [ step - i ] if @xmath311 , pick an arbitrary pin @xmath216 $ ] and perform pin - extraction@xmath181 .
[ step - ii ] else if @xmath312 , then let @xmath223\in { { \mathcal v}}$ ] be a path where @xmath313 is minimal , and perform v - shortcut@xmath222 .
[ step - iii ] else if there exist @xmath239 such that @xmath314 , do : 1 .
[ step - iiia ] let @xmath43 be the rightmost node where @xmath254 .
[ step - iiib ] if @xmath315 for all @xmath109 , @xmath316 and @xmath317 , where @xmath318 is the unique neighbor of @xmath105 on @xmath98 , do l - shortcut@xmath253 .
[ step - iiic ] else let @xmath105 be the leftmost node such that @xmath316 and @xmath319 .
if @xmath320 satisfies [ cond : b3 ] do l - shortcut@xmath321 , otherwise halt and report that @xmath5 is not weakly simple .
[ step - iv ] else if there exist @xmath239 such that @xmath322 , repeat steps [ step - iiia][step - iiic ] with left
right and @xmath323@xmath324 interchanged .
( notice the use of @xmath240 instead of @xmath252 .
same applies to ( vi ) ) .
[ step - v ] else if there exist @xmath239 such that @xmath325 , repeat steps [ step - iiia][step - iiic ] using @xmath326 and @xmath327 in place of @xmath323 and @xmath324 , resp . , and
left - right interchanged .
[ step - vi ] else if there exist @xmath239 such that @xmath328 , repeat steps [ step - iiia][step - iiic ] using @xmath327 and @xmath326 in place of @xmath323 and @xmath324 , resp .
after the loop ends , perform old - bar - expansion ( cf . section [ ssec : bar - expansion ] ) in the ellipse @xmath70 ; + return @xmath5 ( end of algorithm ) .
+ informally , bar - simplification `` unwinds '' each polygonal chain in the bar , while extracting pins and v - chains as they appear , by alternating between steps to ( see fig .
[ fig : cross - lifecycle ] ) . note that step uses @xmath329 ( instead of @xmath330 ) to avoid an infinite loop . , [ step - iii ] , [ step - iv ] , and [ step - vi ] . ]
[ lem : simplfication - existance ] the operations performed by bar - simplification@xmath310 are ws - equivalent , and maintain properties [ cond : a1][cond : a3 ] and [ inv : tree][inv : subdiv ] inside @xmath70 .
the algorithm either removes all nodes from the ellipse @xmath70 , or reports that @xmath5 is not weakly simple .
the l - shortcut operations performed by the algorithm create at most two crimps in each cross - chain in @xmath153 .
we show that the algorithm only uses operations that satisfy their preconditions , and reports that @xmath5 is not weakly simple only when @xmath5 contains a forbidden configuration . * steps [ step - i][step - ii ] . *
since every pin can be extracted from a polygon satisfying [ inv : tree][inv : subdiv ] , we may assume that @xmath331 .
suppose that @xmath312 .
let @xmath223\in { { \mathcal v}}$ ] be a v - chain such that @xmath313 is minimal .
since @xmath331 , the only obstacle for the precondition is an edge @xmath226 $ ] such that segment @xmath227 is in the interior of the wedge @xmath228 ( or else the image graph would have a crossing ) .
this edge is part of a path @xmath332 $ ] .
node @xmath333 must be on @xmath99 between @xmath229 and @xmath230 , otherwise there would be a node - crossing at @xmath33 , contrary to [ cond : a4 ] .
however , @xmath334 , otherwise @xmath332 $ ] would be a pin .
consequently , @xmath332 $ ] is a v - chain where @xmath335 , contrary to the choice of @xmath223\in { { \mathcal v}}$ ] .
this confirms that v - shortcut@xmath222 satisfies all preconditions .
henceforth , assume that @xmath331 and @xmath336 .
* step [ step - iii][step - iv ] .
* by symmetry , we consider only step [ step - iii ] .
we distinguish between two cases .
* case 1 : the conditions of [ step - iiib ] are satisfied .
* we need to show that l - shortcut@xmath253 satisfies [ cond : b1][cond : b3 ] .
since @xmath331 , condition [ cond : b1 ] is met .
suppose there is an edge @xmath261 $ ] such that segment @xmath262 is in the interior of the wedge @xmath337 . clearly , @xmath265 is a top node
. then edge @xmath261 $ ] is part of a path @xmath338 $ ] .
however , @xmath333 must be in the closed wedge @xmath339 otherwise there would be a node - crossing at @xmath140 , contrary to [ cond : a4 ] .
also , @xmath333 can not be a top vertex on @xmath99 since @xmath340 , and @xmath333 can not be on @xmath98 by the choice of node @xmath43 .
this confirms [ cond : b2 ] .
we argue similarly for [ cond : b3 ] .
suppose there is an edge @xmath264 $ ] such that @xmath265 is a top vertex and @xmath266 , @xmath267 .
this edge is part of a path @xmath341 $ ] .
node @xmath342 must be on or above @xmath98 , otherwise there would be a node - crossing at @xmath333 .
it can not be a top vertex , since @xmath340 .
it can not be to the left of @xmath333 , otherwise the conditions of [ step - iiib ] are satisfied ; and it can not be to the right of @xmath333 , otherwise @xmath343 with @xmath344 , contrary to the choice of @xmath43 .
this confirms that l - shortcut@xmath253 satisfies [ cond : b2]-[cond : b3 ] and can be performed .
* case 2 : the conditions of [ step - iiib ] are not satisfied .
* let the path @xmath345\in l_{v'}^{tl}$ ] be selected in l - shortcut@xmath321 by the algorithm .
conditions [ cond : b1][cond : b2 ] are satisfied similar to case 1 .
if [ cond : b3 ] fails , there is an edge @xmath264 $ ] such that @xmath265 is a top vertex and @xmath266 , @xmath346 ( recall that left and right are interchanged in @xmath347 ) .
edge @xmath264 $ ] is part of a path @xmath341 $ ] , where @xmath348 , similar to case 1 .
this implies @xmath341\in l_p^{tr}\cup l_p^{tl}$ ] .
if @xmath349 , then either @xmath343 , which contradicts the choice of @xmath43 , or @xmath350 , which contradicts the choice of @xmath105 . consequently , @xmath351 .
this implies @xmath352 , so the paths @xmath353 $ ] and @xmath354 $ ] form a forbidden configuration described in corollary [ cor : forbidden](2 ) ; see also fig .
[ fig : forbidden](b )
. therefore the algorithm correctly finds that @xmath5 is not weakly simple . * steps [ step - v][step - vi ] . * if steps [ step - i][step - iv ] do not apply , then @xmath355 .
that is , for every path @xmath241\in l^{tr}$ ] , we have @xmath145\in m_{cr}$ ] . in particular , there are no top chains .
the operations in [ step - v][step - vi ] do not change these properties .
consequently , once steps [ step - v][step - vi ] are executed for the first time , steps [ step - iii][step - iv ] are never executed again . by a symmetric argument , steps [ step - v][step - vi ] eliminate all paths in @xmath356 . if the while loop terminates , every edge in @xmath98 is necessarily in @xmath163 and @xmath357 .
consequently , by lemma [ lem : irreducible ] , @xmath98 contains no spurs and old - bar - expansion is ws - equivalent .
this operation eliminates all nodes in the interior of @xmath70 . *
* each pin - extraction and v - shortcut operation reduces the number of vertices of @xmath5 within @xmath70 .
operation l - shortcut@xmath358 , @xmath359 , either reduces the number of interior vertices , or produces a crimp if edge @xmath145 $ ] is a longest edge of a cross - chain . for termination , is enough to show that , for each cross - chain @xmath360 , the algorithm introduces a crimp at most once in steps [ step - iii][step - iv ] , and at most once in steps [ step - v][step - vi ] . without loss of generality ,
consider step [ step - iii ] .
we apply an l - shortcut in two possible cases .
we show that it does not introduce crimps in case 2 . in step [ step - iiic ]
, we only perform l - shortcut@xmath321 if [ cond : b3 ] is satisfied and @xmath317 .
so for all @xmath361\in l_{v'}^{tl}$ ] , we have @xmath362 .
suppose , for contradiction , that @xmath363 $ ] is the only longest edge of some cross chain ( and hence l - shortcut would introduce a crimp ) .
then , @xmath363\in m_{cr}$ ] is inside @xmath364 $ ] , contradicting [ cond : a3 ] .
consider case 1 .
notice that l - shortcut@xmath253 is executed only if there exists a top node @xmath155 with @xmath365 such that @xmath366 .
suppose that l - shortcut@xmath253 introduces a crimp in the path @xmath241\in l_v^{tr}$ ] .
this operation removes this subpath of a cross chain from @xmath246 , but introduces @xmath367 $ ] into @xmath368 . by the time the algorithm executes l - shortcut@xmath369 , we know that for every top vertex @xmath155 with @xmath365 , @xmath370
this implies that , after l - shortcut@xmath369 is performed , although a path @xmath371 $ ] is introduced in @xmath372 , operation l - shortcut@xmath373 can never be performed .
the same arguments apply to steps [ step - v][step - vi ] .
[ lem : simplfication - time ] algorithm * bar - simplification*@xmath310 takes @xmath374 time , where @xmath375 is the number of vertices in @xmath98 .
operations pin - extraction , v - shortcut , and l - shortcut each make @xmath376 changes in the image graph .
operations pin - extraction and v - shortcut decrease the number of vertices inside @xmath70 .
each l - shortcut does as well , but they may jointly create @xmath377 crimps , by lemma [ lem : spur - simplification ] .
so the total number of operations is @xmath378 .
when @xmath241\in l_v^{tr}$ ] and @xmath285 , l - shortcut replaces @xmath241 $ ] by @xmath379 $ ] : vertex @xmath380 $ ] shifts to @xmath381 $ ] , but no vertex is eliminated . in the worst case ,
one l - shortcut modifies @xmath382 paths , so in @xmath382 operations the total number of vertex shifts is @xmath383 .
our implementation does not maintain the paths in @xmath153 explicitly .
instead , we use set operations . we maintain the sets @xmath243 , @xmath244 , and @xmath384 , with @xmath239 and @xmath385 , in sorted lists .
the pins @xmath216\in { { \mathcal pin}}$ ] are sorted by @xmath386 ; the wedges @xmath223\in { { \mathcal v}}$ ] are sorted by @xmath313 . in every set @xmath384 ,
the first two nodes in the paths @xmath241\in l_v^x$ ] are the same by [ inv : no - spur ] , and so it is enough to store vertex @xmath381 $ ] ; these vertices are stored in a list sorted by @xmath387 .
we also maintain binary variables to indicate for each path @xmath241\in l_v^x$ ] whether it is part of a cross chain , and whether @xmath145 $ ] is the only longest edge of that chain .
steps [ step - i]-[step - ii ] remove pins and v - chains , taking linear time in the number of removed vertices , without introducing any path in any set . consider l - shortcut@xmath253 , executed in one of steps [ step - iii][step - iv ] which can be generalized to other occurrences of the l - shortcut operation .
the elements @xmath388\in l_v^{tr}$ ] are simplified to @xmath290 $ ] . for each of these paths , say that the next edge along @xmath5 is @xmath389 $ ] .
then , the paths @xmath390 $ ] are inserted into either @xmath391 , if @xmath392 is a top vertex , or @xmath368 if @xmath393 .
we can find each chain @xmath388\in l_v^{tr}$ ] , in @xmath376 time , since @xmath246 is sorted by @xmath387 .
finally , all other paths @xmath241\in l_v^{tr}$ ] , where @xmath285 , become @xmath379 $ ] and they form the new set @xmath394 . since we store only the last vertex @xmath381 $ ] , which is unchanged , we create @xmath394 at no cost .
this representation allows the manipulation of @xmath378 vertices with one set operation .
the number of insert and delete operations in the sorted lists is proportional to the number of vertices that are removed from the interior of @xmath70 , which is @xmath378 .
each insertion and deletion takes @xmath395 time , and the overall time complexity is @xmath374 .
after bar - simplification ( section [ sec : bars ] ) , we obtain a polygon that has no forks and every spur is at an interior node of some cluster ( formed on the boundary of some ellipse @xmath396 . in the absence of forks ,
we can decide weak simplicity using @xcite[theorem 5.1 ] , but a nave implementation runs in @xmath2 time : successive applications of spur - reductions would perform an operation at each dummy vertex . in this section ,
we show how to eliminate spurs in @xmath0 time .
* formation of groups .
* we create _ groups _ by gluing pairs of clusters with adjacent roots together .
in the course of our algorithm , an analogue of the pin - extraction operation will extract a spur from one group into an `` adjacent '' group .
this requires a well - defined adjacency relation between groups .
recall that by [ inv : tree ] each cluster induces a tree .
we also modify the image graph , transforming each tree into a binary tree using ws - equivalent primitives . for each node @xmath83 with more than two children , let @xmath397 and @xmath398 be the first two children in counterclockwise order .
create new nodes @xmath399 and @xmath400 by subdivision in @xmath401 and @xmath402 , respectively , and create a segment @xmath403 . use the inverse of node - split to merge nodes @xmath399 and
@xmath400 into a node @xmath404 , reducing the number of children of @xmath83 by one . by construction , if a segment @xmath36 connects nodes in different clusters , both @xmath33 and @xmath43 are leaves or both are root nodes . for
every pair of clusters , @xmath169 and @xmath166 , with adjacent roots , @xmath33 and @xmath43 , create a _ group _ @xmath405 ; see fig [ fig : groups ] . by construction ,
the groups are pairwise disjoint .
two groups are called _ adjacent _ if they have two adjacent leaves in the image graph . , containing clusters @xmath169 and @xmath166 .
leaf nodes are shown as black dots . ]
recall that a maximal path in each cluster is represented by benchmark vertices ( leaves and spurs ) .
we denote by @xmath406 $ ] ( using semicolons ) a maximal path inside a group defined by the benchmark vertices @xmath407 . for
a given group @xmath408 , let @xmath153 denote the set of maximal paths with vertices in @xmath408 ; and let @xmath409 be the set of subpaths in @xmath153 between consecutive benchmark vertices .
[ rem : groups ] by invariants [ inv : tree ] , [ inv : max - path ] and [ inv : no - spur ] , a path in @xmath153 of a group @xmath408 has alternating benchmark vertices between @xmath169 and @xmath166 .
consequently , every path in @xmath409 has one endpoint in @xmath169 and one in @xmath166 , and each spur in @xmath408 is incident to two paths in @xmath409 . * spur - elimination algorithm . *
assume that @xmath410 is a partition of the nodes of the image graph into groups satisfying [ inv : tree][inv : subdiv ] .
we consider one group at a time , and eliminate all spurs from one cluster of that group .
when we process one group , we may split it into two groups , create a new group , or create a new spur in an adjacent group ( similar to pin - extraction in section [ sec : bars ] ) .
the latter operation implies that we may need to process a group several times .
termination is established by showing that each operation reduces the total number of benchmark vertices ( i.e. , spurs and boundary vertices ) .
initially , the number of benchmarks is @xmath7 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ algorithm spur - elimination@xmath411 . + while @xmath5 contains a spur , do : 1 .
choose a group @xmath412 that contains a spur , w.l.o.g .
contained in cluster @xmath169 and compute its supporting data structures .
while @xmath170 $ ] contains an interior node , do : 1 . [
step : merge ] if @xmath33 contains no spurs and is incident to only two edges @xmath36 and @xmath413 , eliminate @xmath33 with a merge operation .
the node @xmath217 becomes the new root of the tree @xmath170 $ ] .
[ step : low ] if @xmath33 contains spurs , eliminate them as described in section [ ssec : low ] .
[ step : split ] if @xmath33 contains no spurs , split @xmath408 into two groups along a chain of segments that contains @xmath36 as described in section [ ssec : split ] . rename a largest resulting group to @xmath408 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the detailed description of steps [ step : low ] and [ step : split ] are in sections [ ssec : low ] and [ ssec : split ] , respectively .
we first present supporting data structures in sections [ ssec : data ] , and then analyze the algorithm in section [ ssec : time ] .
consider a group @xmath408 composed of two trees @xmath170 $ ] and @xmath414 $ ] rooted at @xmath33 and @xmath43 , respectively . recall that @xmath409 denotes the set of benchmark - to - benchmark paths , each with one benchmark in @xmath170 $ ] and one in @xmath414 $ ] . in the algorithm spur - elimination ,
we dynamically maintain the image trees @xmath170\cup t[v]$ ] , and the set of paths @xmath409 . in each group @xmath408
, we maintain only @xmath415 nodes that contain benchmark vertices or have degree higher than 2 .
dummy nodes of degree-2 that contain no benchmark vertices are redundant for the combinatorial representation , and will be eliminated with merge operations . however , a polyline formed by a chain of dummy nodes of degree-2 can not always be replaced by a straight - line segment ( this might introduce artificial crossings ) .
for this reason , we maintain the trees @xmath170 $ ] and @xmath414 $ ] as _ ordered trees _
, recording only the circular order of the segments around each node . the partition of a group into two groups will be driven by the partition of the paths in @xmath409 . for a set @xmath416 of benchmark - to - benchmark paths ,
we define a subtree @xmath417 induced by @xmath418 as follows .
let @xmath419 be the set of nodes that contain endpoints of some path in @xmath409 .
the tree @xmath417 is obtained in two steps : take the minimum subtree of @xmath170\cup t[v]$ ] that contains all nodes in @xmath420 , and then merging all nodes of degree two that are not in @xmath420 .
in particular , the nodes of @xmath417 include @xmath420 and the least common ancestor of any two nodes in @xmath421 and in @xmath422 , respectively .
denote by @xmath423 the _ lowest common ancestor _ of nodes @xmath342 and @xmath83 in @xmath170 $ ] ( resp .
, @xmath414 $ ] ) . for the image graph of @xmath408 , we maintain the following data structures .
* we store trees @xmath170 $ ] and @xmath414 $ ] each using the dynamic data structure of @xcite , which supports @xmath376-time insertion and deletion of leaves , merging interior nodes of degree 2 , subdivision of edges , and lowest common ancestor queries . * for purposes of discussion ,
imagine that @xmath408 is inside an axis - aligned rectangle with the leaves of @xmath170 $ ] along the top edge and leaves of @xmath414 $ ] along the bottom edge ( see fig . [
f : tree - cross - geom ] ) . for each tree
, we maintain a left - to - right euler tour in an order - maintenance data structure @xcite , which supports insertions immediately before or after an existing item , deletions , and precedence queries , each in @xmath376 amortized time .
for any node @xmath217 , let @xmath424 and @xmath425 respectively denote the first and last occurrences of @xmath217 in the euler tour .
we refer to the elements of the euler tour as _
tokens_. we write @xmath426 to denote that some token @xmath427 occurs before ( `` to the left of '' ) another token @xmath428 in their common euler tour . * we also maintain the cyclic list of all leaves of the tree @xmath170\cup t[v]$ ] ( in the order determined by the euler tour above ; note that we have @xmath429 for a leaf @xmath217 ) .
we now describe dynamic data structures for @xmath153 and @xmath409 . for every benchmark - to - benchmark path @xmath430\in { { \mathcal b}}$ ] , we assume that @xmath83 is in @xmath170 $ ] and @xmath431 is in @xmath414 $ ] .
a path @xmath430 $ ] is associated with the intervals @xmath432 $ ] and @xmath433 $ ] . for two consecutive benchmark - to - benchmark paths @xmath434 $ ] , where @xmath431 is in @xmath414 $ ]
, we define the interval @xmath435=[s_1^\flat , s_2^\flat]$ ] .
* the set of benchmark - to - benchmark paths @xmath430\in { { \mathcal b}}$ ] is stored in four lists , sorted by @xmath436 , @xmath437 , @xmath438 , and @xmath439 , respectively .
the sorted lists can be computed in @xmath415 time by an eulerian traversal of the tree . * for each node @xmath83 of @xmath170 $ ] , let @xmath440 denote the set of paths @xmath430 \in { { \mathcal b}}$ ] .
we store @xmath440 in two lists , sorted by @xmath438 and @xmath439 , respectively . * an _ interval tree _ for all intervals
@xmath435 $ ] that can report , for a query node @xmath333 , all intervals containing @xmath333 in @xmath441 time .
all data structures described in this section can be constructed in @xmath415 preprocessing time .
* crossing paths . *
the data structure described above can determine in @xmath376 time whether two paths in @xmath409 cross .
straightforward case analysis implies the following characterization of path crossings ( refer to fig .
[ f : tree - cross - geom ] ) .
[ lem : path - crossing ] let @xmath83 and @xmath33 be arbitrary nodes in tree @xmath442 , and let @xmath431 and @xmath43 be arbitrary nodes in @xmath443 .
paths @xmath430 $ ] and @xmath444 $ ] cross if and only if either ( 1 ) @xmath445 and @xmath446 , or ( 2 ) @xmath447 and @xmath448 . we describe step [ step : low ] of algorithm _ spur - elimination_. suppose that the root node @xmath33 contains a spur
. the following operation eliminates all spurs from @xmath33 , but the resulting cluster @xmath166 need not satisfy [ inv : max - path ] and [ inv : no - spur ] , and we need to perform other operations to restore these properties . refer to fig .
[ fig : spur - shortcut](a)(b ) for an example . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ spur - shortcut@xmath449 .
assume that @xmath408 satisfies invariants [ inv : tree][inv : subdiv ] , and @xmath33 contains a spur .
replace every path @xmath450 $ ] by @xmath451 $ ] .
let @xmath452 be the set of all such modified paths .
_ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ spur - shortcut is ws - equivalent and maintains properties [ inv : tree ] , [ inv : deg ] and [ inv : subdiv ] .
the operation is equivalent to a sequence of spur - reduction operations : first perform spur - reduction@xmath181 . in a bfs traversal of all nodes @xmath427 of @xmath414
$ ] , except for the root , perform spur - reduction@xmath453 .
all these operations satisfy spur - reduction s constraints .
initially , every path through the node @xmath427 has an edge in the segment @xmath454 , by [ inv : max - path ] .
the bfs traversal ensures that this property still holds when the algorithm performs spur - reduction@xmath453
. contain spurs .
( b ) after eliminating spurs , @xmath414 $ ] does not satisfy [ inv : max - path ] .
( c ) the analogues of pin - extraction and v - shortcut . ] note that for every path @xmath450 $ ] , both @xmath455 and @xmath456 are in @xmath414 $ ] ( cf . remark [ rem : groups ] ) and the path @xmath451 $ ] is uniquely defined by [ inv : tree ] .
however , the paths @xmath451\in { { \mathcal s}}$ ] violate [ inv : max - path ] or [ inv : no - spur ] .
we proceed with a sequence of `` repair '' steps to restore them , after which the total number of benchmark vertices decrease by at least @xmath457 .
the repair steps are fairly easy when @xmath455 and @xmath456 are in ancestor - descendent relation , that is , @xmath458 .
let @xmath459 denote the node in @xmath460 farther from the root .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for every path @xmath451\in { { \mathcal s}}$ ] , do 1 .
[ step : trivi1 ] if @xmath458 and @xmath461 , then replace @xmath451 $ ] with @xmath462 $ ] .
[ step : trivi2 ] if @xmath463 , and @xmath455 is not a leaf of @xmath414 $ ] , then replace @xmath451 $ ] with @xmath464 $ ] .
[ step : pin ] if @xmath463 , and @xmath455 is a leaf of @xmath414 $ ] , then do : by [ inv : deg ] , node @xmath455 is adjacent to a unique node @xmath465 . subdivide segment @xmath466 with a new node @xmath467 ( added to the cluster containing @xmath468 ) , subdivide every edge @xmath469 $ ] into @xmath470 $ ] , and then replace every path @xmath471 $ ] with @xmath472 $ ] .
see fig .
[ fig : spur - shortcut](b)(c ) for an example .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ these steps restore [ inv : max - path ] and [ inv : no - spur ] for the affected paths @xmath451\in { { \mathcal s}}$ ] , and are ws - equivalent : steps [ step : trivi1][step : trivi2 ] do not modify the polygon ( they change only the benchmarks ) ; and step [ step : pin ] is analogous to pin - extraction@xmath473 .
we are left with paths @xmath451\in { { \mathcal s}}$ ] where @xmath455 and @xmath456 are in different branches of @xmath414 $ ] . in this case , we perform an elaborate version of the v - shortcut operation , that creates a new group . for every node @xmath171 of @xmath414 $ ] , let @xmath474 be the set of paths @xmath451\in { { \mathcal s}}$ ] such that @xmath475 .
consider every node @xmath171 of @xmath414 $ ] where @xmath476 in a bottom - up traversal of @xmath414 $ ] ; and create a new group @xmath477 as follows ( refer to fig .
[ fig : spur - shortcut ] ) .
let @xmath478 ( resp . ,
@xmath479 ) be the set of nodes @xmath455 ( resp . , @xmath456 ) such that there is a path @xmath451\in { { \mathcal s}}_\ell$ ] , and @xmath455 is in the left ( resp .
, right ) subtree of @xmath171 .
let @xmath480 .
sort the nodes @xmath481 by @xmath482 , and let @xmath483 be the minimum node ; and similarly sort the nodes @xmath484 by @xmath485 , and let @xmath486 be the maximum node .
since all nodes in @xmath420 are descendants of @xmath171 in @xmath414 $ ] , we have @xmath487 for all @xmath481 and @xmath488 for all @xmath484 .
[ lem : group - crossing ] if there is a path @xmath430\in { { \mathcal b}}\setminus { { \mathcal s}}_\ell$ ] such that @xmath489 and @xmath490 , then it crosses some path in @xmath474 , hence @xmath5 is not weakly simple .
let @xmath491 be the path between @xmath483 and @xmath486 in @xmath414 $ ] . by the choice of @xmath171 ( in a bottom - up traversal of @xmath414 $ ] )
, we have @xmath492 for all descendants of @xmath171
. then @xmath430 $ ] reaches @xmath491 at some interior node @xmath493 , and then continues to @xmath171 and farther to @xmath494 . if @xmath495 is in a left ( resp .
, right ) subtree of @xmath171 , then @xmath430 $ ] crosses every path in @xmath474 that starts at @xmath483 ( resp .
, ends at @xmath496 .
we can find the set @xmath497 of nodes @xmath431 such that @xmath489 and @xmath490 in @xmath498 time , by lowest common ancestors queries from the leaves between @xmath483 and @xmath486 .
the algorithm halts and reports that the input polygon is not weakly simple if some node in @xmath497 have a path satisfying lemma [ lem : group - crossing ] .
we can now assume that @xmath499 .
the nodes in @xmath420 induce a binary tree , denoted @xmath500 $ ] , of size at most @xmath501 : its nodes are all nodes in @xmath420 and the lowest common ancestors of consecutive nodes in @xmath478 and @xmath479 respectively .
note that a segment of @xmath500 $ ] might not correspond to a segment @xmath414 $ ] ( see fig .
[ fig : spur - shortcut](b ) ) .
denote by @xmath502 the path between @xmath483 and @xmath486 in @xmath500 $ ] .
we now define the changes in the image graph .
every node @xmath503 is deleted from @xmath408 , and added to the new group .
create two nodes , @xmath504 and @xmath505 , in @xmath506 sufficiently close to @xmath171 in the wedge between the two children of @xmath171 , and connect them by a segment @xmath507 .
duplicate each node @xmath508 , by creating a node @xmath509 ( added to @xmath506 ) sufficiently close to @xmath431 , and add a segment @xmath510 .
subdivide every segment @xmath510 with two new boundary nodes , @xmath511 ( added to @xmath414 $ ] ) and @xmath512 ( added to @xmath506 ) .
the nodes @xmath431 or @xmath509 might now have degree 4 .
adjust the image graph so that the group trees are binary . finally partition the nodes in @xmath506 into two trees , @xmath513 $ ] and @xmath514 $ ] , rooted at @xmath515 and @xmath505 , respectively .
we now define the changes in the polygon .
replace every path @xmath516\in { { \mathcal s}}_\ell$ ] , where @xmath508 , by @xmath517 $ ] if it is adjacent to a path @xmath518\in { { \mathcal s}}_\ell$ ] ; and by @xmath519 $ ] otherwise .
now we can build @xmath418 as the set of benchmark - to - benchmark paths @xmath520 $ ] where @xmath521 . to prove ws - equivalence
, we consider the changes in the polygon . these amount to a sequence of ws - equivalent primitives : a node - split at @xmath171 , a sequence of node - splits along the chain @xmath491 from @xmath171 to @xmath483 and @xmath486 , respectively , and subdivision operations that create the new leaf nodes and merge operations at degree-2 nodes that no longer contain spurs .
the creation of new groups take @xmath522 time and @xmath522 paths in @xmath409 are removed or modified in @xmath408 .
thus the data structures for @xmath408 are updated in @xmath523 time .
overall , operation spur - shortcut@xmath449 and the repair steps that follow take @xmath524 time . in this section
we describe step [ step : split ] of algorithm spur - elimination@xmath411 .
assume that @xmath408 satisfies invariants [ inv : tree][inv : subdiv ] and @xmath33 contains no spur .
denote the left and right child of @xmath33 by @xmath525 and @xmath526 , respectively .
let @xmath527 , resp . , be the set of benchmark - to - benchmark paths that contain @xmath525 and @xmath526 .
we check whether any two paths from @xmath528 and @xmath529 , respectively , cross .
if there is such a crossing , we report that @xmath5 is not weakly simple , otherwise we split @xmath408 into two groups induced by @xmath528 and @xmath529 , respectively .
refer to fig .
[ fig : tree - splitting ] . without loss of generality
, we may assume @xmath530 .
it would be easy to compute the groups induced by @xmath528 and @xmath529 in @xmath415 time .
however , for an overall @xmath0-time algorithm , we can afford @xmath531 time for the split operation , and an additional @xmath441 time for each eliminated spur and each node that we spit into two nonempty nodes .
the group induced by @xmath532 can be computed from scratch in @xmath533 time , and we construct the group for @xmath529 by modifying @xmath408 , and updating the corresponding data structures . .
( a ) changes in the image graph .
( b ) changes in the polygon . ]
first , we find @xmath528 and @xmath529 .
compute @xmath528 using the list of paths @xmath430\in { { \mathcal b}}$ ] sorted by @xmath437 or @xmath436 .
since both lists naturally split into corresponding lists for @xmath528 and @xmath529 , we can split these lists in @xmath531 time . to construct the list of @xmath529 sorted by @xmath439 and @xmath438
, we start with the corresponding lists for @xmath409 , and delete all elements of @xmath528 in @xmath533 time . to compute the lists sorted by @xmath439 and @xmath438 for @xmath528
, we shall first compute the subtree @xmath534 $ ] induced by @xmath528 .
however , we can already find the maximum @xmath439 of a path @xmath430\in { { \mathcal b}}^-$ ] in @xmath533 time .
next , we test for crossings between the paths in @xmath528 and the paths in @xmath529 .
let @xmath535 be the maximum @xmath439 of a path @xmath430\in { { \mathcal b}}^-$ ] , and @xmath536 the minimum @xmath438 of a path @xmath430\in { { \mathcal b}}^+$ ] .
by lemma [ lem : path - crossing ] , there is such a crossing iff @xmath537 , which can be determined in @xmath376 time based on our sorted lists .
if a crossing is detected , the algorithm halts and reports that @xmath5 is not weakly simple .
the tree @xmath538 $ ] and @xmath539 $ ] are simple subtrees of @xmath170 $ ] ; but splitting @xmath414 $ ] is nontrivial .
we use the eulerian cycle of all leaves to find the right - most leaf @xmath540 in @xmath414 $ ] for which @xmath541 .
let @xmath542 $ ] .
we do not compute @xmath543 explicitly , as it may contain more than @xmath533 nodes , but we can test wether a query node @xmath431 of @xmath414 $ ] is in @xmath543 by checking whether @xmath544 in @xmath376 time .
since the paths in @xmath528 and @xmath529 do not cross , all nodes of @xmath534 $ ] are in or to the left of the chain @xmath543 , and all common nodes of @xmath534 $ ] and @xmath545 $ ] are in @xmath543 .
the image graph of @xmath534 $ ] can be computed from scratch using @xmath528 in @xmath546 time .
replace each node @xmath431 of @xmath534 $ ] that is in @xmath543 by a duplicate copy @xmath547 located sufficiently close to @xmath431 , to the right of @xmath431 .
the tree @xmath545 $ ] is computed from @xmath414 $ ] by node deletion and merge operations as follows . first deleting all nodes that are in
@xmath534 $ ] but not in @xmath543 . for every node @xmath431 of @xmath534 $ ] that lies in @xmath543 ,
if @xmath431 has degree-3 in @xmath534 $ ] and @xmath548 , then it is a degree-2 node in @xmath414 $ ] with no spurs , and so we can delete @xmath431 by merging its two incident segments . as a result ,
@xmath414 $ ] becomes a tree induced by @xmath529 .
it remains to resolve the connections between trees .
let @xmath549 denote the set of chains @xmath434 $ ] such that @xmath550\in { { \mathcal b}}^-$ ] and @xmath551\in { { \mathcal b}}^+$ ] .
the spurs at @xmath431 on all chains @xmath434\in { { \mathcal v}}^0 $ ] will be eliminated ( they will become adjacent leaves in the two resulting groups ) .
@xmath549 can be found with a query for @xmath33 in the interval tree .
let @xmath552 be the set of all nodes @xmath431 such that @xmath434\in{{\mathcal v}}^0 $ ] .
each node @xmath553 is in @xmath543 and , therefore , has a copy @xmath547 in @xmath534 $ ] .
create a segment between @xmath431 and @xmath547 , and subdivide the segment @xmath554 with two new nodes @xmath555 and @xmath511 in @xmath534 $ ] and @xmath414 $ ] , respectively .
the degree of nodes @xmath431 or @xmath547 might have increases to 4 ; and so we adjust the image graphs so that both trees are binary .
the image graph is now split into groups @xmath556 and @xmath408 .
we next define the changes in the polygon .
replace every chain @xmath434\in{{\mathcal v}}^0 $ ] with the chain @xmath557 $ ] , replacing also the corresponding paths in the lists @xmath528 and @xmath528 in @xmath558 time . in the sorted lists for @xmath528 and @xmath529 ,
this is done by deleting and reinsertions .
note that all leaves @xmath555 ( resp . , @xmath511 )
are at the end ( resp . , beginning ) of the euler tour of @xmath534 $ ] ( resp . ,
@xmath414 $ ] ) , the insertions take only @xmath559 time .
the changes in the polygon are equivalent to a sequence of ws - equivalent primitives : a node - split operation at @xmath33 , followed by a sequence of node - splits along the chain @xmath543 from @xmath540 to @xmath33 , and subdivision operations that create the new leaf nodes between the two groups
. the interval tree does not need to be updated since its query time remain @xmath441 .
consequently , we can split @xmath408 in @xmath560 time .
[ lem : group - simplfication - time ] given @xmath375 benchmark vertices , spur - elimination@xmath411 takes @xmath374 time .
let @xmath561 be the number of spurs , @xmath562 the number of benchmark vertices at the leaves of clusters , and let @xmath563 .
initially , @xmath564 by [ inv : tree ] .
all operations in spur - elimination monotonically decrease both @xmath561 and @xmath565 .
step [ step : low ] decreases @xmath565 by the number of spurs at @xmath33 , and steps [ step : merge ] and [ step : split ] both maintain @xmath565 . in particular , step [ step : split ] converts some spurs into pairs of adjacent benchmark vertices at leaves .
step [ step : merge ] removes an interior node of degree 2 in @xmath376 time .
interior nodes are created only when they contain a spur , so at most @xmath378 interior nodes are ever created , and steps [ step : merge ] altogether take @xmath378 time .
step [ step : merge ] takes @xmath566 to eliminate @xmath567 spurs . eventually , all spurs are eliminated , thus steps [ step : low ] altogether take @xmath374 time .
step [ step : split ] takes @xmath568 time . by a standard heavy - path decomposition argument
, the terms @xmath569 contributes @xmath374 time .
every chain in @xmath549 corresponds to a spur that is destroyed in a step [ step : split ] ( and no new spurs are created ) , therefore the terms @xmath570 sum to @xmath374 over the course of the algorithm .
since every occurrence of each step [ step : split ] increases the number of groups by one , the step is repeated @xmath378 times , and the @xmath571 terms sum to @xmath374 .
we presented an @xmath0-time algorithm for deciding whether a polygon with @xmath1 vertices is weakly simple .
there is a natural generalization for planar graphs @xcite[appendix d ] .
we can define the _ weak embedding _ for a graph @xmath283 in terms of frchet distance .
the frchet distance between two planar drawings @xmath572 of @xmath283 is defined as @xmath573 , where @xmath565 is an automorphism of @xmath283 .
it is an open problem to decide whether a drawing of a graph @xmath283 is weakly simple , i.e. , whether a drawing @xmath5 of @xmath283 is within @xmath4 frchet distance from an embedding @xmath46 of @xmath283 , for all @xmath3 .
we can also generalize the problem to higher dimensions .
a polyhedron can be described as a map @xmath574 .
a simple polyhedron is an injective function .
a polyhedron @xmath5 is weakly simple if there exists a simple polyhedron within @xmath4 frchet distance from @xmath5 for all @xmath3 .
this problem can be reduced to origami flat foldability .
the results of @xcite imply that , given a convex polygon @xmath5 and a piecewise isometric function @xmath575 ( called _ crease pattern _ ) , it is np - hard to decide if there exists an injective embedding of @xmath5 in three dimensions @xmath576 within @xmath4 frchet distance from @xmath577 for all @xmath3 , i.e. , if @xmath577 is _
flat foldable_. given @xmath5 and @xmath577 , we can construct a continuous function @xmath578 mapping each hemisphere of @xmath579 to @xmath5 ( @xmath580 , for a point @xmath581 , maps to two points in different hemispheres of @xmath579 ) .
then , the polyhedron @xmath582 is weakly simple if and only if @xmath577 is flat foldable .
therefore , it is also np - hard to decide whether a polyhedron is weakly simple .
research by akitaya and tth was supported in part by the nsf awards ccf-1422311 and ccf-1423615 .
research by erickson was supported in part by the nsf award ccf-1408763 .
we thank anika rounds and diane souvaine for many helpful conversations that contributed to the completion of this project .
zachary abel , erik d. demaine , martin l. demaine , david eppstein , anna lubiw , and ryuhei uehara , flat foldings of plane graphs with prescribed angles and edge lengths , in _ proc .
22nd graph drawing _ , lncs 8871 , springer , 2014 , pp . 272283 .
esther m. arkin , michael a. bender , erik d. demaine , martin l. demaine , joseph s.b .
mitchell , saurabh sethia , and steven s. skiena , when can you fold a map ?
, _ computational geometry : theory and applications _ * 29 * ( 2004 ) , 2346 .
michael a. bender , richard cole , erik d. demaine , martin farach - colton , and jack zito .
two simplified algorithms for maintaining order in a list , _ proc .
10th annual european symposium on algorithms _ , lncs 2461 , springer , 2002 , pp .
152164 .
daniel d. sleator and paul f. dietz , two algorithms for maintaining order in a list , in _ proc .
19th acm symposium on theory of computing _ , acm , 1987 , pp .
. full version in tech .
cmu - cs-88 - 113 , carnegie mellon university , 1988 . | we present an @xmath0-time algorithm that determines whether a given planar @xmath1-gon is weakly simple .
this improves upon an @xmath2-time algorithm by chang , erickson , and xu @xcite .
weakly simple polygons are required as input for several geometric algorithms . as such
, how to recognize simple or weakly simple polygons is a fundamental question . |
it is a well known fact in quantum field theory ( qft ) that any kind of unification of internal ( gauge ) symmetries with spacetime symmetries is not evident at all .
the celebrated coleman - mandula theorem and its various versions @xcite prohibit the most simple unification scenarios : under quite generic conditions one can not find a larger symmetry group for a qft model , which is composed as the semi - direct product of the poincar group and the internal symmetry group .
one of the important assumptions of the pertinent no - go theorem is that the lie group of internal symmetries is restricted to products of copies of @xmath0 group and compact semisimple lie groups .
this means that the total symmetries of matter fields at a fixed spacetime point or equivalently , at a fixed point of momentum space is a product of copies of @xmath0 group and a semisimple lie group .
our aim is to show a possible , physically natural mathematical example , when the total symmetry group of matter fields at a spacetime point or equivalently , at a fixed point of momentum space is some nontrivial extension of the ( conformal ) lorentz group , or more precisely , its covering group .
the pertinent extended group can be viewed as the automorphisms of the creation operator algebra for a spin 1/2 particle along with its antiparticle .
the non - semisimpleness of the discussed group gives possibility to nontrivially extend the group of spacetime symmetries .
the extended part becomes an idempotent normal subgroup , which can be regarded as `` dressing transformations '' of pure one - particle states in a formal qft context at a fixed momentum .
the presented algebraic construction may also help to improve mathematical formulation of qfts : it shows a possibility to avoid building the theory based on first constructing the one - particle theory and then constructing the corresponding multi - particle model , _ a posteriori_. that is because the `` dressing transformations '' mix the pure @xmath1-particle,@xmath2-antiparticle states with each - other to some extent , putting the possible particle combination states in a unified multiplet .
first we review the properties of the automorphism group of finite dimensional grassmann algebras as prototype problem @xcite .
physically , an @xmath3-generator grassmann algebra can be thought of as the algebra of creation operator polynomials in a formal qft at fixed momentum of a fermion particle with @xmath3 internal degrees of freedom .
( grassmann algebra , canonical generator system ) a finite dimensional complex associative algebra @xmath4 with unit is called a * grassmann algebra * if there exists a minimal generating system @xmath5 of @xmath4 such that @xmath6 such a minimal generating system shall be referred to as * canonical generator system*. we will now explore the basic properties of the group of automorphisms @xmath7 of @xmath4 .
these are the @xmath8 invertible complex - linear transformations , which preserve the algebraic product on @xmath4 .
the following basic properties are well known@xcite .
\(i ) grassmann algebras do exist : the exterior algebra @xmath9 of a finite dimensional complex vector space @xmath10 is a grassmann algebra , where a basis of @xmath10 are canonical generators . in fact , all grassmann algebras are ( not naturally ) isomorphic to some exterior algebra @xmath9 .
\(ii ) if @xmath5 is a canonical generator system and @xmath11 , then @xmath12 are also .
if , in addition , @xmath13 is an other system of canonical generators , then a mapping @xmath14 ( @xmath15 ) uniquely determines an automorphism @xmath11 .
that is , automorphisms can uniquely be characterized by their action on an arbitrarily chosen canonical generator system .
\(iii ) given a chosen system of canonical generators , the linear subspace of the pure @xmath16-th order polynomials of them are called the space of @xmath16-forms , and are denoted by @xmath17 .
the linear subspace of the pure even / odd polynomials of them are called the space of even / odd forms , and are denoted by @xmath18 / @xmath19 . as such , one has @xmath20 and @xmath21 .
these splittings of @xmath4 are referred to as @xmath22 and @xmath23-grading , respectively .
\(iv ) let the unity be denoted by @xmath24 and its complex linear span by @xmath25 .
let @xmath26 be the linear subspace of at least @xmath27-forms and observe that it is the maximal ideal of @xmath4 , and therefore is an @xmath7-invariant subspace .
because of that , we have the @xmath7-invariant splitting @xmath28 with corresponding @xmath7-invariant complementing projection operators @xmath29 and @xmath30 . because of @xmath7-invariance of unity
, @xmath31 can be written with uniquely determined @xmath32 , @xmath7-invariant map , picking out the scalar component .
\(v ) since @xmath33 is @xmath7-invariant , all its powers @xmath34 are @xmath7-invariant ( @xmath35 ) .
\(vi ) the center of @xmath4 , denoted by @xmath36 , consists of all elements commuting with @xmath4 , and they form an @xmath7-invariant subspace .
( d. z. djokovic @xcite ) let us define the following subgroups of @xmath7 , given a canonical generator system @xmath5 of @xmath4 .
\(i ) let @xmath37 be the @xmath22-grading preserving automorphisms .
these are of the form @xmath38 ( @xmath15 ) with @xmath39 .
\(ii ) let @xmath40 be the @xmath23-grading preserving automorphisms acting on @xmath41 factor space as unity .
these are of the form @xmath42 ( @xmath15 ) with @xmath43 .
\(iii ) let @xmath44 be the inner automorphisms , i.e. the ones of the form @xmath45 ( with some @xmath46 ) .
these are of the form @xmath47 $ ] ( @xmath15 ) with some @xmath46 . with these , the semi - direct product splitting @xmath48 holds . as a consequence , we have that the @xmath17 subspaces of @xmath16-forms are not @xmath7-invariant .
in fact , the list of indecomposable @xmath7-invariant subspaces are the followings : @xmath25 , @xmath49 ( @xmath35 ) , @xmath50 ( @xmath51 ) .
physicswise , this means that if we think of @xmath4 as the algebra of creation operator polynomials in a formal qft of an @xmath3 internal degrees of freedom fermion particle at fixed momentum , and we assume that the full @xmath7 acts on this algebra as symmetry group , then it becomes a unified multiplet . particularly , the only @xmath7-invariant decomposition is @xmath28 , i.e. the splitting to @xmath52-particle and to at - least-@xmath27-particle states .
the reason is that the normal subgroup @xmath53 mixes higher particle content to lower particle states .
this motivates to call @xmath54 the _ * dressing transformations * _ , being an idempotent normal subgroup of @xmath7 .
motivated by the above findings , we define a physically more relevant setting .
( @xmath55-algebra ) a finite dimensional complex associative algebra @xmath56 with unit shall be called a * @xmath55-algebra * if it is equipped with a conjugate - linear involution satisfying @xmath57 for all @xmath58 .
it is important that in the above definition the @xmath55-adjoint does not reverse the order of products , i.e. it is slightly different than that of a usual @xmath59-algebra .
this will , physically , model the charge conjugation in our construction .
( spin algebra ) a finite dimensional complex associative @xmath55-algebra @xmath56 with unit shall be called a * spin algebra * if there exists a minimal generating system @xmath60 of @xmath56 such that @xmath61 such a minimal generating system shall be referred to as * canonical generator system * , and the notation @xmath62 will occasionally be used .
physicswise , a spin algebra can be thought of as the algebra of creation operator polynomials in a formal qft of a @xmath63 internal degree of freedom fermion particle along with its antiparticle , at fixed momentum .
it is important to note that in this construction the annihilation operators of particles are not yet identified with the creation operator of antiparticles , and therefore a spin algebra is very different than that of a canonical anticommutation relation ( car ) algebra , also referred to as mixed exterior algebras@xcite . here , the @xmath55-adjoining only models the charge conjugation operator , exchanging particle and antiparticle creation operators in a conjugate - linear way .
we will now explore the basic properties of the group of automorphisms @xmath64 of @xmath56 .
these are the @xmath65 invertible complex - linear transformations , which preserve the algebraic product on @xmath56 along with the @xmath55-adjoining .
similar properties hold as for a grassmann algebra .
\(i ) spin algebras do exist : if @xmath66 is a complex @xmath63 dimensional vector space , called to be the cospinor space , and we take @xmath67 , then it naturally becomes a spin algebra .
in fact , all spin algebras are ( not naturally ) isomorphic to this .
( the notation @xmath68 means complex conjugation . )
\(ii ) if @xmath5 is a canonical generator system and @xmath69 , then @xmath12 are also .
if , in addition , @xmath13 is an other system of canonical generators , then a mapping @xmath14 ( @xmath15 ) uniquely determines an automorphism @xmath69 .
that is , automorphisms can uniquely be characterized by their action on an arbitrarily chosen canonical generator system .
\(iii ) given a chosen system of canonical generators , the linear subspace of the pure polynomials of @xmath1 pieces of @xmath70 and @xmath2 pieces of of @xmath71 are called the space of @xmath72-forms , and are denoted by @xmath73 .
the subspace of those polynomials for which @xmath74 holds are called the space of @xmath16-forms .
the linear subspace of the pure even / odd @xmath17-s are called the space of even / odd forms , and are denoted by @xmath18 / @xmath19 . as such , one has @xmath75 , @xmath76 and @xmath77 .
these splittings of @xmath56 are referred to as the @xmath78 , @xmath22 and @xmath23-grading , respectively .
\(iv ) let the unity be denoted by @xmath24 and its complex linear span by @xmath25 .
let @xmath26 be the linear subspace of at least @xmath27-forms and observe that it is the maximal ideal of @xmath56 , and therefore is an @xmath64-invariant subspace . because of that
, we have the @xmath64-invariant splitting @xmath79 with corresponding @xmath64-invariant complementing projection operators @xmath29 and @xmath30 . because of @xmath64-invariance of unity
, @xmath31 can be written with uniquely determined @xmath80 , @xmath64-invariant map .
\(v ) since @xmath33 is @xmath64-invariant , all its powers @xmath34 are @xmath64-invariant ( @xmath35 ) .
\(vi ) the center of @xmath56 , denoted by @xmath81 , consists of all elements commuting with @xmath56 , and they form an @xmath64-invariant subspace .
in fact , @xmath82 .
( a. lszl @xcite ) let us define the following subgroups of @xmath64 , given a canonical generator system @xmath83 of @xmath56 .
\(i ) let @xmath84 be the @xmath78-grading preserving automorphisms .
these are of the form @xmath85 and @xmath86 ( @xmath87 ) with @xmath88 . here , as before , @xmath68 denotes complex conjugation .
\(ii ) let @xmath89 be the two element group of automorphisms , @xmath90 being the identity and @xmath91 being the involutive complex - linear operator of particle - antiparticle label exchanging : @xmath92 , @xmath93 , @xmath94 , @xmath95 .
\(iii ) let @xmath96 be a subgroup of the @xmath23-grading preserving automorphisms defined by the relations @xmath42 and @xmath97 ( @xmath87 ) with uniquely determined parameters @xmath98 .
\(iv ) let @xmath99 be the inner automorphisms , i.e. the ones of the form @xmath45 ( with some @xmath100 ) .
these are of the form @xmath47+\frac{1}{2}\left[a,\left[a , e_{i}\right]]\right]$ ] ( @xmath15 ) with uniquely determined parameter @xmath101 . with these
, the semi - direct product splitting @xmath102 holds . the proof is based on the fact that @xmath64 elements are the linear transformations preserving the canonical generator relations . although the proof is theoretically not complicated , it is a quite extended calculation , and will be published in a more detailed paper @xcite . as a consequence , we have that the @xmath103 subspaces of @xmath72-forms are not @xmath64-invariant .
in fact , the list of indecomposable @xmath64-invariant subspaces are the followings : @xmath25 , @xmath49 ( @xmath104 ) , @xmath105 , @xmath106 , @xmath107 , @xmath108 .
these are illustrated in fig.[figspininvsubsp ] .
, @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] + , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] + , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] + , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath56 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent indecomposable @xmath64-invariant subspaces.,title="fig : " ] + physicswise , this result means that if we think of @xmath56 as the algebra of creation operator polynomials in a formal qft of a @xmath63 internal degrees of freedom fermion particle along with its antiparticle , at fixed momentum , and we assume that the full @xmath64 acts on this algebra as symmetry group , then it becomes a unified multiplet . particularly , the only @xmath64-invariant decomposition is @xmath79 , i.e. the splitting to @xmath52-particle and to at - least-@xmath27-particle states .
the reason is that the normal subgroup @xmath109 mixes higher particle content to lower particle states .
this motivates to call @xmath54 the _ * dressing transformations * _ , being an idempotent normal subgroup of @xmath64 . when dealing with the @xmath78-grading preserving part of @xmath64 , i.e. the action of the subgroup @xmath110 , a fixed @xmath78 grading may be used , and therefore the whole formalism can be thought of as an ordinary two - spinor calculus @xcite on the tensor algebra of cospinors and complex conjugate cospinors .
using this , it is immediately seen , that @xmath64 acts on the maximal forms @xmath111 as a scaling by a positive real number , i.e. they are orientation preserving . therefore , @xmath112 may be split to the cones of positive and negative maximal forms @xmath113 and @xmath114 , preserved by @xmath64 . in order to proceed ,
we study the transpose action of @xmath64 on the dual vector space @xmath115 of @xmath56 . from simple linear algebra , it automatically follows , that the indecomposable @xmath64-invariant subspaces of @xmath115 are : @xmath116 , @xmath117 , @xmath118 ( @xmath119 ) , @xmath120 , @xmath121 , @xmath122 , where @xmath123 denotes the annulator subspace in @xmath115 .
these are illustrated in fig.[figspindualinvsubsp ] . ,
@xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] + , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] + , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] , @xmath22 and @xmath23-graded structure of the dual @xmath115 of a spin algebra @xmath56 .
other panels : illustration of @xmath64-invariant indecomposable subspaces of @xmath115 .
one unit box depicts one complex dimension on all panels , shaded framed regions depict the pertinent @xmath64-invariant subspaces.,title="fig : " ] we shall show that @xmath64 acts as the conformal lorentz group on the four real dimensional @xmath64-invariant vector space @xmath124 . for this , we use the fact that a spin algebra may be equipped with a hopf algebra structure , given a chosen @xmath78-grading . as usual in the hopf algebra context@xcite , the unit is viewed as an @xmath125 linear map , and the algebraic product as an @xmath126{180}{$\triangle$}}}:\,a\otimes a\rightarrow a$ ] linear map .
the counit shall be an @xmath127 linear map , and we set @xmath128
. the antipode shall be an @xmath129 linear map , and we set @xmath130 .
the swapping involution shall be an @xmath131 linear map , and we set to @xmath132 ( for all @xmath58 ) .
the coproduct shall be a @xmath133 linear map , defined by the requirements @xmath134 , @xmath135 ( for all @xmath136 ) , and that @xmath137 is an @xmath138 algebra homomorphism , where @xmath139 is equipped with the skew - natural product @xmath140{180}{$\triangle$}}}\otimes{\mathbin{\rotatebox[origin = c]{180}{$\triangle$}}})\circ(i\otimes(j\circ\mathcal{i})\otimes i)$ ] , @xmath90 being the @xmath65 identity map and @xmath91 being the @xmath141 swapping map .
the holding of the hopf relations on @xmath142{180}{$\triangle$}}},\varepsilon,\triangle , s,\mathcal{i})$ ] may be checked by direct calculations . given the fixed spin algebra part @xmath142{180}{$\triangle$}}})$ ] it is seen that the coalgebra part @xmath143 is not conserved by all @xmath64 transformations , only under the @xmath22-grading preserving transformations @xmath144 : the dressing transformations @xmath54 deform the part @xmath145 to an other compatible coalgebra structure . equipped with the hopf algebra notions
, direct calculation shows that if we choose an @xmath146 positive maximal form element of @xmath56 , the bilinear map @xmath147 is a lorentz signature metric , where @xmath148 denotes the duality pairing form .
the action of @xmath64 preserves @xmath149 up to a scaling factor , i.e. it acts on @xmath150 as the conformal lorentz group .
the construction does not depend on the choice of the coproduct @xmath137 , i.e. does not depend on grading .
the choice of @xmath151 is unique up to a scaling factor , and thus @xmath149 is unique up to that as well .
[ remmetric ] next , we show that there are two charge - conjugate dirac bispinor spaces embedded ( not uniquely ) in @xmath56 , and thus the relation to the usual clifford algebra formulation becomes clear . given a coproduct @xmath137 , the four real dimensional @xmath64-invariant subspace @xmath152 may be embedded into @xmath153 using the rule @xmath154 ( for each @xmath155 ) , which we call a _ * pauli embedding*_. given a four real dimensional vector space @xmath156 ( modeling the spacetime tangent vectors at a spacetime point , or the momentum vectors of momentum space ) , a linear injection @xmath157 may be taken , which we call a _ * pauli injection*_. the composition of a pauli embedding and a pauli injection shall be called _ * pauli map * _ , which is a @xmath158 linear map .
if we take such a pauli map @xmath159 , in ordinary two - spinor calculus it corresponds to @xmath160 in penrose abstract indices@xcite , where @xmath161 corresponds to a @xmath162 index , while @xmath163 and @xmath164 are spinor indices . keeping these in mind
, one can see that the object @xmath165 is a lorentz metric on @xmath156 , because of remark [ remmetric ] ( dirac adjoint , dirac gamma map ) let us fix a pauli map @xmath166 and a positive maximal form @xmath151 .
the conjugate - linear map @xmath167 is called the * dirac adjoint*. the linear map @xmath168 is called the * dirac gamma map*. [ clifforddef ] in order to see the properties of the dirac adjoint and the dirac gamma map , the following important subspaces of @xmath56 are introduced : let @xmath169 and @xmath170 , where the @xmath78-grading is understood to be the one subordinate to the pauli map @xmath159 , or equivalently , the one used in the underlying coproduct . clearly , these are four complex dimensional subspaces of @xmath56 with @xmath171 and @xmath172 , @xmath173 .
direct calculations show that @xmath174 becomes non - degenerate and @xmath175 satisfies the clifford relations @xmath176 over @xmath177 and @xmath178 .
these subspaces , however , are not preserved by the dressing transformations @xmath54 , but are transformed to other similarly behaving embedded dirac bispinor spaces in @xmath56 .
this embedding is illustrated in fig.[figcliffordembedding ] . and
@xmath178 into the spin algebra @xmath56 , given a pauli mapping @xmath159 .
one unit box depicts one complex dimension .
shaded regions indicate the subspaces @xmath177 and @xmath178 , respectively . ] finally , the relation of @xmath179-adjoining to the usual hilbert scalar product adjoining is given , which is used for identification of antiparticle creation operators to particle annihilation operators in a traditional qft setting .
let @xmath180 be a future directed timelike or null vector in terms of @xmath181 , modeling a momentum vector on mass shell .
then , the bilinear form @xmath182 is a positive semidefinite @xmath64-invariant inner product .
given a coproduct @xmath137 , this induces a scalar product on @xmath56 , and the adjoining operator @xmath183 with respect to that is the usual hilbert adjoining identifying antiparticle creation operators and particle annihilation operators at a fixed momentum @xmath180 .
since it is momentum dependent ( has a parameter @xmath184 ) , and is not @xmath64-invariant , it is not practical to consider @xmath183 as basic ingredient of the internal degrees of freedom , encoded in the mathematical structure of @xmath56 : it is rather constructed as a derived quantity , as shown .
also in this approach , the problematics of normal ordering does not appear : all polynomial expressions of @xmath56 are automatically normal ordered , by construction .
a finite dimensional algebra was presented and its group of automorphisms was studied .
the constructed algebra can physically be thought of as the creation operator algebra of a formal quantum field theory at fixed momentum of a spin 1/2 particle and its antiparticle .
it was shown that the essential part of the pertinent automorphism group is basically the conformal lorentz group .
on the other hand , the remaining normal subgroup can be thought of as `` dressing transformations '' making `` dressed '' states from pure one - particle states .
the proposed construction may be used in construction of quantum field theories on a non - perturbative basis : our approach is different than that of usual approach when the one - particle theory is defined , and then a fock space of multiparticle states is built on top of that .
in addition , a similar approach could be used in gut attempts , as it provides a possibility to circumvent coleman - mandula theorem : the non - semisimpleness of the pertinent symmetry group could allow to have the spacetime and internal symmetries to be connected via the non - semisimple part , without introducing susy .
an interesting idea would be to study these results in the light of @xcite , which provides an @xmath185 unification attempt , also using an algebra automorphism group .
i would like to thank mtys domokos and richrd rimnyi for valuable discussion and expert s opinion on the algebra theory content of the paper , furthermore to pter vecsernys for ideas related to the hopf algebra context .
i would like to thank tams matolcsi , pter vn , tams flp and dezs varga for motivations and discussion on the physical context of the presented idea .
i would like to thank jlia nyri for inviting the submission of this manuscript .
this work was supported in part by the momentum ( `` lendlet '' ) program of the hungarian academy of sciences , as well as the jnos bolyai research scholarship of the hungarian academy of sciences .
v. v. bavula , _ the jacobian map , the jacobian group and the group of automorphisms of the grassmann algebra _ , lecture note for talk held on `` automorphisms and homology in noncommutative algebraic geometry '' cnrs workshop , ` arxiv : math/0703352 ` , 2007 . | in this paper a finite dimensional unital associative algebra is presented , and its group of algebra automorphisms is detailed .
the studied algebra can physically be understood as the creation operator algebra in a formal quantum field theory at fixed momentum for a spin 1/2 particle along with its antiparticle .
it is shown that the essential part of the corresponding automorphism group can naturally be related to the conformal lorentz group .
in addition , the non - semisimple part of the automorphism group can be understood as `` dressing '' of the pure one - particle states .
the studied mathematical structure may help in constructing quantum field theories in a non - perturbative manner .
in addition , it provides a simple example of circumventing coleman - mandula theorem using non - semisimple groups , without susy . |
the first generation of stars so - called population iii stars ushered the universe out of the cosmic dark ages and brought the first light in the cosmos .
they are presumed to be assembled in dark matter halos of @xmath8 at @xmath9 @xcite and influenced the subsequent structure formation via mechanical , chemical and radiative feedback .
pop iii stars enriched the intergalactic medium with metals and led to the second generation of stars known as pop ii stars . according to the hierarchical paradigm of structure formation
, the first galaxies are formed in massive primordial halos of @xmath10 at about @xmath11 most likely hosting both stellar populations . during the epoch of reionization ,
radiation emitted by the pop ii stars not only photo - ionized the gas but also photo - dissociated @xmath12 and hd molecules .
the molecular hydrogen is the only coolant in primordial gas which can bring the gas temperature down to few hundred k and may induce star formation . in the presence of an intense uv flux ,
the formation of @xmath13 remains suppressed ( due to photo - detachment of @xmath0 which is the main pathway of @xmath13 formation ) in pristine halos . under these conditions , cooling only proceeds via lyman alpha radiation in massive primordial halos of @xmath10 and massive objects are expected to form and are known as direct collapse black holes ( dcbhs ) .
the formation of dcbhs requires a critical strength of uv flux above which halos remain @xmath13 free @xcite .
such a uv flux can only exist in the close vicinity of star forming galaxies @xcite and the mass scales of the resulting stars have been explored in recent studies . in fact , high resolution numerical simulations show that supermassive stars of @xmath14 form in the presence of strong uv flux @xcite and agree with theoretical predictions . employing a cosmological framework following the formation and accretion of the resulting supermassive objects , ferrara et al .
( 2014 ) have derived detailed predictions on the mass function of the first high - mass black holes . on the other hand , moderate strengths of the background uv flux
may occur more often during the epoch of reionization without requiring the presence of nearby sources . in the ubiquity of such moderate uv fluxes ,
the formation of @xmath13 does occur and may lead to star formation .
particularly , during the epoch of reionization the dominant background flux is emitted by pop ii stars . the strength of such uv flux to dissociate @xmath13 formation was quantified in a recent study by @xcite and found that @xmath15 ranges from 400 - 700 higher than previous estimates .
therefore , halos exposed below @xmath15 are expected to be more abundant , for @xmath16 the fraction of halos is seven orders of magnitude higher compared to the @xmath17 ( see figure c1 of * ? ? ?
thus it is desirable to explore the typical mass scales of stars forming in halos illuminated by moderate uv fluxes emitted by pop ii stars .
@xcite have performed three dimensional simulations to study the fragmentation in massive halos irradiated by lyman - werner flux of strength @xmath18 for @xmath19 and found that a dense turbulent core of 10@xmath20 m@xmath7 forms in the center of the halo . in this study , we explore the mass scales of stars formed under the moderate strengths of the background uv flux . to achieve this goal
, we perform three dimensional cosmological simulations for two halos of a few times @xmath21 and vary the strength of uv flux ( hereafter called @xmath3 , i.e. uv flux with energy below 13.6 ev ) .
we employed a jeans resolution of 32 cells during the course of simulations and make use of sink particles to follow the evolution for 10 , 000 years after reaching the maximum refinement level in simulations .
the main objective of this work is to determine the mass scale of stars for a moderate uv radiation field .
this has potential implications for the formation of intermediate mass black holes in stellar clusters at earlier cosmic times .
we note that the expected ionizing uv feedback by these protostars is still weak for accretion rates higher than 0.1 m@xmath7/yr .
the outline of this article is as follows . in section 2
, we describe our numerical methodology and give a brief overview of chemical network .
we present our main results in section 3 and confer our conclusions in section 4 .
we employed the open source code enzo to perform three dimensional cosmological simulations following the collapse in massive primordial halos @xcite .
enzo is an adaptive mesh refinement , parallel , eulerian code .
it makes use of the piece - wise parabolic method to solve the hydrodynamical equations .
the dark matter dynamics is solved by using the particle - mesh technique .
our simulations are started with cosmological initial conditions at @xmath22 .
we first run simulations with a uniform grid resolution of @xmath23 cells and select the most massive dm halos in a computational domain with comoving size of 1 @xmath24 .
the simulations are commenced with two additional nested refinement levels each with a resolution of @xmath23 cells in addition to a top grid resolution of @xmath23 cells and are centered on the most massive halo .
we employed 5767168 particles to solve the dark matter dynamics . to follow the collapse of a halo additional 18 levels of refinement
are applied during the course of simulations with a fixed jeans resolution of 32 cells .
we further make use of sink particles to follow the accretion for 10,000 years after the formation of the first sink .
detailed discussions about the sinks particles and the simulation setup can be found in our previous studies @xcite . to follow the evolution of chemical species , the rate equations of the following species @xmath25
, @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath0 , @xmath13 , @xmath31 are solved in our cosmological simulations .
we used publicly available krome package @xcite to compute the evolution of chemical and thermal processes . the chemical network and
the microphysics employed here are the same as in @xcite , including the self - shielding model from @xcite .
we have not included hd chemistry in our simulation as hd is very fragile molecule and even weaker flux such as @xmath32 is found to be sufficient to dissociate it .
the same was observed in simulations of @xcite . in this study ,
simulations are performed for two distinct halos by varying the strength of uv flux below the lyman limit .
the properties of simulated halos and the strength of uv flux in units of @xmath3 are listed in table [ table1 ] .
ccccccc model & @xmath3 & mass & redshift & spin parameter & sink particle masses + no & in units of j@xmath33 & @xmath34 & z & @xmath35 & @xmath34 + + a & & & & 0.034 & + & 10 & @xmath36 & 10.98 & & 1461 , 231 + & 100 & @xmath37 & 10.63 & & 7337 + & 500 & @xmath38 & 10.60 & & 22739 + b & & & & 0.03 & + & 10 & @xmath39 & 11.73 & & 623 + & 100 & @xmath40 & 11.38 & & 2592 + & 500 & @xmath40 & 11.20 & & 24785 + [ table1 ] c c & [ cols="^ " , ] c c &
in total , we have performed six cosmological simulations for two different halos of a few times @xmath21 and varied the strength of @xmath3 ( i.e. 10 , 100 and 500 ) .
our results show that initially the gas is heated up to @xmath41 k in the presence of the background uv flux and then cools by lyman alpha cooling . depending on the strength of the uv flux
, @xmath13 formation takes place and cooling due to the molecular hydrogen becomes effective . for @xmath42 10 , 100 ,
the central temperature in the halo is about 1000 k while for @xmath43 the temperature in the core is a few thousand k as shown in figure [ fig1 ] .
it is also found that for a weaker flux ( i.e. @xmath42 10 ) , @xmath13 cooling kicks in at densities around @xmath44 while for @xmath42 100 it becomes effective around @xmath45 .
for the strongest flux , it starts at even higher densities of about @xmath46 .
the density structure for various strengths of the uv field is shown in figure [ fig2 ] .
it can be seen that for @xmath42 10 the gas in the halos becomes more clumpy and indicates the possibility to form multiple clumps .
these clumps are formed by locally induced @xmath13 cooling which lowers the gas temperature down to a few hundred k as shown in figures [ fig3 ] and [ fig4 ] . for @xmath42 10 ,
the clumps with no sinks are gravitational unbound and have masses of the order of few solar masses . for @xmath42 100 ,
the halos have less structure and the number of cold clumps is reduced .
this is due to the fact that the overall temperature in the halo is higher particularly in the outskirts which suppresses fragmentation .
similarly , the fraction of molecular hydrogen is significantly reduced in the surrounding of the halo . for the strongest flux ( @xmath42 500 )
, there is no sign of fragmentation as the central temperature in the halos is a few thousand k and in one halo later approaches an isothermal state due to the temperature dependence of the @xmath13 collisional dissociation rate @xcite .
the fraction of molecular hydrogen is significantly lower and is limited to only core of the halos . to further quantify the masses of the clumps formed in our simulated halos , we employed sink particles which represent the protostars in our simulations and followed the evolution for many free - fall times .
the masses of the sinks are listed in table [ table1 ] .
it is found that for @xmath42 10 , the masses of the sinks are a few hundred to about 1500 m@xmath7 . in halo a two sinks are formed . for @xmath42 100 ,
only one massive sink is formed in each halo and their masses are 7400 and 2500 m@xmath7 .
the masses of the sinks are above 22 , 000 solar masses for the strongest flux case and a single sink is formed per halo .
overall , our results suggest that fragmentation in the halos is reduced by increasing the strength of background uv flux according to the theoretical expectations . for @xmath42 10 and 100 , although the central temperatures are quite similar the masses are higher by a factor of a few for @xmath42 100 .
this is because of the warmer gas in the surroundings of the halo which leads to higher accretion rates and consequently higher sink masses at the end of our simulations .
it is found that the masses of the sinks in halo b are about a factor of two lower compared to halo a. overall , the average density in halo b is lower and has a higher rotational support .
this leads to lower accretion rates in halo b and consequently lower sink masses .
the physical properties of the halos are depicted in figures [ fig5 ] and [ fig6 ] .
the maximum density in the halos is @xmath47 .
the density profile follow an r@xmath48 behavior at larger radii for almost all fluxes and becomes flat in the core of the halo corresponding to the jeans length .
the deviations from this behavior are observed for weaker values of @xmath3 due to the additional substructure inside the halo .
the difference between weaker and strongest flux are more prominent in halo b as it collapses almost isothermally for @xmath43 .
the overall density in halo b is lower by a factor of a few compared to the halo a around 10 pc . for @xmath49 , cooling due
to @xmath13 becomes effective in the central 10 pc of the halo . on the other hand for @xmath18
it is limited to the central pc of the halo while for @xmath43 cooling due to the molecular hydrogen is not very effective due to its lower abundance .
the impact of different thermal evolutions is also reflected in the infall velocities of the halos .
higher temperatures lead to higher infall velocities and also indicate lower molecular hydrogen fractions in those halos .
similarly lower infall velocities are observed for the weaker fluxes indicating that the halo is in the molecular cooling phase .
halo b has lower radial infall velocities compared to halo a. large accretion rates of about 0.1 m@xmath7/yr are observed .
the decline in accretion rates towards the center is due to the increasing pressure support within the jeans length .
it is further found that the accretion rates are slightly higher for stronger fluxes and are lower in halo b compared to halo a. the enclosed mass in the halos is higher for stronger fluxes .
this is due to the higher infall velocities and higher accretion rates for these cases .
it is also found that halo b has higher rotational support .
we further noted that the accretions rates increased during the course of simulations and are shown for a representative case in figure [ fig7 ] . we show the expected stellar masses for various values of @xmath3 in figure [ fig7 ] .
it is observed that the stars formed in the presence of uv feedback are very massive typically above one thousand solar masses .
this figure shows an increase in the stellar masses with the strength of the uv flux , i.e. the higher the uv flux the more massive the star .
fragmentation may occur at larger scales but may not be able to prevent the formation of massive stars .
we expect that these massive to supermassive stars are the potential candidates for the formation of intermediate to supermassive black holes .
in this article , we explored fragmentation in massive primordial halos of a few times @xmath1 m@xmath7 irradiated by different uv fluxes below the lyman limit . to achieve this goal , we performed high resolution cosmological simulations for two distinct halos collapsing at @xmath5 and varied the strength of the uv flux from 10 - 500 in units of @xmath3 .
we exploited the adaptive mesh refinement technique to follow the collapse of the halo by employing additional 18 dynamical refinement levels during the course of simulations which yields an effective resolution of about 100 au . to further follow the evolution for longer time scales , sink particles representing protostars
were employed .
a fixed jeans resolution of 32 cells were mandated during the entire course of simulations .
our findings show that the formation of @xmath13 gets delayed and no strong fragmentation occurs in the halos illuminated by the uv flux of 10 - 500 in units of @xmath3 . for @xmath49 , in our simulated halos
binary or multiple systems may likely form but for higher values of @xmath3 only one single object is formed . at the end of our simulations , massive protostars of a few 100 - 10 , 000 solar masses are formed and large accretion rates of about @xmath50 m@xmath7/yr are observed .
these relatively high accretion rates distinguish them from normal star formation mode in minihalos .
these massive stars formed in the early universe are the potential candidates for the formation of intermediate mass black holes .
@xmath13 cooling reaches local thermal equilibrium at densities higher than @xmath51 and the cooling rate scales as density while compressional heating varies with @xmath52 .
so the expected polytropic index stiffens ( i.e. @xmath53 @xmath541 ) and suppresses fragmentation @xcite .
of course , the possibility of fragmentation at higher density can not be completely ruled out and low mass stars may form in addition to the most massive object .
@xcite performed similar simulations for a halo of similar mass irradiated by the background uv flux emitted by population iii stars .
they also found that the gas self - shields from the uv flux and a dense core of @xmath55 m@xmath7 develops .
they further used sink particles to follow the simulations for longer times and found that a massive sink of about 1000 solar masses is formed .
their study is comparable to @xmath49 in our simulations due to the background uv flux emitted by two different stellar populations .
they have employed radiation spectrum with @xmath56 k in their calculations which mainly dissociates the molecular hydrogen via the solomon process and @xmath13 self - shielding is more important in contrast to our study .
in particular , we compare the accretion rates at the same times in both simulations and our results appear rather consistent .
we note however that there is no one - to - one correspondence , as the thermal evolution does not precisely match and our halo is somewhat warmer on larger scales .
in addition , there are of course fluctuations from halo to halo , so some variation between the halos is expected , as we also see here when comparing halos a and b. to further explore the possible fragmentation at higher densities , we in total added 27 refinement levels which yield resolution down to sub au scales ( 0.25 au ) and the peak densities of @xmath57 .
we however do not see any indication of further fragmentation for these runs .
such verification further confirms that even if fragmentation occurs at later stages the mass of central object would not change significantly .
our results for @xmath49 are in agreement with @xcite who simulated about 100 minihalos and found that even massive stars of thousand solar masses can form in the presence of similar mass accretion rates that found in this study .
we also note that our simulations do not include the ionizing uv feedback by the protostar .
as long as the accretion rates remain higher than @xmath58 m@xmath7/yr , no strong feedback is expected .
ionizing feedback may kick in once the accretion rates drop more significantly , potentially limiting the growth of the resulting objects .
we also assumed that halos are metal free and given the patchy distribution of the metal such halos may exist down to z=6 @xcite .
however , if metal enrichment takes places in these halos it will lead to fragmentation and most likely star cluster of low mass stars @xcite .
even a small amount of dust , i.e. @xmath59 is sufficient to boost the h@xmath60 fraction and induce fragmentation .
in fact , recent observations suggest the potential presence of intermediate mass black holes in globular clusters @xcite . from the results given here , we deduce that massive to supermassive stars may be more common than previously expected , and can also form for radiative backgrounds @xmath61 . based on the recent estimates of @xcite for the expected density of halos for various uv field strengths ,
the expected fraction of halos exposed to @xmath61 is few orders of magnitude higher .
we expect such stars to produce a considerable amount of uv feedback once they are on the main sequence , thus considerably contributing to the ionization of their environment , and perhaps also to the epoch of reionization . at the end of their lives
, they may collapse to an intermediate mass black hole , and the x - rays released during their formation may contribute to establish a global background @xcite .
the simulations described in this work were performed using the enzo code , developed by the laboratory for computational astrophysics at the university of california in san diego ( http://lca.ucsd.edu ) .
we acknowledge research funding by deutsche forschungsgemeinschaft ( dfg ) under grant sfb @xmath62 ( project a12 ) and computing time from hlrn under project nip00029 .
drgs and sb thank the dfg for funding via the schwerpunktprogram spp 1573 `` physics of the interstellar medium '' ( grant schl @xmath63 ) .
the simulation results are analyzed using the visualization toolkit for astrophysical data yt @xcite . | radiative feedback from populations ii stars played a vital role in early structure formation .
particularly , photons below the lyman limit can escape the star forming regions and produce a background ultraviolet ( uv ) flux which consequently may influence the pristine halos far away from the radiation sources .
these photons can quench the formation of molecular hydrogen by photo - detachment of @xmath0 . in this study
, we explore the impact of such uv radiation on fragmentation in massive primordial halos of a few times @xmath1 m@xmath2 . to accomplish this goal
, we perform high resolution cosmological simulations for two distinct halos and vary the strength of the impinging background uv field in units of @xmath3 .
we further make use of sink particles to follow the evolution for 10,000 years after reaching the maximum refinement level .
no vigorous fragmentation is observed in uv illuminated halos while the accretion rate changes according to the thermal properties .
our findings show that a few 100 - 10 , 000 solar mass protostars are formed when halos are irradiated by @xmath4 at @xmath5 and suggest a strong relation between the strength of uv flux and mass of a protostar .
this mode of star formation is quite different from minihalos , as higher accretion rates of about @xmath6 m@xmath7/yr are observed by the end of our simulations .
the resulting massive stars are the potential cradles for the formation of intermediate mass black holes at earlier cosmic times and contribute to the formation of a global x - ray background . |
several learning problems in modern cyber - physical network systems involve a large number of very - high - dimensional input data .
the related research areas go under the names of _ big - data analytics _ or _ big - data classification_. from an optimization point of view , the problems arising in this area involve a large number of constraints and/or local cost functions typically distributed among computing nodes communicating asynchronously and unreliably .
an additional challenge arising in big - data classification problems is that not only the number of constraints and local cost functions is large , but also the dimension of the decision variable is big and may depend on the number of nodes in the network .
we organize the literature in two parts .
first , we point out some recent works focusing the attention on _ big - data optimization _ problems ,
i.e. , problems in which all the data of the optimization problem are big and can not be handled using standard approaches from sequential or even parallel optimization .
the survey paper @xcite reviews recent advances in convex optimization algorithms for big - data , which aim to reduce the computational , storage , and communications bottlenecks .
the role of parallel and distributed computation frameworks is highlighted . in @xcite big - data , possibly non - convex ,
optimization problems are approached by means of a decomposition framework based on successive approximations of the cost function . in @xcite dictionary
learning tasks motivate the development of non - convex and non - smooth optimization algorithms in a big - data context .
the paper develops an online learning framework by jointly leveraging the stochastic approximation paradigm with first - order acceleration schemes .
second , we review distributed optimization algorithms applied to learning problems and highlight their limitations when dealing with big - data problems .
an early reference on peer - to - peer training of support vector machines is @xcite .
a distributed training mechanism is proposed in which multiple servers compute the optimal solution by exchanging support vectors over a fixed directed graph .
the work is a first successful attempt to solve svm problems over networks .
however , the local memory and computation at each node does not scale with the problem and data sizes and the graph is time - invariant . in @xcite
a distributed alternating direction method of multipliers ( admm ) is proposed to solve a linear svm training problem , while in @xcite the same problem is solved by means of a random projected gradient algorithm . both the algorithms are proven to solve the centralized problem ( i.e. , all the nodes reach a consensus on the global solution ) , but again show some limitations : the graph topology must be ( fixed , @xcite , and ) undirected , and the algorithms do not scale with the dimension of the training vector space . in @xcite a survey on admm algorithms applied to statistical learning problems is given .
in @xcite the problem of exchanging only those measurements that are most informative in a network svm problem is investigated . for separable problems
an algorithm is provided to determine if an element in the training set can become a support vector .
the distributed optimization algorithm proposed in @xcite solves part of these problems : local memory is scalable and communication can be directed and asynchronous .
however , the dimension of the training vectors is still an issue .
the core - set idea used in this paper was introduced in @xcite as a building block for clustering , and refined in @xcite . in @xcite
the approach was shown to be relevant for several learning problems and the algorithm re - stated for such scenarios .
a multi - processor implementation of the core - set approach was proposed in @xcite .
however , differently from our approach , that algorithm : ( i ) is not completely distributed since it involves a coordinator , and ( ii ) does not compute a global core - set , but a larger set approximating it .
the main contribution of this paper is twofold .
first , we identify a distributed big - data optimization framework appearing in modern classification problems arising in cyber - physical network systems . in this framework
the problem is characterized by a large number of input data distributed among computing processors .
the key challenge is that the dimension of each input vector is very - high , so that standard local updates in distributed optimization can not be used . for this big - data scenario
, we identify a class of quadratic programs that model several interesting classification problems as , e.g. , training of support vector machines .
second , for this class of big - data quadratic optimization problems , we propose a distributed algorithm that solves the problem up to an arbitrary @xmath0 tolerance and scales both with the number and the dimension of the input vectors . the algorithm is based on the notion of core - set used in geometric optimization to approximate the value function of a given set of points with a smaller subset of points . from an optimization point of view , a subset of active constraints is identified , whose number depends only on the tolerance @xmath0 .
the resulting approximate solution is such that an @xmath0-relaxation of the constraints guarantees no constraint violation .
the paper is organized as follows . in section
[ sec : distrib_optim_framework ] we introduce the distributed optimization problem addressed in the paper and describe the network model .
section [ sec : distributed_svm ] motivates the problem set - up by showing a class of learning problems that can be cast in this set - up . in section
[ sec : core - set consensus ] the core - set consensus algorithm is introduced and analyzed .
finally , in section [ sec : simulations ] a numerical example is given to show the algorithm correctness .
in this section we introduce the problem set - up considered in the paper .
we recall that we will deal with optimization problems in which both the number of constraints and decision variables are `` big '' .
we consider a set of processors @xmath1 , each equipped with communication and computation capabilities .
each processor @xmath2 has knowledge of a vector @xmath3 and needs to cooperatively solve the quadratic program @xmath4 the above quadratic program is known in geometric optimization as _ minimum enclosing ball _
problem , since it computes the center of the ball with minimum radius enclosing the set of points @xmath5 . by applying standard duality arguments
, it can be shown that solving is equivalent to solving its dual @xmath6 with @xmath7 , @xmath8 .
the problem can be written in a more compact form as @xmath9 where @xmath10 \in{{\mathbb{r}}}^{d\times n}$ ] , @xmath11 is the vector with elements @xmath12 , @xmath8 , @xmath13^t \in{{\mathbb{r}}}^n$ ] and @xmath14 is meant component - wise .
we will show in the next sections that this class of quadratic programs arises in many important big - data classification problems .
each node has computation capabilities meaning that it can run a routine to solve a local optimization problem .
since the dimension @xmath15 can be big , the distributed optimization algorithm to solve problem needs to be designed so that the local routine at each node scales `` nicely '' with @xmath15 . the communication among the processors is modeled by a time - varying ,
directed graph ( digraph ) @xmath16 , where @xmath17 represents a slotted universal time , the node set @xmath18 is the set of processor identifiers , and the edge set @xmath19 characterizes the communication among the processors .
specifically , at time @xmath20 there is an edge from node @xmath2 to node @xmath21 if and only if processor @xmath2 transmits information to processor @xmath21 at time @xmath20 .
the time - varying set of outgoing ( incoming ) _ neighbors _ of node @xmath2 at time @xmath20 , i.e. , the set of nodes to ( from ) which there are edges from ( to ) @xmath2 at time @xmath20 , is denoted by @xmath22 ( @xmath23 ) .
a static digraph is said to be _ strongly connected _ if for every pair of nodes @xmath24 there exists a path of directed edges that goes from @xmath2 to @xmath21 . for the time - varying communication graph we rely on the concept of a jointly strongly connected graph .
[ ass : periodicconnectivity ] for every time instant @xmath25 , the union digraph @xmath26 is strongly connected .
it is worth noting that joint strong connectivity of the directed communication graph is a fairly weak assumption ( it just requires persistent spreading of information ) for solving a distributed optimization problem , and naturally embeds an asynchronous scenario .
we want to stress once more that in our paper all the nodes are peers , i.e. , they run the same local instance of the distributed algorithm , and no node can take any special role .
consistently , we allow nodes to be asynchronous , i.e. , nodes can perform the same computation at different speed , and communication can be unreliable and happen without a common clock ( the time @xmath20 is a universal time that does not need to be known by the nodes ) .
in this section we present a distributed set up for some fundamental classification problems and show , following @xcite , how they can be cast into the distributed quadratic programming framework introduced in the previous section .
we consider classification problems to be solved in a distributed way by a network of processors following the model in section [ sec : distrib_optim_framework ] .
each node in the network is assigned a subset of input vectors and the goal for the processors is to cooperatively agree on the optimal classifier without the help of any central coordinator .
informally , the svm training problem can be summarized as follows .
given a set of positively and negatively labeled points in a @xmath27-dimensional space , find a hyperplane separating `` positive '' and `` negative '' points with the maximal separation from all the data points .
the labeled points are commonly called _
examples _ or _ training vectors_. linear separability of the training vectors is usually a strong assumption . in many important concrete scenarios
the training data can not be separated by simply using a linear function ( a hyperplane ) . to handle the nonlinear separability ,
nonlinear kernel functions are used to map the training samples into a feature space in which the resulting features can be linearly separated .
that is , given a set of points @xmath28 in the _ input space _
they are mapped into a _ feature space _ through a function @xmath29 .
the key aspect in svm is that @xmath30 does not need to be known , but all the computations can be done through a so called _ kernel function _
@xmath31 satisfying @xmath32 .
it is worth noting that the dimension of the feature space can be much higher than the one of the input space , even infinite ( e.g. , gaussian kernels ) .
following @xcite and @xcite we will adopt the following common assumption in svm . for any @xmath33 in the input space @xmath34 with @xmath35 independent of @xmath2 .
this condition is satisfied by the most common kernel functions used in svm as , e.g. , the isotropic kernel ( e.g. , gaussian kernel ) , the dot - product kernel with normalized inputs or any normalized kernel .
for fixed @xmath36 , let @xmath37 , @xmath8 , be a set of @xmath38 feature - points with associated label @xmath39 .
the training vectors are said to be linearly separable if there exist @xmath40 and @xmath41 such that @xmath42 for all @xmath8 . the _ hard - margin svm training _
problem consists of finding the optimal hyperplane @xmath43 , @xmath44 , ( @xmath45 is a vector orthogonal to the hyperplane and @xmath46 is a bias ) that linearly separates the training vectors with maximal margin , that is , such that the distance @xmath47 is maximized . combining the above equations it follows easily that @xmath48 .
thus the svm training problem may be written as a quadratic program @xmath49 in most concrete applications the training data can not be separated without outliers ( or training errors ) . a convex program that approximates
the above problem was introduced in @xcite .
the idea is to introduce positive slack variables in order to relax the constraints and add an additional penalty in the cost function to weight them .
the resulting classification problems are known as _ soft marging problems _ and the solution is called _ soft margin hyperplane_. next , we will concentrate on a widely used soft - margin problem , the _
@xmath50-norm problem _ , which adopts a quadratic penalty function .
following @xcite , we will show that its dual version is a quadratic program with the structure of .
the @xmath50-norm optimization problem turns out to be @xmath51 solving problem is equivalent to solving the dual problem @xmath52 where @xmath53 if @xmath54 and @xmath55 otherwise .
the vector @xmath45 defining the optimal hyperplane can be written as linear combination of training vectors , @xmath56 , where @xmath57 and @xmath58 only for vectors satisfying @xmath59 .
these vectors are called _ support vectors_. support vectors are basically active constraints of the quadratic program .
now , we can notice that defining @xmath60 , it holds @xmath61 so that the constant term @xmath62 , can be added to the cost function .
thus , posing @xmath63 with @xmath64 the @xmath2th canonical vector ( e.g. , @xmath65^t$ ] ) , problem can be equivalently rewritten as @xmath66 which has exactly the same structure as problem .
it is worth noting that even if the dimension @xmath15 of the training samples in the feature space ( @xmath67 ) is small compared to the number of samples @xmath68 ( so that in problem the dimension of the decision variable is much smaller than the number of constraints ) , in the `` augmented '' soft - margin problem we have @xmath69 .
thus , in the primal problem the dimension of the decision variable is of the same order as the number of constraints .
next , we recall from @xcite that also some unsupervised soft margin classification problems can be cast into the same problem set - up of the paper . first , from @xcite and
references therein , it can be shown that problem is equivalent to the hard - margin support vector data description ( svdd ) problem .
indeed , given a kernel function @xmath31 and feature map @xmath30 , the hard - margin svdd primal problem is @xmath70 in other words , problem is simply problem in the feature space .
another unsupervised learning problem that can be cast into the problem set - up is the so called _ one - class l2 svm _ , @xcite .
given a set of unlabeled input vectors the goal is to separate outliers from normal data . from an optimization point of view , the problem can be written as problem , but with @xmath71 and @xmath72 for all @xmath8 .
thus , using the same arguments as in the previous subsection , the problem can be rewritten in the form . to conclude this motivating section
, we recall from @xcite that algorithms solving problem are important building blocks for clustering problems .
in this section we introduce the core - set consensus algorithm to solve problem ( or equivalently its dual ) in a distributed way .
we start by introducing the notion of core - set borrowed from geometric optimization and a routine from @xcite that is proven to compute an efficient core - set for , which in geometric optimization is known as _ minimum enclosing ball problem_. in the following we will a little abuse notation by denoting with @xmath73 both the @xmath74 matrix and the set of @xmath75 vectors ( or points ) of dimension @xmath15 .
let @xmath76 be a matrix of `` points '' @xmath3 , @xmath8 , ( i.e. , a matrix in which each column represents a vector in @xmath77 ) with @xmath78 and @xmath79 respectively the center and radius of the minimum enclosing ball containing the points of @xmath80 .
we say that @xmath81 is an _ @xmath0-core - set _ for the minimum enclosing ball ( meb ) problem , , if all the points of @xmath80 are at distance at most @xmath82 from the center @xmath83 of the minimum enclosing ball containing @xmath84 .
note that @xmath85 , with @xmath86 being the optimal value of .
next , we introduce the algorithm in @xcite that is proven to compute a core - set of dimension @xmath87 for the minimum enclosing ball problem . given a set of points @xmath88 , the algorithm can be initialized by choosing any subset @xmath89 of @xmath87 points .
then the algorithm evolves as follows : * select a point @xmath90 of @xmath88 farthest from the center of the minimum enclosing ball of @xmath84 ; * let @xmath91 ; * remove a point @xmath92 so that the minimum enclosing ball of the set @xmath93 is the one with largest radius ; * if the new radius is equal to the radius of minimum enclosing ball of @xmath84 , then return @xmath84 .
otherwise set @xmath94 and repeat the procedure .
more formally , the routing is described in the following table .
as before , we let @xmath95 and @xmath96 be respectively the center and radius of the minimum enclosing ball containing all the points in a set of points @xmath88 .
* function * @xmath97
@xmath98 @xmath91 @xmath99 * if * @xmath100 @xmath94 and go to step 1 * else * * return * @xmath84 * end if * it is worth pointing out once more that if @xmath101 , with @xmath80 the one in , then the @xmath102 algorithm finds an @xmath0-core - set for ( or equivalently ) . the @xmath102 algorithm will be the local routine implemented in the distributed optimization algorithm we propose in this paper . that is , each node will use the algorithm to solve a ( smaller ) local version of the main problem .
next we provide a lemma that states the results in @xcite by formally itemizing the properties of the algorithm that we will need in our distributed optimization algorithm .
[ lem : badoiu ] let @xmath103 be any point set in @xmath77
. then 1 .
@xmath88 has an @xmath0-core - set of size at most @xmath87 ; 2 .
the @xmath102 algorithm computes an @xmath0-core - set for @xmath88 in a finite number of iterations ; 3 . for any @xmath104
the radius of @xmath105 is larger than or equal to the radius of @xmath84 . statements ( i ) and ( ii ) are proven in ( * ? ? ?
* theorem 3.5 ) , while ( iii ) follows immediately by step 4 of the algorithm .
let @xmath80 be the matrix characterizing problem or consistently the set of vectors @xmath3 , @xmath8 . as stated in section [ sec : distrib_optim_framework ] , each node is assigned one input vector .
this assumption is just for clarity of presentation and can be easily removed .
in fact , the algorithm can be run even if each node is assigned more than one vector .
for this reason we denote @xmath106 the set of initial vectors , so that under the above assumption we have @xmath107 .
an informal description of the _ core - set consensus _ distributed algorithm is the following .
each node stores a candidate core - set @xmath108 , i.e. , a set of @xmath87 vectors that represent node-@xmath2 s current estimate for the core - set of @xmath80 . at each communication
round each node receives the candidate core sets ( sets of @xmath87 vectors ) from its in - neighbors and initializes its local routine to the core - set with highest value .
let @xmath109 be the set of vectors from all neighboring core - sets plus the initial vectors assigned to node @xmath2 .
the local routine at each node finds a core set ( of @xmath87 vectors ) of @xmath110 , and updates the candidate core set with the returned value .
a pseudo - code of the algorithm is given in the following table .
we assume each node can run two routines , namely @xmath111 and @xmath112 returning respectively the core - set of a given set of vectors @xmath113 ( the routine is initialized with @xmath114 ) and the value of a given core set , i.e. , the optimal value of problem with matrix @xmath115 ( squared radius of the minimum enclosing ball ) .
@xmath116 @xmath107 , @xmath117 @xmath108 @xmath118 * return * @xmath108 the algorithm works also if a larger set of vectors is assigned to each node .
only the initialization needs to be changed . if a node is assigned more than @xmath87 vectors , it will initialize @xmath108 with a random set of @xmath87 vectors .
it is worth noting that the nodes need to know the common tolerance @xmath0 ( and thus the core - set dimension @xmath87 ) to run the algorithm . to analyze the algorithm
, we associate a universal , discrete time @xmath119 to each step of the distributed algorithm evolution , i.e. , to each computation and communication round . this time @xmath20 is the one defining the time varying nature of the communication graph and , thus , the same used in assumption [ ass : periodicconnectivity ] .
we are now ready to analyze the convergence properties of the algorithm .
[ ass : non_degenerate ] given @xmath80 in , for any @xmath120 with @xmath121 , then @xmath122 .
the above assumption can be removed by using a total ordering for the choice of @xmath123 in algorithm [ alg : core - set_consensus ] .
for example , if two candidate sets @xmath124 and @xmath125 have @xmath126 , then one of the two could be uniquely chosen by using a lexicographic ordering on the vectors .
consider a network of processors with set of identifiers @xmath127 and communication graph @xmath16 , @xmath25 , satisfying assumption [ ass : periodicconnectivity ] .
suppose problem satisfies assumption [ ass : non_degenerate ] and has a minimum value @xmath128 .
then the core - set consensus algorithm ( algorithm [ alg : core - set_consensus ] ) computes an @xmath0-core - set for problem in a finite - number of communication rounds .
that is , there exists @xmath129 such that 1 .
@xmath130 for all @xmath8 and for all @xmath131 ; 2 .
@xmath132 for all @xmath8 .
we prove the statement in three steps .
first , we prove that each core set converges in a finite - number of communication rounds to a stationary set of vectors .
second , we prove that ( due to assumption [ ass : non_degenerate ] ) all the stationary core sets are equal .
third and finally , we prove that the common steady - state set is a core set for problem . to prove the first part ,
notice that by the choice of @xmath123 in algorithm [ alg : core - set_consensus ] and by lemma [ lem : badoiu ] , @xmath133 is a monotone nondecreasing function for all @xmath8 along the algorithm evolution .
thus , due to the finite possible values that @xmath108 can assume ( it is a set of @xmath87 vectors out of @xmath68 vectors ) , @xmath133 converges to a stationary value in finite time . to prove the second part , suppose that at some time @xmath129 all the @xmath133s have converged to a stationary value and that there exist at least two nodes @xmath2 and @xmath21 such that @xmath134 . without loss of generality , from assumption [ ass : periodicconnectivity ] , we can choose the two nodes so that @xmath135 , i.e. , @xmath24 is an edge in @xmath136 for some time instant @xmath137 .
but from algorithm [ alg : core - set_consensus ] at time @xmath138 node @xmath21 would choose @xmath108 to initialize its @xmath102 routine , thus leading to a contradiction . from assumption [ ass : non_degenerate ]
, it follows @xmath139 for all @xmath8 .
finally , we just need to prove that @xmath140 is a core - set for @xmath80 .
but from the properties of the @xmath102 algorithm , for each node @xmath8 , @xmath139 is a core - set for the a set of points including @xmath141 , so that @xmath140 is a core set for @xmath142 $ ] , thus concluding the proof .
a core - set @xmath84 is a set of `` active constraints '' in problem with a cost @xmath143 ( i.e. , @xmath144 ) .
clearly , some of the constraints will be violated for this value of @xmath145 , but no one will be violated for @xmath146 , with @xmath86 being the optimal value of @xmath145 .
an equivalent characterization for the core - set is that no constraint is violated if @xmath147 is relaxed to @xmath148 .
this test is easier to run , since it does not involve the computation of the optimal value and will be used in the simulations .
in this section we provide a numerical example showing the effectiveness of the proposed strategy .
we consider a network with @xmath149 nodes communicating according to a directed , time - varying graph obtained by extracting at each time - instant an erds - rnyi graph with parameter @xmath150 .
we choose a small value , so that at a given instant the graph is disconnected with high probability , but the graph turns out to be jointly connected .
we solve a quadratic program , , with @xmath151 and choose a tolerance @xmath152 so that the number of vectors in the core - set is @xmath153 . in figure
[ fig : rr_transient ] and figure [ fig : cc_transient ] the evolution of the squared - radius and center - norm of the core - sets at each node are depicted . as expected from the theoretical analysis ,
the convergence of the radius to the consensus value is monotone non - decreasing . , @xmath8 .
] , @xmath8 . ]
in this paper we have proposed a distributed algorithm to solve a special class of quadratic programs that models several classification problems .
the proposed algorithm handles problems in which not only the number of input data is large , but furthermore their dimension is big .
the resulting learning area is known as _ big - data classification_. we have proposed a distributed optimization algorithm that computes an approximate solution of the global problem .
specifically , for any chosen tolerance @xmath0 , each local node needs to store only @xmath87 active constraints , which represent a solution for the global quadratic program up to a relative tolerance @xmath0 .
future research developments include the extension of the algorithmic idea , based on core - sets , to other big - data optimization problems .
v. cevher , s. becker , and m. schmidt , `` convex optimization for big data : scalable , randomized , and parallel algorithms for big data analytics , '' _ ieee signal processing magazine _ , vol .
31 , no . 5 , pp . 3243 , 2014 .
k. slavakis and g. b. giannakis , `` online dictionary learning from big data using accelerated stochastic approximation algorithms , '' in _ 2014 ieee international conference on acoustics , speech and signal processing ( icassp ) _ , 2014 , pp .
y. lu , v. roychowdhury , and l. vandenberghe , `` distributed parallel support vector machines in strongly connected networks , '' _ ieee transactions on neural networks _ , vol .
19 , no . 7 , pp .
11671178 , 2008 .
s. boyd , n. parikh , e. chu , b. peleato , and j. eckstein , `` distributed optimization and statistical learning via the alternating direction method of multipliers , '' _ foundations and trends in machine learning _ ,
vol . 3 , no . 1 , pp . 1122 , 2011 .
d. varagnolo , s. del favero , f. dinuzzo , l. schenato , and g. pillonetto , `` finding potential support vectors in separable classification problems , '' _ ieee transactions on neural networks and learning systems _ , vol . 24 , no . 11 , pp . 17991813 , 2013 .
g. notarstefano and f. bullo , `` distributed abstract optimization via constraints consensus : theory and applications , '' _ ieee transactions on automatic control _ ,
56 , no .
10 , pp . 22472261 , october 2011 .
s. lodi , r. nanculef , and c. sartori , `` single - pass distributed learning of multi - class svms using core - sets , '' in _ proceedings of the 2010 siam international conference on data mining _ , 2010 , pp .
257268 .
s. s. keerthi , s. k. shevade , c. bhattacharyya , and k. r. murthy , `` a fast iterative nearest point algorithm for support vector machine classifier design , '' _ ieee transactions on neural networks _ ,
11 , no . 1 ,
124136 , 2000 . | a new challenge for learning algorithms in cyber - physical network systems is the distributed solution of big - data classification problems , i.e. , problems in which both the number of training samples and their dimension is high . motivated by several problem set - ups in machine learning ,
in this paper we consider a special class of quadratic optimization problems involving a `` large '' number of input data , whose dimension is `` big '' . to solve these quadratic optimization problems over peer - to - peer networks ,
we propose an asynchronous , distributed algorithm that scales with both the number and the dimension of the input data ( training samples in the classification problem ) .
the proposed distributed optimization algorithm relies on the notion of `` core - set '' which is used in geometric optimization to approximate the value function associated to a given set of points with a smaller subset of points . by computing local core - sets on a smaller version of the global problem and exchanging them with neighbors ,
the nodes reach consensus on a set of active constraints representing an approximate solution for the global quadratic program . distributed optimization , big - data optimization , support vector machine ( svm ) , machine learning , core set , asynchronous networks . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.